CN114493375A - Construction Safety Macro Evaluation System and Method - Google Patents

Abstract

The invention relates to a construction safety macro-evaluation system and method. The system includes: a data processing module, which collects data images of the construction site and performs labeling and identification processing to determine each worker and construction machinery on the construction site, and each worker on the construction site. The center point coordinates and contour coordinates corresponding to the construction machinery; the safety evaluation module determines the daily worker safety according to each worker and construction machinery at the construction site, as well as the center point coordinates and contour coordinates corresponding to each worker and the construction machinery. Unsafe behavior value, worker risk awareness value and management agility value of wearing helmets; training module, according to the unsafe behavior value, worker risk awareness value and management agility value within a preset time period, and the corresponding preset construction safety The macro evaluation value is trained by fuzzy neural network algorithm, and the construction safety macro evaluation model is obtained; the evaluation module uses the construction safety macro evaluation model to carry out the construction safety macro evaluation on the target construction site.

Description

Construction Safety Macro Evaluation System and Method

technical field

The invention relates to the technical field of data processing on construction sites, in particular to a construction safety macro evaluation system and method.

Background technique

The ubiquitous unsafe behavior on construction sites is the biggest challenge facing the construction phase of the construction industry today. According to the accident causation theory, the unsafe behavior of people and the unsafe state of things are the direct causes of asset loss and casualties. At present, the main method to reduce safety accidents on construction sites is to patrol by safety-related personnel, and to identify potential safety hazards and avoid risks by manpower observing unsafe behaviors and unsafe conditions on the site. But there are deeper system reasons behind the frequent occurrence of unsafe behaviors and unsafe states. For safety issues in construction, the researchers found that there are three key elements worth paying attention to, namely safety force, safety culture and safety behavior.

Security decision-making is defined as the interaction process between decision-makers and subordinates to achieve organizational security goals. In this process, decision-makers can exert influence on followers by relying on the combined action of organizational and personal factors. This definition is widely accepted by researchers. Security decision-making power is regarded as a sub-concept of decision-making power. Numerous studies have demonstrated that positive transformational or contractual safety decision-making can influence lower-level managers or workers at the lower level: subjectively achieve their good perception of higher-level managers' safety attitudes and commitment to safety, shape correct safety values and attitudes toward safety goals Positive recognition; objectively motivate subordinates or workers to avoid unsafe production activities by setting incentives and punishments based on safety performance and regulatory procedures.

Safety culture is defined as the views and beliefs of all members of an organization about the risks, accidents, and health issues that the organization faces during production. Safety culture is a sub-concept of organizational culture, subordinate to organizational culture, and its main connotations include beliefs, attitudes and values. Safety culture can be condensed as "the combination of safety beliefs and values built on people, things and things in the collective, reflecting the views and behavior habits of life and safety agreed by all the people in the collective". The safety culture of construction projects is manifested in three aspects: workers' behavior, workers' subjective perceptions, and the objective environment of workers' operations.

Safety behaviors are defined as management behaviors and manipulation behaviors made by people in the process of interacting with people, things, and things, driven by safety goals or constrained by safety requirements. Traditional building safety studies have shown that worker unsafe behavior is the main cause of accidents. A large number of existing studies have classified and identified unsafe behaviors on construction sites, and many studies have summarized the categories of typical worker safety behaviors (or unsafe behaviors) on construction sites from different perspectives. The most common construction site unsafe behaviors are excessive proximity to hazards and incorrect wearing of personal protective equipment.

By studying the connotation and interaction of safety decision-making power, safety culture and safety behavior, a building safety decision-making power-culture-behavior (LCB) method with decision-driven cultural development and behavior control as the core is proposed. The LCB approach emphasizes the role of safety decision-making, not only directly reducing unsafe behaviors, but also fundamentally changing the causes of unsafe behaviors through safety culture development, ultimately achieving the goal of sustainable reductions in unsafe behaviors and accident prevention.

SUMMARY OF THE INVENTION

In order to overcome the problems existing in the related art, the present invention provides a macro-evaluation system and method for construction safety based on the LCB theory.

According to a first aspect of the embodiments of the present invention, a construction safety macro evaluation system is provided, including:

The data processing module is used to collect data images of the construction site and perform labeling and identification processing to determine each worker and each construction machine on the construction site, as well as the corresponding center point coordinates and contours of each worker and each construction machine coordinate;

The safety evaluation module is used to determine the daily unsafe behavior of workers without helmets according to each worker and each construction machine on the construction site, as well as the corresponding center point coordinates and outline coordinates of each worker and each construction machine value, worker risk awareness value and management agility value;

The training module is used for training with fuzzy neural network algorithm according to the unsafe behavior value, worker risk awareness value and management agility value within a preset time period, and the corresponding preset macroscopic evaluation value of construction safety to obtain the macroscopic construction safety value. evaluation model;

The assessment module is used to perform macro-assessment of construction safety on the target construction site using the macro-assessment model of construction safety.

In one embodiment, preferably, the data processing module includes:

The data acquisition unit is used to collect the data images of the construction site through the camera device;

Calibration unit for daily calibration of the camera;

The marking unit is used to mark the special roads for construction machinery and high-risk areas for construction operations according to the received marking instructions before daily construction;

The identification unit is used to train workers, mobile construction machinery, and tower cranes based on the Mask-RCNN and Deepsort algorithms through pre-collected and labeled construction images, and to identify workers, mobile, and mobile in the data images of the construction site in minutes The outline coordinates of mobile construction machinery and tower cranes are collected, and each identification object is tracked; in the database, the total number of identified workers, mobile construction machinery and tower cranes is counted in minutes; When contouring, identify whether the worker wears a helmet, and count the number of the corresponding number of not wearing a helmet, number each identified worker, mobile construction machinery and tower crane, and set the corresponding contour coordinates. Synchronized with the center point coordinates and stored in the corresponding data sets of workers, mobile construction machinery and tower cranes.

In one embodiment, preferably, the security evaluation module is used to:

Store the number of people without helmets in the database as a time series array UB _i , i = 0,1,2... n -1, let Wi = UB _{i +1} - UB _i , W ₀ = 0, where _n is the time length of the time series. If Wi < 0, let Wi ₌ 0, construct an array Wi _, and use the following first calculation formula to calculate the daily unsafe behavior value:

Among them, Ni is the total number of workers on the construction site at time _i ;

According to the corresponding center point coordinates and contour coordinates of each worker and each construction machine in the database, count the number of workers exposed to work in the risk area at each moment, and use the following second calculation formula to calculate the risk awareness score of the first worker:

表示每日暴露在风险区作业的工人数量值，即第一工人风险意识分值；Among them, NW _i represents the number of workers exposed to the risk area at time i, N _i is the total number of workers on the construction site at time i,

Represents the value of the number of workers exposed to the risk area every day, that is, the risk awareness score of the first worker;

According to the corresponding center point coordinates and contour coordinates of each worker and each construction machine in the database, count the number of workers exposed to the special road for workers and machinery at each moment, and use the following third calculation formula to calculate the second worker's risk awareness score :

Among them, NRW _i represents the number of workers exposed on the special road for workers and machinery at time i, Ni is the total number of workers on the construction site at time i, and N _rw represents _the daily value of the number of workers exposed on the special road for workers and machinery, that is, the second The worker's risk awareness score, wherein the worker's risk awareness value includes the first worker's risk awareness score and the second worker's risk awareness score;

The daily time T _wh of workers not wearing helmets, the time T _wr of workers entering the special road for construction machinery, the time T _r that the special road for construction machinery is occupied, the unit of time is minutes, and the final output management agility value T = ( T _wh , T _wr , T _r ).

In one embodiment, preferably, the training module includes:

The normalization unit is used to use the MinMaxScaler normalization algorithm to normalize the unsafe behavior value, worker risk awareness value, and management agility value, and obtain the normalized unsafe behavior value and worker risk awareness value. and managing agility values

The conversion unit is used to convert the unsafe behavior value, worker risk awareness value and management agility value into values in the range of 0-1, and obtain the macro-evaluation value of construction safety preset by experts on the construction site;

The training unit is used to use the normalized unsafe behavior value, worker risk awareness value and management agility value as the input x of the fuzzy neural network algorithm, and the corresponding preset macroscopic evaluation value of construction safety as the fuzzy neural network algorithm. Output y, carry out training, and obtain the macroscopic evaluation model of construction safety;

Among them, the macro-evaluation model of construction safety includes fuzzy layer, rule layer, regularization layer, secondary fuzzy calculation layer and defuzzification layer;

The fuzzy layer fuzzifies each input through three membership functions, calculates the corresponding membership value, and outputs it to the rule layer;

The rule layer multiplies each membership value to output the activation degree of each rule to the regularization layer;

The regularization layer performs regularization calculation on the activation degree of each rule to output the regularized activation degree to the secondary fuzzy calculation layer;

The secondary fuzzy calculation layer calculates the new membership value through the regularized activation degree to output to the defuzzification layer;

The defuzzification layer remaps the new membership value back to the exact value in the set of lecture points to output the construction safety macro assessment value.

In one embodiment, preferably, the evaluation module is used to:

Collect target data images of the target construction site and perform labeling and identification processing to determine each worker and each construction machine at the target construction site, as well as the corresponding center point coordinates and outline coordinates of each worker and each construction machine;

According to each worker and each construction machine at the target construction site, as well as the corresponding center point coordinates and outline coordinates of each worker and each construction machine, determine the daily target unsafe behavior value of workers without helmets, target workers Risk awareness value and target management agility value;

According to the target unsafe behavior value, the target worker's risk awareness value, the target management agility value and the construction safety macro evaluation model, the target construction safety macro evaluation value corresponding to the target construction site is calculated.

According to a second aspect of the embodiments of the present invention, a construction safety macro evaluation method is provided for use in a construction safety macro evaluation system, the method comprising:

Collect data images of the construction site and perform labeling and identification processing to determine each worker and each construction machine on the construction site, as well as the corresponding center point coordinates and outline coordinates of each worker and each construction machine;

According to each worker and each construction machine on the construction site, as well as the corresponding center point coordinates and outline coordinates of each worker and each construction machine, the daily unsafe behavior value and worker risk awareness value of workers without helmets are determined and managing agility values;

According to the unsafe behavior value, worker risk awareness value and management agility value within the preset time period, and the corresponding preset construction safety macro evaluation value, the fuzzy neural network algorithm is used for training, and the construction safety macro evaluation model is obtained;

Use the construction safety macro-evaluation model to perform a construction safety macro-evaluation on the target construction site.

In one embodiment, preferably, a data image of the construction site is collected and marked and identified, so as to determine each worker and each construction machine on the construction site, as well as the corresponding center point of each worker and each construction machine Coordinates and contour coordinates, including:

Collect the data images of the construction site through the camera device;

daily calibration of the camera;

Before daily construction, according to the received marking instructions, mark the dedicated roads for construction machinery and high-risk areas for construction operations;

Through pre-collected and labeled construction images, workers, mobile construction machinery, and tower cranes are trained based on the Mask-RCNN and Deepsort algorithms, and workers, mobile construction machinery, and tower cranes are identified in the data images of the construction site in minutes. The coordinates of the contours of the identified objects are tracked, and each identified object is tracked; in the database, the total number of identified workers, mobile construction machinery and tower cranes is counted in minutes; among them, when the contours of the workers are identified, the workers are identified. Whether to wear a safety helmet, and the corresponding statistics of the number of not wearing a safety helmet, number each worker, mobile construction machinery and tower crane identified, and store their corresponding contour coordinates and center point coordinates in the corresponding moment. A dataset of workers, mobile construction machinery and tower cranes.

In one embodiment, preferably, according to each worker and each construction machine on the construction site, and the corresponding center point coordinates and outline coordinates of each worker and each construction machine, the daily number of workers not wearing helmets is determined. Unsafe Behavior Values, Worker Risk Awareness Values, and Management Agility Values, including:

Store the number of people without helmets in the database as a time series array UB _i , i = 0,1,2... n -1, let Wi = UB _{i +1} - UB _i , W ₀ = 0, where _n is the time length of the time series. If Wi < 0, let Wi ₌ 0, construct an array Wi _, and use the following first calculation formula to calculate the daily unsafe behavior value:

Among them, Ni is the total number of workers on the construction site at time _i ;

According to the corresponding center point coordinates and contour coordinates of each worker and each construction machine in the database, count the number of workers exposed to work in the risk area at each moment, and use the following second calculation formula to calculate the risk awareness score of the first worker:

表示每日暴露在风险区作业的工人数量值，即第一工人风险意识分值；Among them, NW _i represents the number of workers exposed to the risk area at time i, N _i is the total number of workers on the construction site at time i,

Represents the value of the number of workers exposed to the risk area every day, that is, the risk awareness score of the first worker;

According to the corresponding center point coordinates and contour coordinates of each worker and each construction machine in the database, count the number of workers exposed to the special road for workers and machinery at each moment, and use the following third calculation formula to calculate the second worker's risk awareness score :

Among them, NRW _i represents the number of workers exposed on the special road for workers and machinery at time i, Ni is the total number of workers on the construction site at time i, and N _rw represents _the daily value of the number of workers exposed on the special road for workers and machinery, that is, the second The worker's risk awareness score, wherein the worker's risk awareness value includes the first worker's risk awareness score and the second worker's risk awareness score;

The daily time T _wh of workers not wearing helmets, the time T _wr of workers entering the special road for construction machinery, the time T _r that the special road for construction machinery is occupied, the unit of time is minutes, and the final output management agility value T = ( T _wh , T _wr , T _r ).

In one embodiment, preferably, according to the unsafe behavior value, worker risk awareness value and management agility value within a preset time period, and the corresponding preset macroscopic evaluation value of construction safety, a fuzzy neural network algorithm is used for training, Obtain a construction safety macro assessment model, including:

The MinMaxScaler normalization algorithm is used to normalize the unsafe behavior value, worker risk awareness value and management agility value to obtain the normalized unsafe behavior value, worker risk awareness value and management agility value.

Convert the unsafe behavior value, worker risk awareness value and management agility value into values ranging from 0 to 1, and obtain the macro evaluation value of construction safety preset by experts on the construction site;

The normalized unsafe behavior value, worker risk awareness value and management agility value are used as the input x of the fuzzy neural network algorithm, and the corresponding preset macroscopic evaluation value of construction safety is used as the output y of the fuzzy neural network algorithm for training. , to obtain the macroscopic assessment model of construction safety;

Among them, the macro-evaluation model of construction safety includes fuzzy layer, rule layer, regularization layer, secondary fuzzy calculation layer and defuzzification layer;

The fuzzy layer fuzzifies each input through three membership functions, calculates the corresponding membership value, and outputs it to the rule layer;

The rule layer multiplies each membership value to output the activation degree of each rule to the regularization layer;

The regularization layer performs regularization calculation on the activation degree of each rule to output the regularized activation degree to the secondary fuzzy calculation layer;

The secondary fuzzy calculation layer calculates the new membership value through the regularized activation degree to output to the defuzzification layer;

The defuzzification layer remaps the new membership value back to the exact value in the set of lecture points to output the construction safety macro assessment value.

In one embodiment, preferably, a macro-assessment of construction safety is performed on the target construction site using a macro-assessment model of construction safety, including:

Collect target data images of the target construction site and perform labeling and identification processing to determine each worker and each construction machine at the target construction site, as well as the corresponding center point coordinates and outline coordinates of each worker and each construction machine;

According to each worker and each construction machine at the target construction site, as well as the corresponding center point coordinates and outline coordinates of each worker and each construction machine, determine the daily target unsafe behavior value of workers without helmets, target workers Risk awareness value and target management agility value;

According to the target unsafe behavior value, the target worker's risk awareness value, the target management agility value and the construction safety macro evaluation model, the target construction safety macro evaluation value corresponding to the target construction site is calculated.

According to a third aspect of the embodiments of the present invention, a construction safety macro evaluation device is provided, the device comprising:

processor;

memory for storing processor-executable instructions;

where the processor is configured as:

Collect data images of the construction site and perform labeling and identification processing to determine each worker and each construction machine on the construction site, as well as the corresponding center point coordinates and outline coordinates of each worker and each construction machine;

According to each worker and each construction machine on the construction site, as well as the corresponding center point coordinates and outline coordinates of each worker and each construction machine, the daily unsafe behavior value and worker risk awareness value of workers without helmets are determined and managing agility values;

According to the unsafe behavior value, worker risk awareness value and management agility value within the preset time period, and the corresponding preset construction safety macro evaluation value, the fuzzy neural network algorithm is used for training, and the construction safety macro evaluation model is obtained;

Use the construction safety macro-evaluation model to perform a construction safety macro-evaluation on the target construction site.

The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

Based on the LCB theory, the invention classifies the identification results of the construction site, the unsafe behavior of people and the unsafe state of objects from the three perspectives of safety decision-making power, safety culture and safety behavior, and evaluates experts through the method of fuzzy neural network. It is described as an automatic algorithm, so as to realize the automatic evaluation of the macro safety level of the construction project, and reflect the actual safety level of the construction site through objective data. In the dimension of safety decision-making power, based on computer vision technology, the safety management level of the construction site is characterized by measuring the unsafe behavior of people and the duration of unsafe state of objects, and the final result is T. In the dimension of safety culture, based on computer vision technology, the risk awareness value of the construction worker group is reflected by measuring the number of workers exposed to risks, thereby reflecting the safety culture level of the construction site, and the final result is N. In the dimension of safety behavior, the frequency of unsafe behaviors is identified by deep learning-based computer vision technology to characterize the safety behavior level of the construction site, and the final result is F. Taking the management time value T, the worker's risk awareness value N, and the unsafe behavior value F as input, a fuzzy neural network is constructed, and finally the macroscopic safety level S of the construction site is formed.

Therefore, the present invention is based on the decision-making-culture-behavior LCB theory, combined with computer vision and fuzzy neural network methods, to construct a construction site macro safety assessment system. Through the global image data of the construction site, the safety management status of the construction project is evaluated from three aspects: unsafe behavior, worker risk awareness and management agility value. Through the objective collection of construction site data, combined with expert knowledge, the project safety management capabilities of construction project managers, owners and other stakeholders can be improved.

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the invention.

Description of drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and together with the description serve to explain the principles of the invention.

Fig. 1 is a block diagram of a macro-evaluation system for construction safety according to an exemplary embodiment.

Fig. 2 is a block diagram of a data processing module in a construction safety macro-evaluation system according to an exemplary embodiment.

Fig. 3 is a block diagram of a data processing module in a construction safety macro-evaluation system according to an exemplary embodiment.

Fig. 4 is a flow chart of a macro-evaluation method for construction safety according to an exemplary embodiment.

Detailed ways

Exemplary embodiments will be described in detail herein, examples of which are illustrated in the accompanying drawings. Where the following description refers to the drawings, the same numerals in different drawings refer to the same or similar elements unless otherwise indicated. The implementations described in the illustrative examples below are not intended to represent all implementations consistent with the present invention. Rather, they are merely examples of apparatus and methods consistent with some aspects of the invention as recited in the appended claims.

Fig. 1 is a block diagram of a macro-evaluation system for construction safety according to an exemplary embodiment.

As shown in FIG. 1 , according to a first aspect of an embodiment of the present invention, a macro-evaluation system for construction safety is provided, including:

The data processing module 11 is used to collect data images of the construction site and perform labeling and identification processing to determine each worker and each construction machine on the construction site, as well as the corresponding center point coordinates and coordinates of each worker and each construction machine. contour coordinates;

The safety evaluation module 12 is used to determine the daily unsafety of workers not wearing helmets according to each worker and each construction machine on the construction site, as well as the corresponding center point coordinates and contour coordinates of each worker and each construction machine Behavioural value, worker risk awareness value and management agility value;

The training module 13 is used for training by using a fuzzy neural network algorithm according to the unsafe behavior value, worker risk awareness value and management agility value within a preset time period, and the corresponding preset construction safety macro-evaluation value, to obtain construction safety macro evaluation model;

The evaluation module 14 is used to perform macro-evaluation of construction safety on the target construction site by using the macro-evaluation model of construction safety.

Fig. 2 is a block diagram of a data processing module in a construction safety macro-evaluation system according to an exemplary embodiment.

As shown in FIG. 2, in one embodiment, preferably, the data processing module 11 includes:

The data collection unit 21 is used to collect data images of the construction site through the camera device; specifically, the data collection unit includes five sub-modules: a tethered airship, a high-definition camera, a pan/tilt, a picture transmission module and a remote control module. In order to ensure that the picture covers the construction site, the high-definition camera is installed at a height of at least 200 meters, and the camera inclination is based on the picture covering the entire construction site. In order to ensure the clarity of the picture, the present invention selects 6K and above high-definition cameras to monitor the construction site in real time. The gimbal module adopts a three-axis self-stabilizing gimbal, which mainly uses the IMU (inertial measurement unit) and the motor magnetic encoder to calibrate the camera lens on the gimbal to ensure the stability of the lens. The remote control module is used to control the orientation of the lens. The image transmission module is used to transmit the pictures collected by the high-definition camera to the ground industrial computer in real time.

The construction site lacked a high-altitude location for installing cameras, so the tethered airship was chosen to provide a location for camera installation. The main reason for choosing a tethered airship here is that it can operate uninterruptedly at high altitudes for a long time, and the price is more economical than UAVs (UAVs, industrial balloons and other equipment combined with the tethered system can also meet the requirements here. ). In the specific implementation, firstly, based on the construction drawings, select a suitable location on the site to lift the airship into the air, and lift the airship to a height of 100 meters through the mooring system under the premise of stable power supply. Then connect the remote control module and the image transmission module through the industrial computer on the ground to adjust the orientation of the high-definition camera. The image transmission module transmits the photos taken by the high-definition camera to the ground industrial computer in real time. In addition, in order to ensure the long-term work of the data acquisition module, the tethered airship is connected to the power supply of the construction site (which can be the power supply of the living area) and supplies power to the overall data acquisition module.

The calibration unit 22 is used for calibrating the camera device every day;

After the tethered airship takes off for the first time, the high-definition camera shooting area is adjusted through the ground industrial computer so that the picture can cover the entire construction site. At this time, the value of the IMU sensor is reset to zero. The role of the IMU sensor is to measure the attitude of the high-definition camera. The IMU sensor module forms a Cartesian coordinate system through a combination unit consisting of three accelerometers and three gyroscopes. With an x-axis, y-axis, and z-axis, the sensor can measure linear motion along each axis, as well as rotational motion around each axis. After zeroing, the linear motion and rotation angle are adjusted every half hour so that each value is zero again. Due to the accumulated error of the IMU, the IMU value is reset to zero after manual calibration at 7:00 every morning. Before use, a 2000mm×2000mm purple square plastic plate should be placed on site as a calibration color block, and the complete existence of the color block should be confirmed in a high-definition camera. Traverse the pixels in a frame of image to find pixel blocks that meet the following requirements:

H ∈ [125,155], S ∈ [43,255], V ∈ [46,255]

And record the pixel block coordinates corresponding to the maximum and minimum values in the horizontal direction and the pixel block coordinates corresponding to the maximum and minimum values in the vertical direction. Calculate the Euclidean distance in pairs, add the obtained 4 distances and calculate the arithmetic average, the corresponding real distance is 2 meters, and obtain the scale of the camera image.

The marking unit 23 is used to mark the special roads for construction machinery and high-risk areas for construction operations according to the received marking instructions before daily construction;

Due to the complex construction site environment, falling from a height is the most common accident type, and the direct cause is that workers work on the edge of a high place. Therefore, calibrating the real risk edge in a complex environment is a problem that the labeling unit needs to solve. Before marking, the user needs to fully inspect the site and record the risk edge to be marked. When labeling, it is necessary to first draw the risk edge with a straight line or curve, and then draw the scope of the risk area. The specific steps are as follows. First, click the direction of the risk area with the mouse to determine whether the risk area is generated inward or outward, then enter the real width of the risk area, such as 5 meters, and then convert it into the distance in the image according to the scale to describe the distance in the image. The risk edge is used as the benchmark, parallel lines are drawn in the selected direction, and the distance is the converted distance. Finally, the area involved in the risk area is highlighted in the figure.

Human-machine collision accidents are the main source of construction accidents. The essential reason is that workers appear in the risk area of construction operations. Therefore, the risk area of construction machinery needs to be demarcated in advance. The risk area of construction machinery includes the surrounding risk area and the preset construction machinery road. In the calibration module, specific construction machinery moving roads need to be calibrated at the beginning of construction, including haul roads for heavy construction machinery, construction roads and temporary roads for normal construction machinery, and the fill color algorithm uses the fillPoly algorithm in OpenCV. The calibration method is similar to that of the calibration risk edge. Pay attention to ensure the integrity and closure of the calibrated area. Finally, each day needs to use a polygonal wireframe to delineate the area to be constructed on that day before construction.

The identification unit 24 is used to train workers, mobile construction machinery, and tower cranes based on the Mask-RCNN and Deepsort algorithms through pre-collected and labeled construction images, and to identify workers, mobile construction machinery, and tower cranes in the data images of the construction site in minutes. The outline coordinates of mobile construction machinery and tower cranes, and each identification object is tracked; in the database, the total number of identified workers, mobile construction machinery and tower cranes is counted in minutes; When the contour is identified, identify whether the worker is wearing a helmet, and count the number of the corresponding number of not wearing a helmet, number each identified worker, mobile construction machinery and tower crane, and assign the corresponding contour coordinates and center point coordinates. Stored in the data set of workers, mobile construction machinery and tower cranes corresponding to this moment.

The identification unit is mainly used to identify the workers and construction machinery on the construction site and obtain the image coordinate points corresponding to their outer contours. The present invention collects and labels construction site data in advance, and trains construction personnel, typical mobile construction machinery (excavators, mobile cranes, forklifts, pile drivers, etc.) and tower cranes based on the Mask-RCNN and Deepsort algorithms. The unit segmented the outlines of construction workers, mobile construction machinery and tower cranes in the images, and tracked the identified objects. For each identified object, the number of that category is incremented by one at the minute's position in the database. For the silhouette identified as a construction worker, carry out helmet identification, if no helmet is identified, add 1 to the number of people without helmets corresponding to that moment in the database. Number each identified worker, mobile construction machinery and tower crane, and store its corresponding contour coordinates and center point coordinates in the data set of workers, mobile construction machinery and tower cranes corresponding to that moment, and the outline of each element in the database The storage schematic is shown in Table 1. Where w stands for workers, c stands for mobile cranes, f stands for forklifts, e stands for excavators, t stands for tower cranes, and tr stands for trucks.

Table 1

In one embodiment, preferably, the security evaluation module is used to:

Store the number of people without helmets in the database as a time series array UB _i , i = 0,1,2... n -1, let Wi = UB _{i +1} - UB _i , W ₀ = 0, where _n is the time length of the time series. If Wi < 0, let Wi ₌ 0, construct an array Wi _, and use the following first calculation formula to calculate the daily unsafe behavior value:

Among them, Ni is the total number of workers on the construction site at time _i ;

According to the corresponding center point coordinates and contour coordinates of each worker and each construction machine in the database, count the number of workers exposed to work in the risk area at each moment, and use the following second calculation formula to calculate the risk awareness score of the first worker:

表示每日暴露在风险区作业的工人数量值，即第一工人风险意识分值；Among them, NW _i represents the number of workers exposed to the risk area at time i, N _i is the total number of workers on the construction site at time i,

Represents the value of the number of workers exposed to the risk area every day, that is, the risk awareness score of the first worker;

According to the corresponding center point coordinates and contour coordinates of each worker and each construction machine in the database, count the number of workers exposed to the special road for workers and machinery at each moment, and use the following third calculation formula to calculate the second worker's risk awareness score :

Among them, NRW _i represents the number of workers exposed on the special road for workers and machinery at time i, Ni is the total number of workers on the construction site at time i, and N _rw represents _the daily value of the number of workers exposed on the special road for workers and machinery, that is, the second The worker's risk awareness score, wherein the worker's risk awareness value includes the first worker's risk awareness score and the second worker's risk awareness score;

In this embodiment, the risks include a fall from height risk and a human-machine collision risk. Due to the need of work, workers sometimes have to work in high-risk areas. Such workers exposed to high-risk areas do not represent a low level of macro safety in the construction project. Therefore, the number of such workers needs to be excluded from the calculation of the number. In principle, workers should not be present on roads specially planned for construction machinery, so the number of workers present on haul roads, construction roads and temporary roads reflects their awareness of their own risks. To sum up, the algorithm of worker risk awareness value calculates the number of two types of workers, namely, the number of workers exposed to the peak area and the number of workers exposed to the road dedicated to construction machinery. The specific algorithm is as follows.

，其中N _i 为i时刻工地的总人数。Set 5 meters as the risk distance of human-machine collision, and convert 5 meters into the pixel distance in the figure through the scale bar of the calibration module. The number of workers exposed to the high-risk area is first calculated by calling the contour data and center point data of the construction machine per minute identified by the identification unit to calculate the coordinates and center of each contour point in the contour data of the construction machine (a mobile crane is used here as an example). The difference of the point coordinates, and the maximum value and the distance between the corresponding pixel points of 5 meters are added to obtain the radius R _ctkm , k = 1,2,...n, m = 1,2,...n, where ct represents the movement type crane, k represents the number of the crane, m represents the mth minute, then draw a circle with the coordinates of the identified construction machinery center point as the center, R _ctkm as the radius, and then count the exposure to the construction machinery circle and the coordinates of the risk area at the edge of the calibration module in this minute. The number of workers in the area, subtract the number of workers exposed to the construction area in this minute to get an array of time series ( UNW _i ), i = 1,2... n , and then count in the database the continuous retention time in the risk area is less than or equal to Number of workers at 5 minutes, counting the number of such workers per minute, revised for ( UNW _i ), i = 1,2... n (plus the number of workers per minute who stay in the risk area for no more than 5 minutes) , get the revised time series ( NW _i ), i = 1,2... n , NW _i represents the number of workers exposed to the risk area at time i, and finally calculate the value of the number of workers exposed to the risk area on that day

, where Ni is the total number of people on the construction site at time _i .

；The number of workers exposed on the special road for construction machinery is calculated by counting the number of workers who appear on the special road for construction machinery (calibration module calibration) per minute, as follows

;

Among them, Ni is the total number of workers at the construction site at time i, and NRW i _is the number of workers on the road dedicated to construction machinery at time _i . Finally output the daily N=( N _hw , N _rw ).

The daily time T _wh of workers not wearing helmets, the time T _wr of workers entering the special road for construction machinery, the time T _r that the special road for construction machinery is occupied, the unit of time is minutes, and the final output management agility value T = ( T _wh , T _wr , T _r ).

The main statistics of management agility are the unsafe state and the duration of unsafe behavior of people, including the time when workers are not wearing safety helmets, and when workers appear in the lanes dedicated to construction machinery and the lanes dedicated to construction machinery are blocked.

During the construction process, construction materials often occupy the road of construction machinery, resulting in unsafe entry and exit of construction machinery. The invention first extracts the range of the special road for construction machinery that is calibrated from the pictures collected by the high-definition camera every minute, then subtracts the range corresponding to the construction machinery and the worker identified by the recognition unit, and then uses the background difference algorithm to count the object points in the foreground for the remaining pictures. Coordinates, then cluster the points in the foreground, use the maximum silhouette coefficient to determine the number of categories, count the maximum distance between the coordinates of the cluster center point and the coordinates of each point in the category and compare it with the distance of the corresponding pixel of 1 meter, if it is greater than 1 meter , the category shall be recorded as occupation of special roads for construction machinery.

The daily time T _wh of workers not wearing helmets, the time T _wr of workers entering the special road for construction machinery, the time T _r that the special road for construction machinery is occupied, the unit of time is minutes, and the final output management agility value T = ( T _wh , T _wr , T _r ).

Fig. 3 is a block diagram of a data processing module in a construction safety macro-evaluation system according to an exemplary embodiment.

As shown in FIG. 3, in one embodiment, preferably, the training module 13 includes:

The normalization unit 31 is configured to use the MinMaxScaler normalization algorithm to normalize the unsafe behavior value, worker risk awareness value, and management agility value, and obtain the normalized unsafe behavior value and worker risk awareness value. value and management agility values;

The conversion unit 32 is configured to convert the unsafe behavior value, the worker's risk awareness value and the management agility value into values in the range of 0-1, and obtain the construction safety macro-evaluation value preset by the expert on the construction site;

The training unit 33 is used for taking the normalized unsafe behavior value, worker risk awareness value and management agility value as the input x of the fuzzy neural network algorithm, and the corresponding preset macroscopic evaluation value of construction safety as the fuzzy neural network algorithm The output y is trained to obtain the construction safety macro evaluation model; the training algorithm adopts the hybrid training algorithm and the hybrid training method in the ANFIS toolbox of matlab. For example, 15 data points can be randomly selected from 20 samples as training data, the other 5 data points are used as the test set, the training error acceptance range is set to 0.001, and the maximum number of training times is set to 1000 times.

Among them, the macro-evaluation model of construction safety includes fuzzy layer, rule layer, regularization layer, secondary fuzzy calculation layer and defuzzification layer;

The fuzzy layer fuzzifies each input through three membership functions, calculates the corresponding membership value, and outputs it to the rule layer;

The model has a total of six inputs ( F , N _hw , N _rw , T _wh , T _wr , T _r ) and outputs a single value S, each input is processed by three membership functions A _{1 i} , A _{2 i} , A _{3 i} Fuzzy, representing the high, medium and low levels of a certain safety performance at the construction site on that day, where the first subscript represents the order of the three membership functions, and the second i represents the six inputs. order. ANFIS can adaptively learn fuzzy logic similar to human reasoning through the (F, N, T, S) data set:

The membership function fuzzifies the classical set theory and expresses the degree to which the input satisfies the required mathematical properties of a set. Commonly used membership functions are triangular and Gaussian. The present invention adopts a Gaussian membership function, and the calculation formula of the output of the fuzzy layer is:

。

.

where c _i and σ _i are the membership function shape parameters, and x is the original data received by the fuzzy layer, that is, the input data. Using the ANFIS toolbox of matlab, first load the collected data set, then set the number of membership functions corresponding to each input to 3, and select the type of membership function as gaussmf. The structural design of ANFIS needs to design the number and types of membership functions in the ANFIS toolbox of matlab.

The rule layer multiplies each membership value to output the activation degree of each rule to the regularization layer; the calculation formula of the rule layer is:

，

,

Its output represents the activation level of each if rule.

。The regularization layer performs regularization calculation on the activation degree of each rule to output the regularized activation degree to the secondary fuzzy calculation layer; the calculation formula is:

.

，The secondary fuzzy calculation layer calculates the new membership value through the regularized activation degree to output to the defuzzification layer; its calculation formula is:

,

where x _i is the _ith input and pi is the corresponding weight.

。The defuzzification layer remaps the new membership value back to the exact value in the set of lecture points to output the construction safety macro assessment value. Its calculation formula is:

.

In one embodiment, the evaluation module 14 is preferably used to:

Collect target data images of the target construction site and perform labeling and identification processing to determine each worker and each construction machine at the target construction site, as well as the corresponding center point coordinates and outline coordinates of each worker and each construction machine;

According to each worker and each construction machine at the target construction site, as well as the corresponding center point coordinates and outline coordinates of each worker and each construction machine, determine the daily target unsafe behavior value of workers without helmets, target workers Risk awareness value and target management agility value;

According to the target unsafe behavior value, the target worker's risk awareness value, the target management agility value and the construction safety macro evaluation model, the target construction safety macro evaluation value corresponding to the target construction site is calculated.

Based on the LCB theory, the invention classifies the identification results of the construction site, the unsafe behavior of people and the unsafe state of objects from the three perspectives of safety decision-making power, safety culture and safety behavior, and evaluates experts through the method of fuzzy neural network. It is described as an automatic algorithm, so as to realize the automatic evaluation of the macro safety level of the construction project, and reflect the actual safety level of the construction site through objective data. In the dimension of safety decision-making power, based on computer vision technology, the safety management level of the construction site is characterized by measuring the unsafe behavior of people and the duration of unsafe state of objects, and the final result is T. In the dimension of safety culture, based on computer vision technology, the risk awareness value of the construction worker group is reflected by measuring the number of workers exposed to risks, thereby reflecting the safety culture level of the construction site, and the final result is N. In the dimension of safety behavior, the frequency of unsafe behaviors is identified by deep learning-based computer vision technology to characterize the safety behavior level of the construction site, and the final result is F. Taking the management time value T, the worker's risk awareness value N and the unsafe behavior value F as input, a fuzzy neural network is constructed, and finally the macroscopic safety level S of the construction site is formed.

Therefore, the present invention is based on the decision-making-culture-behavior LCB theory, combined with computer vision and fuzzy neural network methods, to construct a construction site macro safety assessment system. Through the global image data of the construction site, the safety management status of the construction project is evaluated from three aspects: unsafe behavior, worker risk awareness and management agility value. Through the objective collection of construction site data, combined with expert knowledge, the project safety management capabilities of construction project managers, owners and other stakeholders can be improved.

Fig. 4 is a flow chart of a macro-evaluation method for construction safety according to an exemplary embodiment.

As shown in FIG. 4 , according to a second aspect of the embodiments of the present invention, a construction safety macro evaluation method is provided, which is used in a construction safety macro evaluation system, and the method includes:

Step S401, collecting data images of the construction site and performing labeling and identification processing to determine each worker and each construction machine on the construction site, as well as the center point coordinates and outline coordinates corresponding to each worker and each construction machine;

Step S402, according to each worker and each construction machine on the construction site, and the corresponding center point coordinates and outline coordinates of each worker and each construction machine, determine the daily unsafe behavior value of the worker without a helmet, the worker Risk awareness value and management agility value;

Step S403, according to the unsafe behavior value, worker risk awareness value and management agility value within the preset time period, and the corresponding preset construction safety macro evaluation value, use fuzzy neural network algorithm to train, and obtain the construction safety macro evaluation model ;

Step S404, using the construction safety macro-evaluation model to perform a construction safety macro-evaluation on the target construction site.

In one embodiment, preferably, the above step S401 includes the following steps:

Collect the data images of the construction site through the camera device;

daily calibration of the camera;

Before daily construction, according to the received marking instructions, mark the dedicated roads for construction machinery and high-risk areas for construction operations;

Through pre-collected and labeled construction images, workers, mobile construction machinery, and tower cranes are trained based on the Mask-RCNN and Deepsort algorithms, and workers, mobile construction machinery, and tower cranes are identified in the data images of the construction site in minutes. The coordinates of the contours of the identified objects are tracked, and each identified object is tracked; in the database, the total number of identified workers, mobile construction machinery and tower cranes is counted in minutes; among them, when the contours of the workers are identified, the workers are identified. Whether to wear a safety helmet, and the corresponding statistics of the number of not wearing a safety helmet, number each worker, mobile construction machinery and tower crane identified, and store their corresponding contour coordinates and center point coordinates in the corresponding moment. A dataset of workers, mobile construction machinery and tower cranes.

In one embodiment, preferably, step S402 includes:

Store the number of people without helmets in the database as a time series array UB _i , i = 0,1,2... n -1, let Wi = UB _{i +1} - UB _i , W ₀ = 0, where _n is the time length of the time series. If Wi < 0, let Wi ₌ 0, construct an array Wi _, and use the following first calculation formula to calculate the daily unsafe behavior value:

Among them, Ni is the total number of workers on the construction site at time _i ;

According to the corresponding center point coordinates and contour coordinates of each worker and each construction machine in the database, count the number of workers exposed to work in the risk area at each moment, and use the following second calculation formula to calculate the risk awareness score of the first worker:

表示每日暴露在风险区作业的工人数量值，即第一工人风险意识分值；Among them, NW _i represents the number of workers exposed to the risk area at time i, N _i is the total number of workers on the construction site at time i,

Represents the value of the number of workers exposed to the risk area every day, that is, the risk awareness score of the first worker;

According to the corresponding center point coordinates and contour coordinates of each worker and each construction machine in the database, count the number of workers exposed to the special road for workers and machinery at each moment, and use the following third calculation formula to calculate the second worker's risk awareness score :

Among them, NRW _i represents the number of workers exposed on the special road for workers and machinery at time i, Ni is the total number of workers on the construction site at time i, and N _rw represents _the daily value of the number of workers exposed on the special road for workers and machinery, that is, the second The worker's risk awareness score, wherein the worker's risk awareness value includes the first worker's risk awareness score and the second worker's risk awareness score;

The daily time T _wh of workers not wearing helmets, the time T _wr of workers entering the special road for construction machinery, the time T _r that the special road for construction machinery is occupied, the unit of time is minutes, and the final output management agility value T = ( T _wh , T _wr , T _r ).

In one embodiment, preferably, step S403 includes:

The MinMaxScaler normalization algorithm is used to normalize the unsafe behavior value, worker risk awareness value and management agility value to obtain the normalized unsafe behavior value, worker risk awareness value and management agility value.

Convert unsafe behavior value, worker risk awareness value and management agility value into values ranging from 0 to 1, and obtain the macro evaluation value of construction safety preset by experts on the construction site;

The normalized unsafe behavior value, worker risk awareness value and management agility value are used as the input x of the fuzzy neural network algorithm, and the corresponding preset macroscopic evaluation value of construction safety is used as the output y of the fuzzy neural network algorithm for training. , to obtain the macroscopic assessment model of construction safety;

Among them, the macro-evaluation model of construction safety includes fuzzy layer, rule layer, regularization layer, secondary fuzzy calculation layer and defuzzification layer;

The fuzzy layer fuzzifies each input through three membership functions, calculates the corresponding membership value, and outputs it to the rule layer;

The rule layer multiplies each membership value to output the activation degree of each rule to the regularization layer;

The regularization layer performs regularization calculation on the activation degree of each rule to output the regularized activation degree to the secondary fuzzy calculation layer;

The secondary fuzzy calculation layer calculates the new membership value through the regularized activation degree to output to the defuzzification layer;

The defuzzification layer remaps the new membership value back to the exact value in the set of lecture points to output the construction safety macro assessment value.

In one embodiment, preferably, step S404 includes:

Collect target data images of the target construction site and perform labeling and identification processing to determine each worker and each construction machine at the target construction site, as well as the corresponding center point coordinates and outline coordinates of each worker and each construction machine;

According to each worker and each construction machine at the target construction site, as well as the corresponding center point coordinates and outline coordinates of each worker and each construction machine, determine the daily target unsafe behavior value of workers without helmets, target workers Risk awareness value and target management agility value;

According to the target unsafe behavior value, the target worker's risk awareness value, the target management agility value and the construction safety macro evaluation model, the target construction safety macro evaluation value corresponding to the target construction site is calculated.

According to a third aspect of the embodiments of the present invention, a construction safety macro evaluation device is provided, the device comprising:

processor;

memory for storing processor-executable instructions;

where the processor is configured as:

Collect data images of the construction site and perform labeling and identification processing to determine each worker and each construction machine on the construction site, as well as the corresponding center point coordinates and outline coordinates of each worker and each construction machine;

According to each worker and each construction machine on the construction site, as well as the corresponding center point coordinates and outline coordinates of each worker and each construction machine, the daily unsafe behavior value and worker risk awareness value of workers without helmets are determined and managing agility values;

According to the unsafe behavior value, worker risk awareness value and management agility value within the preset time period, and the corresponding preset construction safety macro evaluation value, the fuzzy neural network algorithm is used for training, and the construction safety macro evaluation model is obtained;

Use the construction safety macro-evaluation model to perform a construction safety macro-evaluation on the target construction site.

It can be further understood that, in the present invention, "a plurality" refers to two or more than two, and other measure words are similar. "And/or", which describes the association relationship of the associated objects, means that there can be three kinds of relationships, for example, A and/or B, which can mean that A exists alone, A and B exist at the same time, and B exists alone. The character "/" generally indicates that the associated objects are an "or" relationship. The singular forms "a," "" and "the" are intended to include the plural forms as well, unless the context clearly dictates otherwise.

It is further understood that the terms "first", "second", etc. are used to describe various information, but the information should not be limited to these terms. These terms are only used to distinguish the same type of information from one another, and do not imply a particular order or level of importance. In fact, the expressions "first", "second" etc. are used completely interchangeably. For example, first information may also be referred to as second information, and similarly, second information may also be referred to as first information, without departing from the scope of the present invention.

It should be further understood that, although the operations in the embodiments of the present invention are described in a specific order in the drawings, it should not be construed as requiring that the operations be performed in the specific order shown or the serial order, or requiring Perform all operations shown to obtain the desired result. In certain circumstances, multitasking and parallel processing may be advantageous.

Other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses or adaptations of the invention which follow the general principles of the invention and which include common knowledge or conventional techniques in the art not disclosed by the invention . The specification and examples are to be regarded as exemplary only, with the true scope and spirit of the invention being indicated by the following claims.

It should be understood that the present invention is not limited to the precise structures described above and illustrated in the accompanying drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present invention is limited only by the appended claims.

Claims (10)

1. a construction safety macro evaluation system, is characterized in that, comprises: The data processing module is used to collect data images of the construction site and perform labeling and identification processing to determine each worker and each construction machine on the construction site, as well as the corresponding center point coordinates and contours of each worker and each construction machine coordinate; The safety evaluation module is used to determine the daily unsafe behavior of workers without helmets according to each worker and each construction machine on the construction site, as well as the corresponding center point coordinates and outline coordinates of each worker and each construction machine value, worker risk awareness value and management agility value; The training module is used for training with fuzzy neural network algorithm according to the unsafe behavior value, worker risk awareness value and management agility value within a preset time period, and the corresponding preset macroscopic evaluation value of construction safety to obtain the macroscopic construction safety value. evaluation model; The assessment module is used to perform macro-assessment of construction safety on the target construction site using the macro-assessment model of construction safety. 2. The construction safety macro-evaluation system according to claim 1, wherein the data processing module comprises: The data acquisition unit is used to collect the data images of the construction site through the camera device; Calibration unit for daily calibration of the camera; The marking unit is used to mark the special roads for construction machinery and high-risk areas for construction operations according to the received marking instructions before daily construction; The identification unit is used to train workers, mobile construction machinery, and tower cranes based on the Mask-RCNN and Deepsort algorithms through pre-collected and labeled construction images, and to identify workers, mobile, and mobile in the data images of the construction site in minutes The outline coordinates of mobile construction machinery and tower cranes are collected, and each identification object is tracked; in the database, the total number of identified workers, mobile construction machinery and tower cranes is counted in minutes; When contouring, identify whether workers wear safety helmets, and count the number of people who do not wear safety helmets, number each identified worker, each construction machine, and each tower crane, and set the corresponding contour coordinates to the center. The point coordinates are synchronously stored in the corresponding data sets of workers, mobile construction machinery and tower cranes. 3. construction safety macro-evaluation system according to claim 2, is characterized in that, safety evaluation module is used for: Store the number of people without helmets in the database as a time series array UB _i , i = 0,1,2... n -1, let Wi = UB _{i +1} - UB _i , W ₀ = 0, where _n is the time length of the time series. If Wi < 0, let Wi ₌ 0, construct an array Wi _, and use the following first calculation formula to calculate the daily unsafe behavior value:

Among them, Ni is the total number of workers on the construction site at time _i ; According to the corresponding center point coordinates and contour coordinates of each worker and each construction machine in the database, count the number of workers exposed to work in the risk area at each moment, and use the following second calculation formula to calculate the risk awareness score of the first worker:

Among them, NW _i represents the number of workers exposed to the risk area at time i, N _i is the total number of workers on the construction site at time i,

Represents the value of the number of workers exposed to the risk area every day, that is, the risk awareness score of the first worker; According to the corresponding center point coordinates and contour coordinates of each worker and each construction machine in the database, count the number of workers exposed to the special road for workers and machinery at each moment, and use the following third calculation formula to calculate the second worker's risk awareness score :

Among them, NRW _i represents the number of workers exposed on the special road for workers and machinery at time i, Ni is the total number of workers on the construction site at time i, and N _rw represents _the daily value of the number of workers exposed on the special road for workers and machinery, that is, the second The worker's risk awareness score, wherein the worker's risk awareness value includes the first worker's risk awareness score and the second worker's risk awareness score; The daily time T _wh of workers not wearing helmets, the time T _wr of workers entering the special road for construction machinery, the time T _r that the special road for construction machinery is occupied, the unit of time is minutes, and the final output management agility value T = ( T _wh , T _wr , T _r ). 4. construction safety macro-evaluation system according to claim 1, is characterized in that, training module comprises: The normalization unit is used to use the MinMaxScaler normalization algorithm to normalize the unsafe behavior value, worker risk awareness value, and management agility value, and obtain the normalized unsafe behavior value and worker risk awareness value. and managing agility values The conversion unit is used to convert the unsafe behavior value, worker risk awareness value and management agility value into values in the range of 0-1, and obtain the macro-evaluation value of construction safety preset by experts on the construction site; The training unit is used to use the normalized unsafe behavior value, worker risk awareness value and management agility value as the input x of the fuzzy neural network algorithm, and the corresponding preset macroscopic evaluation value of construction safety as the fuzzy neural network algorithm. Output y, carry out training, and obtain the macroscopic evaluation model of construction safety; Among them, the macro-evaluation model of construction safety includes fuzzy layer, rule layer, regularization layer, secondary fuzzy calculation layer and defuzzification layer; The fuzzy layer fuzzifies each input through three membership functions, calculates the corresponding membership value, and outputs it to the rule layer; The rule layer multiplies each membership value to output the activation degree of each rule to the regularization layer; The regularization layer performs regularization calculation on the activation degree of each rule to output the regularized activation degree to the secondary fuzzy calculation layer; The secondary fuzzy calculation layer calculates the new membership value through the regularized activation degree to output to the defuzzification layer; The defuzzification layer remaps the new membership value back to the exact value in the set of lecture points to output the construction safety macro assessment value. 5. construction safety macro evaluation system according to claim 1, is characterized in that, evaluation module is used for: Collect target data images of the target construction site and perform labeling and identification processing to determine each worker and each construction machine at the target construction site, as well as the corresponding center point coordinates and outline coordinates of each worker and each construction machine; According to each worker and each construction machine at the target construction site, as well as the corresponding center point coordinates and outline coordinates of each worker and each construction machine, determine the daily target unsafe behavior value of workers without helmets, target workers Risk awareness value and target management agility value; According to the target unsafe behavior value, the target worker's risk awareness value, the target management agility value and the construction safety macro evaluation model, the target construction safety macro evaluation value corresponding to the target construction site is calculated. 6. A construction safety macro-evaluation method, characterized in that, comprising: Collect data images of the construction site and perform labeling and identification processing to determine each worker and each construction machine on the construction site, as well as the corresponding center point coordinates and outline coordinates of each worker and each construction machine; According to each worker and each construction machine on the construction site, as well as the corresponding center point coordinates and outline coordinates of each worker and each construction machine, the daily unsafe behavior value and worker risk awareness value of workers without helmets are determined and managing agility values; According to the unsafe behavior value, worker risk awareness value and management agility value within the preset time period, and the corresponding preset construction safety macro evaluation value, the fuzzy neural network algorithm is used for training, and the construction safety macro evaluation model is obtained; Use the construction safety macro-evaluation model to perform a construction safety macro-evaluation on the target construction site. 7. The construction safety macro-evaluation method according to claim 6, wherein the data images of the construction site are collected and marked and identified to determine each worker and each construction machine on the construction site, and each worker The center point coordinates and contour coordinates corresponding to each construction machine, including: Collect the data images of the construction site through the camera device; daily calibration of the camera; Before daily construction, according to the received marking instructions, mark the dedicated roads for construction machinery and high-risk areas for construction operations; Through pre-collected and labeled construction images, workers, mobile construction machinery, and tower cranes are trained based on the Mask-RCNN and Deepsort algorithms, and workers, mobile construction machinery, and tower cranes are identified in the data images of the construction site in minutes. The coordinates of the contours of the identified objects are tracked, and each identified object is tracked; in the database, the total number of identified workers, mobile construction machinery and tower cranes is counted in minutes; among them, when the contours of the workers are identified, the workers are identified. Whether to wear a safety helmet, and to count the number of the corresponding number of not wearing a safety helmet, to identify each worker, mobile construction machinery and tower crane number, and to store the corresponding contour coordinates and center point coordinates into the corresponding worker synchronously , Data collection of mobile construction machinery and tower cranes. 8. The construction safety macro-evaluation method according to claim 7, characterized in that, according to each worker and each construction machine on the construction site, and the respective center point coordinates and outline coordinates of each worker and each construction machine Determine daily unsafe behavior values, worker risk awareness values, and management agility values for workers without helmets, including: Store the number of people without helmets in the database as a time series array UB _i , i = 0,1,2... n -1, let Wi = UB _{i +1} - UB _i , W ₀ = 0, where _n is the time length of the time series. If Wi < 0, let Wi ₌ 0, construct an array Wi _, and use the following first calculation formula to calculate the daily unsafe behavior value:

Among them, Ni is the total number of workers on the construction site at time _i ; According to the corresponding center point coordinates and contour coordinates of each worker and each construction machine in the database, count the number of workers exposed to work in the risk area at each moment, and use the following second calculation formula to calculate the risk awareness score of the first worker:

Among them, NW _i represents the number of workers exposed to the risk area at time i, N _i is the total number of workers on the construction site at time i,

Represents the value of the number of workers exposed to the risk area every day, that is, the risk awareness score of the first worker; According to the corresponding center point coordinates and contour coordinates of each worker and each construction machine in the database, count the number of workers exposed to the special road for workers and machinery at each moment, and use the following third calculation formula to calculate the second worker's risk awareness score :

Among them, NRW _i represents the number of workers exposed on the special road for workers and machinery at time i, Ni is the total number of workers on the construction site at time i, and N _rw represents _the daily value of the number of workers exposed on the special road for workers and machinery, that is, the second The worker's risk awareness score, wherein the worker's risk awareness value includes the first worker's risk awareness score and the second worker's risk awareness score; The daily time T _wh of workers not wearing helmets, the time T _wr of workers entering the special road for construction machinery, the time T _r that the special road for construction machinery is occupied, the unit of time is minutes, and the final output management agility value T = ( T _wh , T _wr , T _r ). 9. The construction safety macro-evaluation method according to claim 6, characterized in that, according to the unsafe behavior value, worker risk awareness value and management agility value within a preset time period, and the corresponding preset construction safety macro-evaluation value, the fuzzy neural network algorithm is used for training, and the construction safety macro evaluation model is obtained, including: The MinMaxScaler normalization algorithm is used to normalize the unsafe behavior value, worker risk awareness value and management agility value to obtain the normalized unsafe behavior value, worker risk awareness value and management agility value; Convert unsafe behavior value, worker risk awareness value and management agility value into values ranging from 0 to 1, and obtain the macro evaluation value of construction safety preset by experts on the construction site; The normalized unsafe behavior value, worker risk awareness value and management agility value are used as the input x of the fuzzy neural network algorithm, and the corresponding preset macroscopic evaluation value of construction safety is used as the output y of the fuzzy neural network algorithm for training. , to obtain the macroscopic assessment model of construction safety; Among them, the macro-evaluation model of construction safety includes fuzzy layer, rule layer, regularization layer, secondary fuzzy calculation layer and defuzzification layer; The fuzzy layer fuzzifies each input through three membership functions, calculates the corresponding membership value, and outputs it to the rule layer; The rule layer multiplies each membership value to output the activation degree of each rule to the regularization layer; The regularization layer performs regularization calculation on the activation degree of each rule to output the regularized activation degree to the secondary fuzzy calculation layer; The secondary fuzzy calculation layer calculates the new membership value through the regularized activation degree to output to the defuzzification layer; The defuzzification layer remaps the new membership value back to the exact value in the set of lecture points to output the construction safety macro assessment value. 10. The construction safety macro-evaluation method according to claim 6, characterized in that, using a construction safety macro-evaluation model to carry out construction safety macro-evaluation to the target construction site, comprising: Collect target data images of the target construction site and perform labeling and identification processing to determine each worker and each construction machine at the target construction site, as well as the corresponding center point coordinates and outline coordinates of each worker and each construction machine; According to each worker and each construction machine at the target construction site, as well as the corresponding center point coordinates and outline coordinates of each worker and each construction machine, determine the daily target unsafe behavior value of workers without helmets, target workers Risk awareness value and target management agility value; According to the target unsafe behavior value, the target worker's risk awareness value, the target management agility value and the construction safety macro evaluation model, the target construction safety macro evaluation value corresponding to the target construction site is calculated.

Images