CN117133145A - System and method for parking guidance and reverse car seeking based on LSTM network - Google Patents

Abstract

The application discloses a parking guidance and reverse vehicle searching system and method based on an LSTM network, comprising a management server, a parking and vehicle searching system and method, wherein the management server is used for collecting and processing data, calculating an optimal path and planning a parking and vehicle searching line; the identification card is used for acquiring and transmitting RSSI data and self-identity data; the beacon site is used for broadcasting signals outwards in a fixed time interval; the vehicle information identification device is used for identifying the identity data of the device identification card and the vehicle information; the information management system is used for controlling and managing the data of the whole system; the terminal system is used for a user to check vehicle information, parking space information and a parking and vehicle searching line. The LSTM network is trained to realize rapid and accurate positioning of the target vehicle, the parking guidance and the reverse vehicle searching route can be planned rapidly and efficiently, the position of the target vehicle can be determined in the complex parking lot by inputting any license plate number and scanning positioning two-dimensional code during reverse vehicle searching, and the shortest path from the current position to the target parking space can be planned automatically.

Description

System and method for parking guidance and reverse car seeking based on LSTM network

Technical field

This application belongs to the field of Internet of Vehicles technology, and particularly relates to systems and methods for parking guidance and reverse car seeking based on LSTM networks.

Background technique

With the rapid development of science and technology and the rapid popularization of transportation, urban parking resources are becoming increasingly scarce, and the problem of parking space supply has become more and more prominent. The problems of parking and finding a car have not been well solved. According to urban congestion theory, 30% of congestion on urban roads is caused by vehicles looking for parking spaces. According to my country's current standards, the ratio of vehicles to parking spaces in large and medium-sized cities is around 1:0.8. Due to factors such as random parking and management problems in large parking lots, large gaps in parking spaces and high vacancy rates have occurred. question.

In large underground parking lots, due to the large area, many parking spaces, complex space, and large traffic flow, the problems of parking and finding a car are more significant in this environment. Traditional underground parking lots generally adopt manual management or intelligent network equipment system management relying on monitoring equipment, IC cards, and GPS technology, and use image recognition technology or smart sensors, IC cards, etc. to bind parking spaces and vehicle-related information, and then Transfer the data to the server for analysis and processing to achieve guidance to the target. However, the GPS signal in the underground garage is weak, the WiFi signal positioning accuracy is low, and the deployment cost of monitoring equipment, IC cards, smart sensors and other equipment is high, easy to wear and difficult to maintain.

Chinese invention patent (publication number: CN 103337196 A) parking space guidance and reverse car search method and system. During parking guidance, the vehicle needs to swipe the card to enter the parking lot. When the vehicle enters the parking space to complete the parking, it needs to enter the parking space corresponding to the parking space. Swipe the card again on the card positioning machine to achieve positioning. When searching for a car in reverse, the car owner needs to operate on a fixed card swiping machine to obtain the vehicle location information and the car-finding guidance path. However, its operation process is relatively cumbersome. Once you forget to swipe your card, the system will lack important information and cannot implement related functions conveniently and effectively. This positioning method based on manual intervention has poor usability, cumbersome processes, and low intelligence. High, low efficiency.

Chinese invention patent (publication number: CN 113421433 A) is a regional video parking guidance system and parking guidance method. The parking areas are first divided according to the parking lot, and corresponding camera monitoring systems need to be installed in each area. The camera system Identify the number of vehicles entering and exiting each fixed area, and identify the license plate number information and the corresponding driving trajectory to determine the area where the vehicle is parked and the empty parking spaces in the corresponding area, and realize the function of reverse vehicle search by calculating the driving trajectory . This system cannot accurately locate the parking space information where the vehicle is parked, and it only relies on the driving trajectory to achieve reverse vehicle search results with low reliability. The cost of monitoring equipment and maintenance costs are high, and the accuracy of monitoring and recognition alone is limited, resulting in this method. Positioning accuracy and ease of use are low, and deployment costs are high.

The above information disclosed in this Background Art is only for increasing understanding of the Background Art of this application and, therefore, it may contain prior art that does not constitute prior art known to a person of ordinary skill in the art.

Contents of the invention

The purpose of this application is to overcome the shortcomings of the existing technology, provide a parking guidance and reverse car-finding system and method based on the LSTM network, and solve the problem of inaccurate navigation and ease of use of parking guidance and car-finding systems in the existing technology. Technical issues with poor and inefficient parking.

In some embodiments of this application, a parking guidance and reverse car-seeking system based on LSTM network is provided, including:

Management server: receives beacon site data and self-identification data from the identification card, processes the data, establishes a location fingerprint database, trains the LSTM network, predicts the coordinate information of the target site and stores it, and generates authorized and undirected Map, calculate the shortest path, integrate parking space usage and parking vehicle information, receive or feedback instructions and data from the information management system and terminal system.

Identification card: A wireless mobile tag equipped on the target vehicle, used to receive beacon location signal data, and transmit beacon location data and its own identification data to the management server.

Beacon locations: Set up at multiple fixed and known locations in the parking lot area, and broadcast signals to the outside world at fixed time intervals to ensure that the signal data covers the entire parking lot area.

Vehicle information identification equipment: used to identify and monitor vehicle identification card information and vehicle information, and upload relevant data to the management server.

Information management system: used to manage user, vehicle, parking lot and other information, to send relevant instructions and data to the management server, and to display the information fed back by the management server in the view layer to achieve data control and management of the entire system.

Terminal system: used to obtain user instruction information, communicate with the information management system and server, and mainly realize the display of relevant information such as user self-service car search, vehicle search, and parking guidance.

In some embodiments of the present application, the terminal system is a smart phone, a tablet computer, a notebook computer, and other types of user terminals. The terminals all run car-finding applications for providing car-finding services, such as car-finding customers. terminal or car search applet, etc.

In some embodiments of this application, the terminal system is a mobile app system.

In some embodiments of the present application, the vehicle information identification device is installed at key locations such as entrances and exits.

In some embodiments of the present application, the vehicle information identification equipment includes surveillance cameras, identification card identification equipment, and vehicle detectors, which are used to identify and detect information about passing vehicles.

In some embodiments of the present application, the vehicle information includes identification and detection of vehicle passing, presence, vehicle length, color, license plate number and other information.

In another embodiment of the present application, a parking guidance and reverse car-seeking method based on an LSTM network is also provided, using the parking guidance and reverse car-seeking system based on the LSTM network described in the above embodiment;

The method includes the following steps:

Establish a two-dimensional coordinate system in the parking lot, and store the information of each parking space and the corresponding center coordinate point in the service manager. By setting up multiple beacon points at different fixed positions, the identification card obtains its corresponding information at each reference measurement point. The RSSI data of k beacons communicated. At the same time, the vehicle information identification equipment will identify and monitor the vehicle's identification card information and vehicle information and upload it to the management server. The management server will filter the RSSI data to form a fingerprint database, and the data in the database will be processed according to the After the time series is recombined and sorted, the input model sequence S _n is generated, which is used as the input sequence of the LSTM network. The coordinate points are used as the output sequence of the LSTM network, and the corresponding regression model is established to complete the training of the neural network. Input the processed data set and predict the coordinates of the target point.

The management server stores the output sequence of the trained LSTM network as an adjacency matrix, calculates the edge weights of each adjacent vertex according to the weight algorithm, and stores each vertex data and edge data. Generate a weighted undirected graph and use the car-finding method based on the A* algorithm to plan the optimal path.

In some embodiments of the present application, the collected data are standardized and denoised through Kalman filtering. The normalization formula is:

in, Indicates the i-th RSSI data in the data collected at the m-th reference measurement point, max(RSSI) and min(RSSI) respectively represent/> The maximum and minimum values of , the processed RSSI data are iteratively calculated through Kalman filtering.

In some embodiments of the present application, the training process for the LSTM network is to use the processed RSSI data set S _n as the input sequence, where S _n = (X _i-M+1 ,X _iM ,,X _i ), M is the dimension of S _n , indicating that it took M time _to receive this set of data, where Xi ₌ (RSSI ₁ , RSSI ₂ ,, RSSI _k ) represents the data received at time i, k is the dimension of Among the k RSSI values received at any time, X _i is the last position recorded by S _n , that is, the current position, corresponding to the position ( _xi , y _i ) in the two-dimensional coordinate system. The two-dimensional coordinate position corresponding to the reference point is used as the output sequence, a corresponding regression model is established, and the training of the LSTM network is completed.

In some embodiments of this application, k target location information predicted through the LSTM network is stored as an adjacency matrix. The storage form of the location adjacency matrix is:

Calculate the weights of all adjacent vertex edges according to the weight algorithm, set the weight of non-adjacent vertices to "∞", and the weight of its own vertices to "0". The weight algorithm is:

Represents the weight of the edge (a, b) of two vertices with an adjacent relationship, (x _a , y _a ) and (x _b , y _b ) represent the two-dimensional coordinates of the two points.

In some embodiments of the present application, on the basis of storing location information and generating a weighted undirected graph, the A* algorithm is used to plan the two-point shortest path. The user can use a method according to claim 1 The terminal system described in the parking guidance and reverse car-seeking system based on the LSTM network can autonomously select a parking space when parking. When reverse-seeking a car, it can be realized by inputting the license plate number of any nearby vehicle or scanning the positioning QR code. Optimal path planning from the current location to the target parking space.

In some embodiments of the present application, a parking guidance and reverse car-seeking method based on an LSTM network uses the parking guidance and reverse car-seeking system based on the LSTM network described in the above embodiments;

Specifically, it includes the following steps:

Establish a two-dimensional coordinate system in the parking lot, store the information of each parking space and the corresponding center coordinates in the service manager, and achieve full-area signal coverage by setting up multiple beacon points at different fixed positions. When the vehicle is used as an identification card When carrying a mobile tag into the parking lot signal coverage area, the identification card and the beacon location communicate with each other. The identification card will obtain the RSSI of the k beacons communicating with it at each reference measurement point. The identification card will transmit the RSSI data and its own data to Management server, the management server will filter the RSSI data to form a fingerprint database, recombine and sort the data in the database according to time series to generate the input model sequence S _n , and use S _n as the input sequence of the LSTM network, where S _n =( _{Xi - M + 1 , _ _ Represents} the data _{received at time} i _, k is the dimension of

First, by selecting multiple sets of reference measurement points, the obtained RSSI data is processed and then a location fingerprint database is established. The input model sequence S _n and the corresponding two-dimensional plane coordinates (x _i , y _i ) are used as the input sequence of the LSTM network to complete For the training of the LSTM network, the RSSI data collected by the target vehicle identification card and its own data are transmitted to the management server. The LSTM network is used to predict the positioning results to obtain the corresponding two-dimensional coordinate position of the target vehicle. The management server will record all For the coordinate point information of vehicles and key locations, each coordinate point is regarded as a node. The node stores location information and vehicle information. The nodes are connected to each other within a certain weight range to form a weighted undirected graph. In reverse car search During the process, the user only needs to input the license plate information of nearby vehicles into the terminal system to generate the shortest path information from the current location to the target vehicle location, and finally use the car-finding method based on the A* algorithm to find the optimal path.

In some embodiments of the present application, the specific method of establishing a fingerprint database is as follows:

Set up multiple beacons at different fixed positions in the space, select multiple sets of reference measurement points, each beacon broadcasts to the outside world at fixed time intervals, and use identification cards to collect multiple sets of RSSI information at different reference points;

The data collection model uses P ₀ to P _n to represent the positions of n+1 beacons, and R(0,0) to R(n,m) to represent the reference measurement points used to collect information. By recording the RSSI data and self-data collected at multiple reference measurement points, the data is transmitted to the service manager. Since the data signal is susceptible to environmental interference during the collection process, the collected data needs to be processed by Kalman filtering. Improve the data accuracy of the database. The data recording format is shown in Table 1.

Table 1 Fingerprint data record format table

In some embodiments of this application, the process of processing the collected data through Kalman filtering is as follows:

First, using the standardized formula Perform Gaussian distribution processing on the collected RSSI data, where,/> Indicates that in the data collected at the m-th reference measurement point, the i-th RSSI data max(RSSI) and min(RSSI) respectively represent/> The maximum and minimum values of , the processed RSSI data are iteratively operated through Kalman filtering to eliminate the influence of noise in the environment.

In some embodiments of this application, the training and prediction process of the LSTM network using the data set is as follows:

First, read the RSSI data processed by Kalman filtering, use the RSSI measurement data at time t as the input layer of the LSTM network, and use the coordinate points corresponding to the reference measurement point at time t in the two-dimensional coordinate system as the output of the LSTM network. layer, establish the corresponding regression model, and complete the training of the neural network. After training, input the verification data set S _n , predict the corresponding coordinates, and verify the prediction results.

Compared with the prior art, this application has at least the following beneficial effects:

In terms of system design, the parking lot is intelligent and unmanned. To meet the needs of parking lot management in the Internet era, use mobile payment technology and intelligent network equipment to achieve scientific billing and management, and achieve intelligent management of parking areas, vehicles, fees and other information. Intelligent recognition of license plates is realized through relevant interfaces, and cloud control enables autonomous entry and exit of vehicles in the booth-less mode, reducing labor costs. At the same time, managers can control and integrate relevant information. Equipped with a terminal system to realize parking guidance and reverse car search functions, it integrates information management, license plate recognition, parking guidance, and intelligent car search integrated platforms to improve corporate management capabilities, solve parking efficiency problems, and enhance user parking experience.

The LSTM network is used to achieve rapid target positioning. By filtering the signal data, the impact of environmental noise is reduced and the positioning accuracy is improved. Fast and efficient vehicle positioning, parking guidance route planning, and reverse vehicle search can be achieved. function, reducing the use of manual assistance and monitoring camera equipment and infrared equipment that the current existing system relies on, and realizing the intelligence of the system. Improve the efficiency of finding cars in large and medium-sized parking lots. Users only need to input the information of the nearest vehicle or location information into the terminal system at any location in a complex parking lot, and the system can generate the shortest path plan from the current location to the target parking space. , easy and fast to use, greatly improving the efficiency of parking guidance and automatic car search.

Description of the drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application. For those skilled in the art, other drawings can also be obtained based on these drawings without exerting creative efforts.

Figure 1 is an overall system structure diagram of some embodiments of the present application;

Figure 2 is a parking guidance diagram of some embodiments of the present application;

Figure 3 is a reverse car search flow chart of some embodiments of the present application;

Figure 4 is a data collection model diagram of some embodiments of the present application;

Figure 5 is a schematic plan view of a parking lot according to some embodiments of the present application;

Figure 6 is a generated weighted undirected graph according to some embodiments of this application;

Figure 7 is an algorithm logic flow chart of some embodiments of the present application.

Detailed ways

In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application. For those skilled in the art, other drawings can also be obtained based on these drawings without exerting creative efforts.

Some embodiments of the present application will be described in detail below with reference to Figures 1-7.

Some embodiments of the present application provide a parking guidance and reverse car-finding system based on the LSTM network, including: a management server, an identification card, a beacon location, a vehicle information identification device, an information management system, and a terminal system. The terminal system described above takes the mobile app system as an example.

Management server: used to receive and process identification card RSSI data and identity data, generate a location fingerprint database, use the data set in the location fingerprint database to train the LSTM network, use the LSTM network to predict the target location, store coordinate information, and To generate an undirected graph of parking rights, integrate the usage of parking spaces, vehicles parked in the parking spaces and other information, and send information such as the optimal path for parking and reverse car search to the information management system and terminal system.

Identification card: A wireless mobile tag equipped on the target vehicle that communicates with the beacon site and receives RSSI data. It is used to send the collected RSSI data and identity data to the management server.

Beacon location: It is located at a known fixed location in the parking lot area and broadcasts to the outside world at fixed time intervals. Its signal strength weakens as the distance increases. Obstacles in the broadcast environment will also affect the signal. Impact, the beacon location set needs to cover the entire parking lot area.

Vehicle information identification equipment: Set up at key locations such as entrances and exits, it is used to identify the license plate, color and other information of entering and exiting vehicles. At the same time, it records the identity data of the vehicle identification card and uploads the data to the management server for identification card location information, identity data, vehicle Data binding.

Information management system: used to manage user, vehicle, parking lot and other information, used to receive vehicle parking information, parking space usage information and vehicle information from the management server and display them in the view layer, send relevant instructions and data to the management server, and display Information such as the optimal path for parking and reverse car search enables data control and management of the information management system and terminal system.

Mobile app system: It interacts with users and is used to obtain user command information, communicate with the information management system and servers, display information such as the optimal path for parking and reverse car search, and mainly realizes user self-service car search, vehicle search and parking. Display of car guidance and other related information.

In some embodiments of the present application, the parking guidance process is also included, specifically including: starting to select a parking space → locating the current vehicle position → calculating the shortest path and planning the parking route → displaying the planned route in the terminal system → updating the vehicle position → judging whether to follow the Plan the route → If not, reposition the current vehicle position. If not, then determine whether it has reached the target parking space → If it reaches the target parking space, directly determine whether it has parked in an empty parking space. If not, record the last location information according to the server → Then determine whether Parked in an empty parking space→If yes, end, if not, reposition the current vehicle position.

In some embodiments of the present application, a reverse car search process is also included, which specifically includes: starting to input data, the data includes inputting any vehicle information or inputting target vehicle information → locating the current location of inputting any license plate information or inputting the target license plate The target location of the information → calculate the shortest path and plan the car-seeking route → display the planned route on the terminal system → update the current location → determine whether the planned route is followed → if not, reposition the current location; if so, then determine whether the target parking space is reached → if If the target parking space is not reached, then return to update the current position; if so, guide the vehicle to the exit → then record the vehicle information to release the parking space occupation → end.

In some embodiments of the present application, a parking guidance and reverse car-seeking method based on the LSTM network can be implemented using the above-mentioned parking guidance and reverse car-seeking system, involving the establishment of a fingerprint database and the processing of RSSI data. , generate a weighted undirected graph, and design a car-finding method based on the A* algorithm.

The steps are as follows:

Step 1: Select the ground projection as the two-dimensional coordinate plane, establish a two-dimensional coordinate system, store each parking space information and the corresponding center coordinate point in the service manager, and set multiple beacon points at different fixed positions to fix the The projected position of the beacon is its two-dimensional coordinate point. Select multiple sets of reference measurement points with known two-dimensional coordinates, and use identification cards to collect multiple sets of RSSI information at different reference points, where P ₀ to P _n represent n+1 The positions of the beacons, R(0,0) to R(n,m), represent different reference measurement points used to collect information.

Step 2: Apply the collected RSSI data to the standardized formula The collected RSSI data is processed with Gaussian distribution, and the processed RSSI data is iteratively calculated through the Kalman filter to eliminate the influence of noise in the environment.

Step 3: Read the RSSI data processed by Kalman filtering, input the sequence S _n = (X _i-M+1 ,X _iM ,,X _i ), and use the RSSI measurement data at time t as the input layer of the LSTM network. The corresponding coordinate point of the reference measurement point in the two-dimensional coordinate system at time t is used as the output layer of the LSTM network, and the corresponding regression model is established to complete the training of the neural network. After training, input the training data set S _n , predict the corresponding coordinates, and verify the output results.

Step 4: The management server stores the output results of the LSTM network as an adjacency matrix and generates a weighted undirected graph. The occupied parking spaces are activated nodes and the rest are inactive nodes. The management server stores each vertex information and edge information. The weight algorithm of the edges (a, b) between two vertices with adjacent relationships is: Set the weight of non-adjacent vertices to "∞" and the weight of its own vertices to "0". The storage form of the adjacency matrix is:

Step 5: When the user queries and selects an available parking space for parking or enters the license plate information of any nearby vehicle and the target license plate information for reverse car search, the management server will use the optimal path algorithm based on the A* algorithm based on the positioning information to pass The evaluation function f(n)=g(n)+h(n) calculates the shortest path between two vertices, where g(n) is the established cost function and h(n) is the estimated cost function. If three The vertices (a, b, c) of the adjacency relationship, when b is the current vertex, the evaluation function is:

By creating an open set and a closed empty set to store all vertices, select the nodes in the open and store them in closed. Repeat until open is empty. In closed, search in reverse order to the starting point to obtain the optimal path. The algorithm logic flow is shown in Figure 7 display, and finally transmit the relevant information to the information management system and terminal system display to complete the automatic parking and reverse car-finding process.

In some embodiments of this application, the algorithm logic includes the following process: create open and closed sets, add the starting node in open → then determine whether open is empty → if yes, fail, if not, select open f(n) minimum node best → then put it into the closed set → then determine whether best is the end point → if so, that is, best is the end point, then put it into the closed set, and then search the previous nodes in reverse to generate the optimal path, which is successful; if No, obtain the subsequent node and determine whether the subsequent node is empty. If the subsequent node is empty, return to determine whether open is empty. If the subsequent node is not empty, obtain any node s, and then calculate the path length G from best to node s ( s)=g(best)+c(best,s), determine whether s is in the open set. If not, add s to the open set. If so, determine whether G(s) is less than g(s). If it is less, delete it. s, then s is added to the open set. If G(s) is not less than g(s), return to determine whether the subsequent node is empty → add s to the open set and calculate f(s)=G(s)+h(s) →Return whether the subsequent node is empty.

The above are only preferred embodiments of the present application. It should be noted that those of ordinary skill in the art can also make several improvements and modifications without departing from the technical principles of the present application. These improvements and modifications It should also be regarded as the protection scope of this application.

Claims (10)

1. A LSTM network-based park guiding and reverse vehicle locating system, comprising:

the management server: the method comprises the steps of receiving beacon site data and self identification data from an identification card, processing the data, establishing a position fingerprint database, training an LSTM network, predicting coordinate information of a target site, storing the coordinate information, generating a weighted undirected graph, calculating a shortest path, integrating the use condition of a parking space and parking space vehicle information, and receiving or feeding back instructions and data of an information management system and a terminal system;

identification card: the wireless mobile tag is provided with a target vehicle and is used for receiving the beacon point signal data and transmitting the beacon point data and the self identification data to the management server;

beacon site: a plurality of known sites are fixed in the parking area, signals are broadcast outwards at fixed time intervals, and signal data are ensured to cover the whole area of the parking area;

vehicle information identifying apparatus: the system comprises a management server, a vehicle identification card information and a vehicle information monitoring server, wherein the management server is used for managing the vehicle identification card information and the vehicle information;

an information management system: the system is used for managing information of users, vehicles, parking lots and the like, sending related instructions and data to a management server, and realizing data control and management of the whole system;

terminal system: the information management system is used for collecting instruction information for users, communicating with the information management system and the server, and displaying information of reverse vehicle searching, vehicle searching and parking guiding routes of the users.

2. The LSTM network-based parking guidance and reverse vehicle finding system according to claim 1, wherein the vehicle information recognition device includes a monitoring camera, an identification card recognition device, a vehicle detector for recognizing and detecting information of passing vehicles.

3. The LSTM network-based parking guidance and reverse vehicle finding system according to claim 1, wherein the terminal system is a mobile phone app system.

4. The LSTM network-based parking guidance and reverse vehicle finding system of claim 1, wherein the information of the vehicle includes identification and detection of passing, stay presence, length, color, license plate number information of the vehicle.

5. A method for guiding parking and reverse vehicle searching based on LSTM network, which applies the system for guiding parking and reverse vehicle searching based on LSTM network as claimed in any one of claims 1-4, said method comprising the steps of:

establishing a two-dimensional coordinate system in a parking lot, storing parking space information and corresponding central coordinate points in a service manager, acquiring RSSI data of k beacons communicated with the parking space information by an identification card at each reference measurement point by setting a plurality of beacon points at different fixed positions, simultaneously identifying and monitoring identification card information of vehicles and vehicle information by a vehicle information identification device and uploading the identification card information and the vehicle information to a management server, filtering the RSSI data by the management server to form a fingerprint database, recombining and sequencing data in the database according to a time sequence to generate an input model sequence S _n As an input sequence of the LSTM network, taking the coordinate point as an output sequence of the LSTM network, establishing a corresponding regression model, and completing training of the neural network; inputting the processed data set, and predicting the coordinates of the target point;

the management server stores the output sequence of the trained LSTM network as an adjacent matrix, calculates the edge weight of each adjacent vertex according to a weight algorithm, and stores the vertex data and the edge data; generating a weighted undirected graph, and planning an optimal path by using a vehicle searching method based on an A-algorithm.

6. The LSTM network-based parking guidance and reverse vehicle finding method according to claim 5, wherein the collected data is normalized and noise reduction is performed by kalman filtering, and the normalization processing formula is:

wherein,represents the ith RSSI data, max (RSSI) and min (RSSI) respectively representing +.>And (3) carrying out iterative operation on the processed RSSI data through Kalman filtering.

7. The LSTM network-based parking guidance and reverse vehicle searching method of claim 5, wherein the LSTM network training process is to process the RSSI data set S _n As an input sequence, wherein S _n ＝(X _i-M+1 ,X _i-M ,…,X _i ) M is S _n Represents the use of M time instants for receiving the set of data, where X _i ＝(RSSI ₁ ,RSSI ₂ ,…,RSSI _k ) Representing data received at time i, k is X _i Represents k RSSI values received at time i, X _i Is S _n The recorded last position, i.e. the current position, corresponds to the position (x _i ,y _i ) Is a position of (c). And taking the two-dimensional coordinate position corresponding to the reference point as an output sequence, and establishing a corresponding regression model to complete the training of the LSTM network.

8. The LSTM network-based parking guidance and reverse vehicle finding method according to claim 5, wherein k pieces of target position information predicted through the LSTM network are stored as an adjacency matrix, and the stored form of the position adjacency matrix is:

calculating the edge weights of all adjacent vertexes according to a weight algorithm, setting the non-adjacent vertexes as "infinity", and setting the self-vertex weights as "0", wherein the weight algorithm is as follows:

weights (x) representing sides (a, b) of two vertices having an adjacency relationship _a ,y _a ) And (x) _b ,y _b ) Representing two-dimensional coordinates of two points.

9. The method for guiding and reversely searching vehicles based on LSTM network according to claim 5, wherein the user uses the terminal system to plan the optimal path by using the A-algorithm on the basis of storing the position information and generating the authorized map when searching and selecting the idle parking space to park or inputting any vehicle license plate information and target license plate information nearby to reversely search vehicles.

10. The method for guiding and reversely searching vehicles according to claim 5, wherein the management server calculates the shortest path of two vertexes in the authorized undirected graph according to the positioning information by using the optimal path algorithm based on the a-based algorithm according to the positioning information, through the valuation function f (n) =g (n) +h (n), wherein g (n) is a given cost function, h (n) is an estimated cost function, if three vertexes (a, b, c) with adjacency are provided, and if b is the current vertex, the valuation function is:

and (3) through creating an open set and a closed empty set for storing all vertexes, optimally storing nodes in the open into the closed, repeating until the open is empty, searching to a starting point in the closed through a reverse sequence to obtain an optimal path, and finally transmitting relevant information to a terminal system for displaying, so that automatic parking and reverse vehicle searching are completed.