A garbage throwing abnormal behavior monitoring method, device, equipment and medium

Abstract

The present application relates to the technical field of garbage classification, and discloses a garbage throwing abnormal behavior monitoring method, device, equipment and medium, first, a video screenshot is acquired, a target detection algorithm is used to judge the position of a human body and the position of a garbage can, when it is identified that there is contact between the human body and the garbage can, a garbage throwing abnormality detection model is started, whether there is a garbage throwing illegal behavior is judged, when a garbage throwing illegal behavior is detected, garbage throwing illegal behavior data is stored and alarm information is sent, realizing real-time identification of classified garbage throwing illegal behavior, introducing a multiple judgment mechanism, improving the accuracy of identification, and having low requirements for the detection environment, in the case of complex background in dark light environment, high identification accuracy can be realized, and the coverage and accuracy of identification and classification can be continuously expanded through data sample accumulation.

Description

Technical Field

[0001] This invention relates to the field of waste sorting technology, and in particular to a method, device, equipment and medium for monitoring abnormal waste disposal behavior. Background Technology

[0002] With the continuous expansion of urban scale, the amount of urban household waste has risen sharply. The problems of land erosion and environmental pollution caused by "garbage surrounding cities" not only seriously threaten public health, but have also become one of the factors affecting and restricting the sustainable development of cities in my country and even the world. Promoting household waste sorting and building a waste sorting and disposal system has become an urgent task and an inevitable trend. In order to solve the pain points of waste sorting that are common among Chinese residents, such as "unwillingness to sort, laziness in sorting, and difficulty in grasping sorting standards," and to break through the existing intelligent waste sorting devices on the market, which monitor and capture users' waste disposal process to trace abnormal waste disposal behavior, most of the existing methods for tracing abnormal behavior require manual work. It is necessary to check the video or photo captures of violations after discovering waste disposal violations. This method is inefficient and cannot deal with abnormal behavior in waste sorting in a timely manner, making it difficult to effectively solve the problems of littering, improper sorting, mixed disposal, and disorderly stacking. Summary of the Invention

[0003] In view of this, the present invention provides a method, device, equipment and medium for monitoring abnormal waste disposal behavior, which solves the technical problems of difficulty in identifying violations of classified waste disposal and low timeliness.

[0004] According to a first aspect of the present invention, a method for monitoring abnormal waste disposal behavior is provided, comprising:

[0005] Get video screenshots;

[0006] The location of the human body and the trash can are determined using an object detection algorithm;

[0007] When contact between a human body and a trash can is detected, the abnormal trash disposal detection model is activated to determine whether there is any illegal trash disposal behavior.

[0008] When illegal waste disposal is detected, the data on the illegal waste disposal is stored and an alarm message is sent.

[0009] According to a second aspect of the present invention, a device for monitoring abnormal waste disposal behavior is provided, comprising:

[0010] The acquisition module is used to acquire video screenshots;

[0011] The first judgment module is used to determine the location of the human body and the location of the trash can using a target detection algorithm;

[0012] The second judgment module is used to activate the abnormal waste disposal detection model when contact between a human body and the trash can is detected, and to determine whether there is any illegal waste disposal behavior.

[0013] The processing and execution module is used to store data on illegal waste disposal and send alarm information when illegal waste disposal is detected.

[0014] According to a third aspect of the present invention, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described method for monitoring abnormal waste disposal behavior.

[0015] According to a fourth aspect of the present invention, a computer-readable storage medium is provided, which stores a computer program that, when executed by a processor, implements the steps of the above-described method for monitoring abnormal waste disposal behavior.

[0016] By employing the above technical solution, this invention provides a method, device, equipment, and medium for monitoring abnormal waste disposal behavior. First, a video screenshot is acquired. Then, a target detection algorithm is used to determine the location of the human body and the location of the trash can. When contact between the human body and the trash can is detected, the abnormal waste disposal detection model is activated to determine whether there is any illegal waste disposal behavior. When illegal waste disposal behavior is detected, the data on illegal waste disposal behavior is stored and an alarm message is sent. This achieves real-time identification of illegal waste disposal behavior. The introduction of multiple judgment mechanisms improves the accuracy of identification and has low requirements for the detection environment. It can achieve high identification accuracy even in low-light environments and complex backgrounds. Furthermore, the coverage and accuracy of identification classification can be continuously expanded through the accumulation of data samples.

[0017] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention are listed below. Attached Figure Description

[0018] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this invention, illustrate exemplary embodiments of the invention and are used to explain the invention, but do not constitute an undue limitation of the invention. In the drawings:

[0019] Figure 1 A schematic diagram of the structure of the waste disposal abnormality monitoring device provided in an embodiment of the present invention is shown;

[0020] Figure 2This diagram illustrates a flowchart of a method for monitoring abnormal waste disposal behavior provided in an embodiment of the present invention.

[0021] Figure 3 This invention provides a schematic diagram of another method for monitoring abnormal waste disposal behavior.

[0022] Figure 4 A schematic diagram of a waste disposal abnormality monitoring device provided in an embodiment of the present invention is shown. Detailed Implementation

[0023] The present invention will be described in detail below with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in the embodiments of the present invention can be combined with each other.

[0024] The following describes a method for monitoring abnormal waste disposal behavior provided in this invention, with reference to specific embodiments.

[0025] like Figure 1This invention provides a camera that integrates multi-angle, multi-component, multi-functional, multi-algorithm, high-performance, and large-capacity capabilities for all-round tracking, capture, identification, positioning, analysis, and monitoring. It includes an automatically rotating base and a camera. The automatically rotating base has a rotating shaft embedded inside, and an automatic reset module is installed inside the rotating shaft. The camera or base contains an algorithm box (i.e., a processing circuit board). The algorithm box enables intelligent terminal analysis and cloud-based decision-making. It can detect data through local edge acquisition or transmit data to the cloud for processing, achieving integrated linkage of deep intelligence, target positioning, recognition and capture, multi-behavior analysis, and autonomous tracking. The camera contains a shooting component, and the shooting angle and position are controlled in real time through position adjustment and angle adjustment components, thereby achieving precise tracking and positioning. Operation can be controlled via a terminal, which drives the automatically rotating base to achieve rotation and lifting functions. Simultaneously, to ensure stability, the rotating shaft drives the base turntable to rotate, changing the lens angle, making the shooting angle adjustable and facilitating the capture of clearer, multi-angle photos of violations. In practical applications, facial images are recognized by cameras to identify different feature points on the face and establish a database. This enables functions such as facial information collection, comparison, search, and alerts, achieving intelligent image capture. Artificial intelligence computing resources can be deployed in the cloud or at the edge, combined with behavioral analysis technology and machine learning algorithms to model and analyze the behavior of target objects, predict and judge abnormal behaviors, and perform intelligent image capture. This is used for monitoring violations of waste sorting and disposal regulations. Specifically, a waste disposal abnormality monitoring method can be implemented to obtain video screenshots; target detection algorithms can be used to determine the location of the human body and the waste bin; when contact between the human body and the waste bin is detected, the waste disposal abnormality detection model is activated to determine whether there is a waste disposal violation; when a waste disposal violation is detected, the data is stored and an alarm message is sent, achieving real-time identification of waste sorting and disposal violations. The introduction of multiple judgment mechanisms improves the accuracy of identification and has low requirements for the detection environment. High recognition accuracy can be achieved even in low-light environments and complex backgrounds, and the coverage and accuracy of identification and classification can be continuously expanded through data sample accumulation.

[0026] The hardware module includes an automatic rotation drive base, inside which a rotating shaft is embedded, and inside the rotating shaft is an automatic reset module. The algorithm box includes intelligent terminal analysis and cloud decision-making. This device can utilize edge computing or cloud computing capabilities locally to achieve a series of functions, as follows:

[0027] The feature extraction function is performed by training a convolutional neural network algorithm model: extracting feature information of the target from video image sequences, and using the feature information to classify and analyze the target, automatically learning and extracting features from the image. This feature extraction function can be completed locally, i.e., at the edge computing part, or sent to the cloud.

[0028] Behavior classification function: Application of support vector machine classification algorithm model, naive Bayes algorithm model and neural network algorithm model; through machine learning classification algorithms, identify and judge the behavior of the target and classify it.

[0029] Recognition technology: Automatically recognizes captured information through cameras, and combines facial information, time and other information to achieve functions such as automatic capture, recognition, database entry, query and comparison.

[0030] Intelligent detection technology: By setting different detection areas and rules, it can automatically detect people, objects, etc., identify them, and intelligently capture images, thereby improving monitoring efficiency and accuracy.

[0031] 3D modeling technology: By acquiring 3D image data, it enables the modeling and analysis of target objects and performs intelligent image capture, featuring high precision, high robustness, and high level of recognition accuracy.

[0032] To more clearly describe the method for monitoring abnormal waste disposal behavior provided in this invention, it is now described in detail below:

[0033] Example 1

[0034] In practical applications, such as Figure 2 As shown, the method for monitoring abnormal waste disposal behavior provided in Embodiment 1 includes:

[0035] Step 201: Obtain video screenshots;

[0036] Step 202: Use a target detection algorithm to determine the location of the human body and the trash can;

[0037] Step 203: When contact between a human body and the trash can is detected, the abnormal trash disposal detection model is activated to determine whether there is any illegal trash disposal behavior.

[0038] Step 204: When a violation of garbage disposal rules is detected, store the garbage disposal violation data and send an alarm message.

[0039] Specifically, step 202 may include:

[0040] Step 202-1: Real-time detection of human and trash can positions using the YOLOv3 object detection algorithm;

[0041] In analyzing abnormal user behavior, accurate detection and localization of user targets are crucial while maintaining detection efficiency. The YOLO series of algorithms omits the proposal extraction process, using only a single convolutional neural network to predict and recognize the entire image, resulting in very high speed. Therefore, this embodiment selects the YOLOv3 algorithm as the target detection algorithm, ultimately outputting the top-left corner coordinates (xmin, ymin) and bottom-right corner coordinates (xmax, ymax) of the predicted bounding box for each user in the video, along with a prediction score. This localization is relatively accurate and can meet the needs of subsequent detection of illegal waste disposal.

[0042] Step 202-2: Calculate the area of ​​the overlapping region between the human body position and the trash can position.

[0043] Specifically, step 203 may include:

[0044] Step 203-1: When the area of ​​the overlapping region between the human body position and the trash can position is greater than a preset threshold, the abnormal trash disposal detection model is activated to detect whether there is any illegal trash disposal behavior.

[0045] Step 203-2: Collect continuous frame screenshots of human motion and perform human pose estimation using OpenPose;

[0046] OpenPose is a human pose recognition project, an open-source library developed by Carnegie Mellon University using convolutional neural networks and supervised learning, with Caffe as the framework. This library can estimate poses such as human movements, facial expressions, and finger movements. It is suitable for single and multi-person scenarios, exhibits excellent robustness, and is the world's first real-time multi-person 2D pose estimation application based on deep learning. To estimate actions during waste disposal, it is necessary to extract human skeletal features. For user skeleton extraction, OpenPose, a representative bottom-up approach, is chosen as the algorithm framework for extracting key points from the user's skeleton. Based on the already obtained user bounding box, key point information such as the user's body skeleton, face, and fingers is extracted. While maintaining the same image size, background information is separated, and only the human skeletal key point image is fed into a C2D network for classification.

[0047] The advantage of using this method in this embodiment of the invention is that it can ignore the interference of external information such as the user's clothing and the background of the delivery environment, so that the model can focus on learning human behavior and reduce the complexity of behavior recognition.

[0048] Step 203-3: Classify and recognize human movements based on human pose estimation and image classification models;

[0049] In this embodiment, the ResNet50 image classification model is used to classify and identify behaviors in images to determine whether they are suspected violations. This embodiment compares the performance of four classification algorithms—VGG16, VGG19, InceptionV3, and ResNet50—on datasets and test videos. The detection results show that while VGG16 and VGG19 can filter out some abnormal behaviors, they also have false negatives. InceptionV3 and ResNet50 models have no false negatives; however, InceptionV3 has too many false positives, failing to effectively reduce workload. Therefore, considering both accuracy and recall, the ResNet50 model performs best and best meets the requirements of the intelligent monitoring system. It can be used as a feature extractor and classifier to classify the input user human body keypoint image and detect suspected garbage disposal violations. The default output dimension of the ResNet50 model is 1000. This is because the pre-trained ResNet50 model is trained on the ImageNet dataset, which contains 1000 categories. In addition, supervised training of ResNet50 is required before this step. In the actual data collection and training process, photos were taken from different angles under different lighting conditions, and the illegal behavior of littering was labeled in order to carry out fully supervised learning. The specific training method is existing technology and will not be described again.

[0050] Step 203-4: Based on the human motion classification and recognition results, determine whether there is any suspected violation.

[0051] In this process, classification is performed based on the vector representation output by ResNet50, and the classification corresponds to various abnormal behaviors in garbage disposal.

[0052] Step 203-5: Input the continuous frame screenshots of human movement suspected of violating regulations into the garbage disposal anomaly detection model to determine whether there is any violation of garbage disposal regulations.

[0053] The garbage disposal anomaly detection model is specifically a C3D+attention+SVM model (an attention mechanism is added to the C3D 3D convolutional neural network to extract image features, which are then classified and detected using linear SVM to identify illegal garbage disposal behavior; the features extracted by the C3D network are dimensionality-reduced using PCA, and then these features are fed into the SVM for prediction). Step 203-5 may specifically include: inputting continuous frame screenshots of suspected illegal human movement into the C3D network for feature extraction; and using the extracted features to determine whether illegal garbage disposal behavior exists using the garbage disposal anomaly detection SVM (Support Vector Machine) classification model. In this embodiment, the C3D network includes a hybrid attention module, which is a channel- and spatial hybrid attention module. C3D is a behavior recognition method based on convolutional neural networks that uses 3D convolution operations to extract features from video sequences. Compared with traditional behavior recognition methods, C3D can better handle time-series data and has higher recognition accuracy. The C3D architecture includes multiple convolutional layers and pooling layers, as well as multiple fully connected layers and a softmax classifier.

[0054] This invention introduces a garbage disposal violation detection method based on a C3D+attention+SVM mechanism. In this method, a C3D network is used as the basic C3D neural network structure, and improvements have been made to optimize it based on processing capabilities. For example... Figure 3As shown, video screenshots captured by the camera form continuous video frames, and multiple video keyframes undergo further preprocessing (image scaling, human detection, environmental background removal, etc.). The network uses 3×3×3 convolutional kernels with a stride of 1×1×1, and pooling layers are 2×2×2 except for the first layer which is 1×2×2. The number of filters is modified to 32, 64, 128, 128, 256, 256, 256, and 256, followed by two 1024-dimensional fully connected layers. To avoid overfitting, a Dropout layer with a dropout probability of 0.5 is added after each fully connected layer. For probability output, a softmax layer follows each fully connected layer, using the softmax function to convert the output value into a probability representation within the 0-1 range. The network input is a video segment of size c×l×h×w, where h and w are the height and width of the video keyframe, respectively, c is the number of channels, and l is the number of consecutive video frames. The improved C3D network takes a 3×8×208×208 image sequence as input, which is eight consecutive frames of 208×208 RGB channel images. During feature extraction, because C3D operations cannot distinguish between background features and human action features, the model's recognition is easily affected by environmental factors. To address this issue, this embodiment adds a channel-space hybrid attention module that integrates spatial and channel elements to the C3D network to reduce the model's sensitivity to changes in the background environment. This results in a C3D+attention+SVM garbage disposal violation recognition method that incorporates a C3D-attention mechanism. The garbage disposal violation dataset is constructed based on the environmental characteristics of different garbage disposal methods and the feasibility of algorithm implementation. During application, common normal behaviors and garbage disposal violations need to be defined. For example, normal behaviors include continuously opening multiple garbage lids, continuously depositing garbage into multiple garbage bins, and depositing garbage after pausing briefly in a garbage bin. Garbage disposal violations can include depositing garbage into only one garbage bin at a time, depositing garbage quickly without pausing in front of the bin, and depositing garbage outside the bin.

[0055] In this embodiment, compared to ResNet50's use of 2D images for identifying illegal garbage disposal, C3D uses action recognition, which has higher time complexity, effectively reducing false detections. This is more pronounced in lower-resolution videos because 2D image methods are easily affected by human skeleton extraction results when video resolution decreases. C3D, on the other hand, extracts motion information from consecutive frames, reducing the impact of resolution degradation. Furthermore, the introduction of an attention module in C3D+attention+SVM also reduces false detections to some extent. The attention mechanism makes the model focus more on the changes in the action of the garbage target, thus learning the relevant features of garbage disposal behavior more accurately, thereby reducing false detections of normal behavior and improving the model's accuracy. This embodiment of the invention employs...

[0056] The ResNet50+C3D+attention+SVM method for secondary violation detection first uses 2D ResNet50 to detect suspected littering violations, and then uses C3D+attention+SVM for action recognition to detect more accurate littering violations. This method can filter out some false detections from both methods and select behaviors that are jointly identified as littering violations, thus achieving the highest accuracy. As the training data increases, it can identify all littering violations without false detections.

[0057] Step 204 may specifically include:

[0058] Step 204-1: When a violation of garbage disposal regulations is detected, extract and store the video clip of the violation.

[0059] Step 204-2: Send a violation alarm message to the audio-visual equipment near the trash can to remind the user to confirm that the trash has been disposed of correctly.

[0060] In this embodiment, the SVM model is trained by collecting sample data to train an SVM classification model for detecting abnormal waste disposal. The specific training process involves the basic training steps of SVM, which will not be repeated here. The overall training steps may include:

[0061] Obtain sample data on actual violations of waste disposal regulations;

[0062] Supervised learning is performed based on sample data of actual waste disposal violations to obtain an SVM classification model for waste disposal anomaly detection.

[0063] During the actual data collection and training process, videos were taken from different angles under varying lighting conditions. These included scenes with both single and multiple individuals, used to construct a dataset of illegal garbage disposal behavior. Frames were extracted from the captured videos, one frame every 8 or 16 frames. Then, OpenPose was used to extract the human skeleton and garbage can images for each user. The background was set to black, retaining only the keyframe images of the user's human skeleton and the edge images of the garbage can. All images were standardized to a size of 208×208, resulting in a dataset of 6541 images (user skeleton images + garbage can images). Of these, 2245 depict illegal garbage disposal and 1233 depict normal behavior. The garbage disposal violation video dataset was preprocessed by cropping the extracted frames from each user's image and converting them to grayscale, with a standardized image size of 208×208.

[0064] To ensure each video segment represents a user's continuous actions, consecutive frames are treated as a single video segment to demonstrate the occurrence of improper waste disposal, thus constructing a video dataset of improper waste disposal behavior. The dataset contains a total of 12,455 video segments, of which 8,404 represent improper waste disposal behavior and 4,051 represent normal behavior. Improper waste disposal behavior includes disposing of waste in only one bin at a time, failing to stop in front of the bin before quickly disposing of waste, and disposing of waste outside the bin. Normal behavior includes opening multiple bin lids consecutively, disposing of waste in multiple bins consecutively, and stopping briefly in one bin before disposing of waste.

[0065] This invention provides a method for monitoring abnormal waste disposal behavior. First, a video screenshot is acquired. Then, a target detection algorithm is used to determine the location of the human body and the location of the trash can. When contact between the human body and the trash can is detected, an abnormal waste disposal detection model is activated to determine whether there is any illegal waste disposal behavior. When illegal waste disposal behavior is detected, the data on illegal waste disposal behavior is stored and an alarm message is sent. This method achieves real-time identification of illegal waste disposal behavior. The introduction of multiple judgment mechanisms improves the accuracy of identification and has low requirements for the detection environment. It can achieve high identification accuracy in low-light environments and complex backgrounds. Furthermore, the coverage and accuracy of identification classification can be continuously expanded through the accumulation of data samples.

[0066] In actual waste sorting and disposal, users' activity range is relatively fixed, and their movements do not vary significantly. Extracting every frame of the video for detection would greatly reduce the detection speed. To quickly identify potential waste disposal violations, an inter-frame difference method is used. This method involves summing the absolute values ​​of the differences between corresponding pixels to achieve detection, and the result is recorded as the inter-frame difference intensity. Maintaining the original order of the video frames, frames with inter-frame difference intensities higher than their immediate preceding and following frames are selected as keyframes for detecting abnormal behavior, used for subsequent detection of waste disposal violations.

[0067] Furthermore, as Figures 2 to 3 In a specific implementation of the method, this embodiment of the invention provides a device for monitoring abnormal waste disposal behavior, such as... Figure 4 As shown, the device includes:

[0068] Module 410 is used to acquire video screenshots;

[0069] The first judgment module 420 is used to determine the position of the human body and the position of the trash can using a target detection algorithm;

[0070] The second judgment module 430 is used to activate the garbage disposal anomaly detection model when contact between a human body and the garbage can is detected, and to determine whether there is any illegal garbage disposal behavior.

[0071] The processing and execution module 440 is used to store data on illegal waste disposal and send alarm information when illegal waste disposal is detected.

[0072] This invention provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it performs the following steps:

[0073] Get video screenshots;

[0074] The location of the human body and the trash can are determined using an object detection algorithm;

[0075] When contact between a human body and a trash can is detected, the abnormal trash disposal detection model is activated to determine whether there is any illegal trash disposal behavior.

[0076] When illegal waste disposal is detected, the data on the illegal waste disposal is stored and an alarm message is sent.

[0077] This invention provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, performs the following steps:

[0078] Get video screenshots;

[0079] The location of the human body and the trash can are determined using an object detection algorithm;

[0080] When contact between a human body and a trash can is detected, the abnormal trash disposal detection model is activated to determine whether there is any illegal trash disposal behavior.

[0081] When illegal waste disposal is detected, the data on the illegal waste disposal is stored and an alarm message is sent.

[0082] It should be noted that the functions or steps that can be implemented by the computer-readable storage medium or computer device described above can be referred to the relevant descriptions on the server side and client side in the foregoing method embodiments. To avoid repetition, they will not be described one by one here.

[0083] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided by this invention can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0084] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0085] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for monitoring abnormal waste disposal behavior, characterized in that, include: Get video screenshots; Real-time detection of human and trash can locations using the YOLOv3 object detection algorithm; Calculate the area of ​​the overlapping region between the human body and the trash can. When the area of ​​the overlapping region between the human body and the trash can exceeds a preset threshold, the garbage disposal anomaly detection model is activated to detect whether there is any illegal garbage disposal behavior; continuous frame screenshots of human movement are collected, and human pose is estimated using OpenPose; based on the human pose estimation and image classification model, human movements are classified and identified. Based on the human motion classification and recognition results, it is determined whether there is any suspected violation; the continuous frame screenshots of the human motion of the suspected violation are input into the garbage disposal anomaly detection model to determine whether there is any garbage disposal violation. Specifically, the continuous frame screenshots of the human motion of the suspected violation are input into the C3D network for feature extraction; based on the extracted features, the garbage disposal anomaly detection SVM classification model is used to determine whether there is any garbage disposal violation. When illegal waste disposal is detected, the data on the illegal waste disposal is stored and an alarm message is sent.

2. The method according to claim 1, characterized in that, The C3D network includes a hybrid attention module, which is a hybrid attention module that combines channel and spatial attention.

3. The method according to claim 1, characterized in that, The step of storing data on illegal waste disposal and sending an alarm message when illegal waste disposal is detected includes: When illegal garbage disposal is detected, video clips of the illegal garbage disposal are extracted, output, and stored. Send violation alerts to audio-visual devices near the trash can to remind users to confirm that the trash has been disposed of correctly.

4. The method according to claim 1, characterized in that, Before the step of inputting continuous frame screenshots of the suspected violation of human movement into the garbage disposal anomaly detection model to determine whether there is a violation of garbage disposal regulations, the following steps are included: Obtain sample data on actual violations of waste disposal regulations; Supervised learning is performed based on sample data of actual waste disposal violations to obtain an SVM classification model for waste disposal anomaly detection.

5. A device for monitoring abnormal waste disposal behavior, characterized in that, include: The acquisition module is used to acquire video screenshots; The first judgment module is used to detect the position of human body and trash can in real time using the YOLOv3 target detection algorithm; Calculate the area of ​​the overlapping region between the human body and the trash can. The second judgment module is used to activate the garbage disposal anomaly detection model to detect whether there is any illegal garbage disposal behavior when the area of ​​the overlapping area between the human body position and the garbage can position is greater than a preset threshold; to collect continuous frame screenshots of human movement and to estimate human pose through OpenPose; and to classify and identify human movements based on human pose estimation and image classification model. Based on the human motion classification and recognition results, it is determined whether there is any suspected violation; the continuous frame screenshots of the human motion of the suspected violation are input into the garbage disposal anomaly detection model to determine whether there is any garbage disposal violation. Specifically, the continuous frame screenshots of the human motion of the suspected violation are input into the C3D network for feature extraction; based on the extracted features, the garbage disposal anomaly detection SVM classification model is used to determine whether there is any garbage disposal violation. The processing and execution module is used to store data on illegal waste disposal and send alarm information when illegal waste disposal is detected.

6. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the waste disposal abnormal behavior monitoring method as described in any one of claims 1 to 4.

7. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the waste disposal abnormal behavior monitoring method as described in any one of claims 1 to 4.

Images