CN110466911A - Automatic sorting garbage bin and classification method - Google Patents

Abstract

An automatic sorting trash bin and a sorting method. The present invention designs an automatic sorting trash can based on the YOLOV3 technical framework, so as to classify and recycle garbage, reduce garbage pollution, and recycle waste with maximum cost. The present invention is designed with an upper and lower double-layer trash can. The upper box body is provided with a garbage throwing inlet and a camera is installed. The upper top of the upper box body is provided with a partition for installing solar panels and batteries to meet the power supply requirements of the system. . The camera captures the input garbage, and uses the recognition algorithm to identify and classify the types of garbage. The lower box is designed as a double box or multiple boxes according to the garbage category. The connection between the two boxes is equipped with a mechanical and electric control part. Two steering gears are used to control the rotation of the sorting baffle and the control baffle to automatically drive The garbage falls into the sorting box corresponding to its type, and the invention can realize automatic garbage sorting.

Description

Automatic sorting trash can and sorting method

technical field

The invention relates to garbage sorting technology, in particular to an automatic sorting garbage bin and a sorting method.

Background technique

In contemporary society, energy conservation and emission reduction is a topic that everyone pays close attention to, especially in terms of garbage classification and recycling, which is one of the most prominent issues in society at present. Although most of the garbage bins are used in life, according to the survey and life experience, most users still throw garbage at will, and will not throw it into the corresponding garbage bin according to the type of garbage. In the existing sorting garbage bins, all kinds of garbage are still in an indistinguishable state of being mixed together. This will make it impossible to sort and recycle the garbage, resulting in a waste of resources. If sorting is carried out manually, a lot of human and financial resources will be wasted.

Therefore, there is a need for a garbage sorting device with a simple structure and a fast and efficient sorting process to solve the above problems.

Contents of the invention

Aiming at the deficiencies of the prior art, the present invention provides an artificial intelligence-based automatic sorting trash can, its circuit system and a sorting method, so as to automatically identify and classify the garbage put into the trash can. The present invention specifically adopts the following technical solutions.

First, in order to achieve the above object, an automatic sorting trash can is proposed, which includes: a barrel body, at least one garbage throwing inlet is arranged on the upper part of its side wall; a first steering gear is fixed on the outside of the side wall of the barrel body, It is connected with a horizontal central axis column, and the horizontal central axis column penetrates the inside of the cylinder body along the diameter direction of the cylinder body, and the horizontal central axis column is driven to rotate by the first steering gear; the control baffle , comprising at least two pieces, the control baffle is respectively connected to the horizontal central axis column, and divides the inside of the barrel body into upper and lower boxes; the control baffle can be connected with the horizontal central axis column is an axis, flipped with the rotation of the horizontal central axis column to open or close the bottom of the upper box, so that the objects contained in the upper box fall into the lower box; the second steering gear is set On the upper top of the upper box, it is connected with a vertical central axis column, and the vertical central axis column penetrates the inside of the upper box along the axial direction of the cylinder body; the vertical central axis column is formed by the The second steering gear is driven to rotate; the classification baffle, the middle of which is connected to the lower end of the vertical central axis column, divides the upper box into two parts, and the classification baffle can take the vertical central axis column as the axis , rotate with the rotation of the vertical central axis column to drive the objects contained in the upper box to move to the top of the control baffle corresponding to its classification; the lower box is based on the horizontal central axis column The boundary is divided into a plurality of classification boxes, each of the control baffles corresponds to one of the classification boxes, and the upper parts of the classification boxes are respectively closed by a control baffle; the camera is arranged in the upper box The upper top of the body is used to collect the image of the object contained in the upper box; the control board is used to obtain the image collected by the camera, identify the type of the object in the image, and drive the second object according to the type of the object. The steering gear drives the sorting baffle through the vertical central axis column, moves the object contained in the upper box to the top of the control baffle corresponding to its classification, and then drives the first steering gear to pass through the classification baffle. The horizontal central axis column drives the control baffle to turn over, so that the object falls into the classification box corresponding to the type of the object under the control baffle.

The above-mentioned automatic sorting trash can also includes a first gear and a second gear; the first steering gear is fixedly connected with the first gear, and the first gear meshes with the second gear to turn the first gear The driving force output by a steering gear is transmitted to the second gear; the horizontal central axis column includes two, and each of the horizontal central axis columns is respectively connected with a control baffle; the two horizontal central axis columns are respectively Connected with the first gear and the second gear, the two horizontal central shaft columns are respectively driven by the first gear and the second gear to drive the connected control baffle to turn over to open or close the upper box The bottom of the body, so that the objects contained in the upper box fall into the classification box corresponding to the type of the object in the lower box.

Optionally, in the above-mentioned automatic sorting trash can, the garbage throwing inlets include two respectively arranged on the upper sides of the two control baffles.

Optionally, in the above-mentioned automatic sorting trash bin, it also includes: LED lights, batteries, solar panels; the LED lights, batteries, solar panels and the control panel are all accommodated on the top of the upper box. top.

At the same time, in order to achieve the above object, the present invention also provides an automatic garbage sorting method, which is used in the above-mentioned automatic sorting garbage can, the steps of which include: the first step, when an object enters the upper box, through the camera The image of the object is collected; the second step is to perform dehazing and clear enhancement processing on the image; the size of the image is adjusted to an integer multiple of 32; the third step is to circulate the processed image through the Yolo v3 method The product neural network is trained to carry out category identification to the object in the image after described processing; The fourth step, drives described second steering gear according to the object category identified in the third step to drive through the vertical central axis column The sorting baffle moves the object contained in the upper box to the top of the control baffle corresponding to its classification, and then drives the first steering gear to drive the control baffle through the horizontal central axis column Turn over so that the object falls into the classification box corresponding to the type of the object under the control baffle.

Optionally, in the above-mentioned automatic garbage classification method, the specific steps of the third step include: step 301, performing grid division on the image obtained in the second step; step 302, using the method of k-means or IOU Obtain the prior frame anchor corresponding to the above grid; step 303, use the Darknet network for training, use the entire image obtained in the second step above as the input of the network, and perform regression calculation to obtain the regression calculation at the output layer of the Darknet network The position of the bounding box and the category to which it belongs, and its accuracy rate is calculated; step 304, the position of the bounding box obtained above, its category and accuracy rate are filtered using NMS (non-maximum value suppression method) Processing, filtering out the bounding boxes whose accuracy rate is lower than the set threshold, and outputting the result of category recognition according to the position of the bounding box corresponding to the reserved bounding box and the category to which it belongs.

Optionally, in the above automatic garbage classification method, in the training of the circular convolutional neural network, the training parameters are set as: decay=0.005, learning_rate=0.001, steps=500000; the circular convolutional neural network adopts sum-squared error loss Design a loss function; the loss function includes position coordinate prediction, feature value prediction with objects, feature value prediction without objects, and category prediction.

Beneficial effect

The present invention designs an automatic sorting garbage can device based on the YOLOV3 technical framework, so as to classify and recycle garbage, reduce garbage pollution, and recycle waste at the maximum cost. The present invention is designed with an upper and lower double-layer trash can. The upper box body is provided with a garbage throwing inlet and a camera is installed. The upper top of the upper box body is provided with a partition for installing solar panels and batteries to meet the power supply requirements of the system. . The camera captures the input garbage, and uses the recognition algorithm to identify and classify the types of garbage. The lower box is designed as a double box or multiple boxes according to the garbage category. The connection between the two boxes is used as a mechanical and electric control part. Two steering gears are used to control the rotation of the control baffle and the classification baffle to automatically drive the garbage to fall. into the classification box corresponding to its type, the present invention can realize the automatic classification of garbage.

Additional features and advantages of the invention will be set forth in the description which follows, and in part will be apparent from the description, or may be learned by practice of the invention.

Description of drawings

The accompanying drawings are used to provide a further understanding of the present invention, and constitute a part of the description, and together with the embodiments of the present invention, are used to explain the present invention, and do not constitute a limitation to the present invention. In the attached picture:

Fig. 1 is the geometric relationship diagram of the Anchor prediction frame position used in the YOLOV3 object recognition method adopted in the present invention;

Fig. 2 is the flowchart of the automatic garbage sorting method based on artificial intelligence of the present invention;

Fig. 3 is the recognition result figure that the present invention utilizes YOLOV3 object recognition method to test;

Fig. 4 is the sectional view of the automatic sorting trash can based on artificial intelligence of the present invention;

Fig. 5 is the overall structure schematic diagram of the automatic sorting trash can based on artificial intelligence of the present invention;

Fig. 6 is a schematic diagram of the driving mode of the first steering gear in the automatic sorting trash can based on artificial intelligence of the present invention;

Fig. 7 is a schematic diagram of the driving mode of the second steering gear in the automatic sorting garbage bin based on artificial intelligence of the present invention;

Fig. 8 is a block diagram of the circuit system in the automatic sorting trash can based on artificial intelligence of the present invention.

In the figure, 1 represents the cylinder body; 11 represents the first classification box; 12 represents the second classification case; 2 represents the first steering gear; 21 represents the first gear; 22 represents the second gear; 3 represents the horizontal central axis column; 4 represents the control baffle; 5 represents the second steering gear; 6 represents the vertical central axis column; 7 represents the sorting baffle; 8 represents the garbage throwing entrance.

Detailed ways

In order to make the purpose and technical solutions of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings of the embodiments of the present invention. Apparently, the described embodiments are some, not all, embodiments of the present invention. Based on the described embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

Those skilled in the art can understand that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. It should also be understood that terms such as those defined in commonly used dictionaries should be understood to have a meaning consistent with the meaning in the context of the prior art, and will not be interpreted in an idealized or overly formal sense unless defined as herein explain.

The meanings of "inside and outside" in the present invention refer to that relative to the cylinder body itself, the direction pointing to the inner box is inward, and vice versa is outward; it is not a specific limitation to the device mechanism of the present invention.

The meaning of "connection" in the present invention may be a direct connection between components or an indirect connection between components through other components.

The meaning of "up and down" in the present invention means that when the user is facing the automatic sorting trash can, the direction from the steering gear to the sorting box is down, and vice versa, it is up, not to the device of the present invention Institution-specific restrictions.

The algorithm identification part of the present invention first collects pictures of common recyclable and non-recyclable garbage, makes a data sample library, and marks the data sample library for training. In the subsequent application process, through the image processing algorithm, the above training results are used to identify and classify garbage. The recognition algorithm here intends to use the yolov3 algorithm, but it is not limited to this algorithm.

The present invention is mainly based on the YOLOV3 object recognition algorithm to develop a set of automatic intelligent classification garbage can device to help people solve the problem of garbage classification. The difficulty lies in the detection of garbage types. The current mainstream target detection algorithms mainly include Faster R-CNN, YOLO, SSD, etc., YOLO has a good effect on recognition accuracy and speed, so the present invention uses the YOLO network model to design an automatic classification device for recyclable garbage and non-recyclable garbage, and the YOLO series of algorithms have their own advantages from v1 to v3. Advantages and disadvantages, YOLO's detection method adopts the end-to-end idea, and uses the Darknet network for training. The model uses the entire image as the input of the network. It uses the regression method to directly return the position of the bounding box (boundary box) and its position in the output layer. The category it belongs to is different from traditional recognition algorithms. For example, RCNN adopts the idea of proposal+classifier, but puts the step of extracting proposal in CNN, while YOLO adopts the idea of direct regression.

Specifically, referring to FIG. 4 , the automatic sorting trash bin provided by the present invention utilizes the above-mentioned method to identify the type of garbage to classify the garbage. The whole is divided into three parts: the box structure, the mechanical transmission part, and the photovoltaic energy supply part:

The upper part of the cylinder body 1 is provided with at least one, or two opposite garbage disposal inlets 8; the cylinder body 1 is provided with a structure of two boxes, and the whole is cylindrical, separated by a control baffle, which is divided into There are upper and lower floors (upper box and lower box); the garbage disposal entrance 8 is set at 10cm away from the control baffle:

The upper box includes the garbage sorting device and the CPU control part of the power supply. The garbage sorting device is composed of a sorting baffle and a control baffle. Its structural diagram is shown in Figure 5, the control baffle 4 includes at least two pieces, which are respectively arranged below the two garbage throwing inlets 8; the control baffle 4 is respectively connected to the horizontal central axis column 3, and the cylinder The interior of the body 1 is divided into upper and lower boxes; the control baffle 4 can take the horizontal central axis column 3 as the axis, and turn over with the rotation of the horizontal central axis column 3 to open or close the the bottom of the upper box so that the objects contained in the upper box fall into the lower box. Specifically, two semicircular control baffles controlled by a horizontal central axis column controlled by fixed gears are placed horizontally, which are controlled by the first steering gear 2 .

The lower box is a two-box structure. It is separated from the center by a vertically placed central partition, and the central partition is placed in a direction parallel to the horizontal central axis column. A semicircular garbage collection bin is placed between the two boxes for sorted garbage collection.

The above-mentioned first steering gear 2 is fixed on the outside of the side wall of the cylinder body 1, and is connected with a horizontal central axis column 3, and the horizontal central axis column 3 penetrates the cylinder body along the diameter direction of the cylinder body 1 1, the horizontal central shaft column 3 is driven by the first steering gear 2 to rotate.

The above-mentioned second steering gear 5 is arranged on the upper top of the upper box, and it is connected with a vertical central axis column 6, and the vertical central axis column 6 penetrates the upper box along the axial direction of the cylinder body 1 inside; the vertical central axis column 6 is driven to rotate by the second steering gear 5; wherein, the first steering gear 2 and the second steering gear 5 can be selected as MG995 type steering gear

Classification baffle 7, with reference to Fig. 7, its middle part connects the lower end of described vertical central axis column 6, and described upper box body is divided into two parts, and described classification baffle 7 can take described vertical central axis column 6 as axis , to rotate with the rotation of the vertical central axis column 6, to drive the objects contained in the upper box to move to the top of the control baffle 4 corresponding to its classification; that is to say, the classification baffle is controlled by the second rudder Machine 5 is controlled, placed vertically above the control baffle, and the placement direction is parallel to the horizontal central axis column. A vertical central axis column set in the center of the classification baffle is fixed on the rotating shaft of the steering gear 2, and the classification is controlled by controlling the steering gear steering. The flapper turns to sort the trash.

The lower box is divided into a plurality of classification boxes with the horizontal central axis column 3 as the boundary, each of the control baffles 4 corresponds to one of the classification boxes, and the upper parts of the classification boxes are respectively Closed by a control baffle 4;

a camera, arranged on the upper top of the upper box, for collecting images of objects contained in the upper box;

Referring to Figure 8, due to the diversification of the application site and environment of the garbage can, the circuit part of the outdoor garbage can can also be provided with LED lights, batteries, solar panels and control panels on the upper top of the upper box. Wherein, the partition plate installed on the top of the trash can and the steering gear 2 is used to install the above-mentioned solar panel as a photovoltaic power supply, and the above-mentioned partition board is also used to install the control panel part. The output terminal of the solar panel is connected to the storage battery, and the storage battery supplies power for the camera, the LED lamp, the control board, the first steering gear 2 and the second steering gear 5; thus, the present invention utilizes solar energy During the day, the battery panel receives solar radiation energy and converts it into electrical energy and stores it in the battery through the controller, which supplies the system with self-energy, realizes energy self-sufficiency, and can still work without external power supply;

The control board includes: an image receiving unit, which is connected to the camera, and is used to obtain the image collected by the camera; an image processing unit, which is used to process the image and identify the type of object in the image; a steering gear control unit, It is connected to the first steering gear 2 and the second steering gear 5, and is used to drive the second steering gear 5 according to the type of the object recognized by the image processing unit so as to drive all the steering gears through the vertical central axis column 6. The classification baffle 7 moves the object contained in the upper box to the top of the control baffle 4 corresponding to its classification, and then drives the first steering gear 2 to drive the object through the horizontal central axis column 3 The control baffle 4 is overturned so that the object falls into the classification box corresponding to the type of the object under the control baffle 4 .

As shown in FIG. 6, in a preferred implementation mode, the horizontal central shaft column 3 is designed to be two, respectively controlling two control baffles, which are respectively driven by the first gear 21 and the second gear 22. . in:

The first steering gear 2 is fixedly connected to the first gear 21, and the first gear 21 meshes with the second gear 22 to transmit the driving force output by the first steering gear 2 to the second gear. gear 22;

The horizontal central axis column 3 includes two, and each of the horizontal central axis columns 3 is respectively connected with a control baffle plate 4; the two horizontal central axis columns 3 are respectively connected with the first gear 21 and the second gear 22 connection, the two horizontal central shaft columns 3 are respectively driven by the first gear 21 and the second gear 22 to drive the connected control baffle 4 to turn over to open or close the bottom of the upper box, so that The objects contained in the upper box fall into the sorting boxes corresponding to the types of objects in the lower box.

The above-mentioned mechanical transmission is controlled by a steering gear. The steering gear is a micro servo system. Its working principle is that the control circuit receives the control pulse from the signal source and drives the motor to rotate; the gear set reduces the speed of the motor by a large multiple and reduces the speed of the motor. The output torque is amplified by the response multiple, and then output; the potentiometer and the final stage of the gear set rotate together to measure the rotation angle of the steering gear shaft; the circuit board detects and judges the rotation angle of the steering gear according to the potentiometer, and then controls the steering gear to rotate to the target angle or maintain at the target angle. The use of steering gear control has a good effect on control accuracy and speed. What the present invention adopts is MG995 steering gear, and MG995 steering gear has good stability and higher control precision, and can stably and accurately control the sorting baffle to turn to sorting and control the opening and closing of the baffle. The control baffle is controlled by the first steering gear 2, and two gears are used to mesh the central shafts of the two control baffles to achieve the effect of synchronous control. As shown in Figure 6, the second steering gear 5 controls the sorting baffle and is responsible for correctly sorting the garbage. Fix the shaft of the steering gear and the central shaft column of the sorting baffle, and control the work of the sorting baffle by controlling the rotation of the steering gear

The steps for the control panel of the above-mentioned garbage bin to classify the garbage can be referred to in Figure 2, including:

In the first step, when an object enters the upper box, the image of the object is collected by the camera;

The second step is to perform dehazing and clear enhancement processing on the image; adjust the size of the image to an integer multiple of 32;

The third step is to carry out cyclic convolutional neural network training on the processed image by the YOLO v3 method, so as to carry out category recognition to the objects in the processed image; the steps are specifically:

Step 301, performing grid division on the image obtained in the second step;

Step 302, using the method of k-means or IOU to obtain the prior frame anchor corresponding to the above grid;

Step 303, use the Darknet network for training, use the entire image obtained in the second step above as the input of the network, and perform regression calculation to obtain the position of the bounding box and its category at the output layer of the Darknet network , calculate its accuracy rate;

Step 304, use NMS to filter the position of the bounding box, its category and accuracy rate obtained above, filter out the bounding box whose accuracy rate is lower than the set threshold, and filter out the bounding box according to the retained bounding box The position of the bounding box corresponding to the bounding box, the category to which it belongs, and the output category recognition result;

The fourth step is to drive the second steering gear 5 according to the type of object identified in the third step to drive the sorting baffle 7 through the vertical central axis column 6, and move the object contained in the upper box to the Above the control baffle 4 corresponding to its classification, and then drive the first steering gear 2 to drive the control baffle 4 to turn over through the horizontal central axis column 3, so that the object falls under the control baffle 4 Inside the classification box corresponding to the type of the object.

The above-mentioned trash can mainly uses the image target detection framework to identify and detect the types of garbage in the trash can. A camera is installed at the entrance of the trash can. When an object is thrown in, the camera is touched to take a photo. For detection, since the application object is a trash can, the speed of its operation is very important, and it is necessary to respond in a timely manner. The present invention has guaranteed the real-time property of algorithm, below the implementation process of the present invention is described further, specifically as follows:

1. When it is detected that there is garbage thrown in, touch the camera to take a picture, and perform dehazing and clear enhancement processing on the picture, so as to obtain a better quality picture, so that the later training network can obtain better image features and increase the accuracy of the result Spend. The image enhancement network here can use the GAN network, but is not limited to this method.

2. The present invention is an automatic sorting garbage bin based on artificial intelligence. The present invention adopts two means of taking photos and searching the Internet to collect common recyclable garbage (plastic bottles, cloth, books, iron wire) and non-recyclable garbage (fruit peels). , Broken glass, leftovers, napkins) a total of 2667 pictures constitute the data sample library of the present invention, and the data set is tagged to make the training result more accurate.

3. After preparing the data set, use the circular convolutional neural network for training. Some of the training parameters are set as follows, decay=0.005, learning_rate=0.001, steps=500000, and the training is performed on the GPU. In the second step, the picture size needs to be processed as a multiple of 32, because YOLO v3 has 5 downsampling, each sampling step is 2, so the maximum stride of the network (the stride refers to the input size of the layer divided by the output ) is 2^5=32. In the implementation, the most important thing is how to design the loss function. The algorithm of the present invention uses sum-squared error loss to design the loss function. The final loss function is as follows:

In this loss function, it is mainly divided into four parts, coordinate prediction, eigenvalue (confidence) prediction with object (object), eigenvalue (confidence) prediction without object (object) and category prediction, using loss function for constraint training network.

4. Using the YOLOV3 algorithm to detect, it is necessary to obtain the anchor (a priori frame). Specifically, a new anchor can be obtained by using k-means, IOU and other methods on the collected data set, but it is not limited to these two methods. The Anchor mechanism refers to setting some reference frame shapes and sizes for each grid. It only needs to refine the reference frame during detection, instead of the position regression of the entire image.

To use the Anchor mechanism, first determine the width and height dimensions of the reference frame. Although the network training process will also adjust the width and height dimensions of the frame, and finally get an accurate frame, but if a more representative reference frame is selected from the beginning, it will be easier for the network to detect the exact position. The convolutional neural network will predict 4 values for each bounding box on each cell, namely the coordinates (x, y) and the width w and height h of the target, which are recorded as xt, yt, wt, ht respectively. If the center of the object is offset (xc, yc) in the cell relative to the upper left corner of the image, and the anchor box has height and width wp, hp, the corrected bounding box is shown in Figure 1. in

b _x =σ(t _x )+c _x

b _y =σ(t _y )+c _y

5. Use the YOLOv3 algorithm to identify the extracted pictures, and mark the category and location of the object. The specific method is as follows:

Use NMS (Non-Maximum Suppression Method) for filtering processing. After convolutional network training, when testing, the class information predicted by each grid is multiplied by the confidence information predicted by the bounding box to get each Category information and accuracy information (class-specific confidence score) of a bounding box:

The first item on the left side of the equation is the category information predicted by each grid, and the second and third items are the confidence predicted by each boundingbox. After obtaining the category information and accuracy information (class-specificconfidence score) of each box, set the threshold, filter out the boxes with low scores, and perform NMS processing on the reserved boxes to obtain the final detection result.

Therefore, the present invention cleverly uses the principle of multi-scale detection of the YOLOv3 algorithm, and the detection accuracy of garbage is very high, especially when the garbage is small, yolov3 has a good accuracy in identifying small objects, so that there will be no missed detection of garbage, The situation of under-inspection occurs, and the method of using the anchor box increases the recall without changing the mAP, and the use of the new network structure reduces the calculation by 33%. The speed is faster than other detection systems (FasterR-CNN, ResNet, SSD), which improves the recall rate and accuracy rate, improves the accuracy of positioning, and maintains the accuracy of classification. With the deepening of the network and the combination of multiple models, the training accuracy can be improved. At the same time, the data of the picture is enhanced to make the extracted features more significant and the picture quality higher. The recognition results are shown in Figure 3.

Thus, after the database is trained, the present invention calls the generated model file, and applies the yolov3 algorithm framework to carry out the garbage picture detection test. The specific test results are shown in FIG. , fruit peel, plastic bag and cloth were tested. The specific recognition diagram is shown in Figure 3. The blue box indicates the position of the object, and the red word in the upper left corner has the label of the object and the corresponding accuracy rate. It can be seen that the recognition effect is good. , can well meet the working requirements of the smart trash can.

The above is only the embodiment of the present invention, and its description is relatively specific and detailed, but it should not be construed as limiting the patent scope of the present invention. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of the present invention, and these all belong to the protection scope of the present invention.

Claims (10)

1. An automatic sorting trash can, characterized in that, comprising: The cylinder body (1), the upper part of the side wall is provided with at least one garbage disposal inlet (8); The first steering gear (2) is fixed on the outside of the side wall of the barrel (1); A control baffle (4), the control baffle (4) is respectively connected to the horizontal central axis column (3), and divides the inside of the barrel (1) into upper and lower boxes; the control The baffle (4) can take the horizontal central axis column (3) as the axis, and turn over with the rotation of the horizontal central axis column (3) to open or close the bottom of the upper box, so that the upper box The objects contained in the body fall into the lower box; The second steering gear (5) is arranged on the upper top of the upper box; A classification baffle (7), the middle of which is connected to the lower end of the vertical central axis column (6), dividing the upper box into two parts; The lower box is divided into a plurality of sorting boxes with the horizontal central axis column (3) as the boundary, each of the control baffles (4) corresponds to one of the sorting boxes, and each of the sorting boxes The upper part of the body is closed by a control baffle (4); a camera, arranged on the upper top of the upper box, for collecting images of objects contained in the upper box; The control board is used to obtain the image collected by the camera, identify the type of object in the image, and drive the second steering gear (5) according to the type of the object to drive the classification through the vertical central axis column (6) The baffle (7), which moves the object contained in the upper box to the top of the control baffle (4) corresponding to its classification, and then drives the first steering gear (2) to pass through the horizontal central axis column (3) Drive the control baffle (4) to turn over, so that the object falls into the classification box corresponding to the type of the object under the control baffle (4). 2. The automatic classification garbage bin as claimed in claim 1, characterized in that, The first steering gear (2) is connected with a horizontal central axis column (3), and the horizontal central axis column (3) penetrates the inside of the cylinder body (1) along the diameter direction of the cylinder body (1), The horizontal central shaft column (3) is driven to rotate by the first steering gear (2). 3. The automatic classification garbage bin as claimed in claim 1, characterized in that, The second steering gear (5) is connected with a vertical central axis column (6), and the vertical central axis column (6) penetrates the interior of the upper box along the axial direction of the cylinder body (1); The vertical central shaft column (6) is driven to rotate by the second steering gear (5). 4. The automatic classification garbage bin as claimed in claim 1, characterized in that, The sorting baffle (7) can take the vertical central axis column (6) as an axis and rotate with the rotation of the vertical central axis column (6), so as to drive the objects contained in the upper box to move to The top of the control baffle (4) corresponding to its classification. 5. The automatic sorting trash can according to claim 1, further comprising a first gear (21) and a second gear (22); The first steering gear (2) is fixedly connected with the first gear (21), and the first gear (21) meshes with the second gear (22) to turn the first steering gear (2) The output driving force is transmitted to the second gear (22); The horizontal central axis columns (3) include two, and each of the horizontal central axis columns (3) is respectively connected with a control baffle (4); the two horizontal central axis columns (3) are respectively connected to the The first gear (21) is connected to the second gear (22), and the two horizontal central shaft columns (3) are respectively driven by the first gear (21) and the second gear (22) to drive the connected The control baffle (4) turns over to open or close the bottom of the upper box, so that the objects contained in the upper box fall into the sorting boxes corresponding to the types of objects in the lower box. 6. The automatic sorting trash can according to claim 1-2, characterized in that, the garbage throwing inlets (8) include two respectively arranged on the upper sides of the two control baffles (4). 7. The automatic sorting trash bin according to claims 1-3, further comprising: LED lights, batteries, and solar panels; The LED lamp, the storage battery, the solar panel and the control panel are all contained in the upper top of the upper box. 8. An automatic garbage sorting method, characterized in that it is used for the automatic sorting garbage can according to claims 1 to 7, and its steps include: In the first step, when an object enters the upper box, the image of the object is collected by the camera; The second step is to perform dehazing and clear enhancement processing on the image; adjust the size of the image to an integer multiple of 32; In the third step, the image is trained with a circular convolutional neural network to identify the types of objects in the image; The fourth step is to drive the second steering gear (5) according to the type of object identified in the third step to drive the sorting baffle (7) through the vertical central axis column (6), and place all objects in the upper box The contained object moves above the control baffle (4) corresponding to its classification, and then drives the first steering gear (2) to drive the control baffle (4) through the horizontal central axis column (3) ) is overturned so that the object falls into the classification box corresponding to the type of the object under the control baffle (4). 9. The automatic garbage classification method according to claim 5, wherein the specific steps of the third step include: Step 301, performing grid division on the image obtained in the second step; Step 302, using the method of k-means or IOU to obtain the prior frame anchor corresponding to the above grid; Step 303, use the Darknet network for training, use the entire image obtained in the second step above as the input of the network, and perform regression calculation to obtain the position of the bounding box and its category at the output layer of the Darknet network , calculate its accuracy rate; Step 304, use NMS to filter the position of the bounding box, its category and accuracy rate obtained above, filter out the bounding box whose accuracy rate is lower than the set threshold, and filter out the bounding box according to the retained bounding box The position of the bounding box corresponding to the bounding box and the category it belongs to output the result of category recognition. 10. The automatic garbage classification method according to claim 5-6, wherein, in the training of the cyclic convolutional neural network, the training parameters are set to: decay=0.005, learning_rate=0.001, steps=500000; the cyclic The convolutional neural network uses sum-squared error loss to design a loss function; the loss function includes position coordinate prediction, feature value prediction with objects, feature value prediction without objects, and category prediction.