Garbage overflow intelligent monitoring system based on YOLO under view angle of unmanned aerial vehicle

Abstract

The invention discloses a YOLO-based garbage overflow intelligent monitoring system under the view angle of an unmanned aerial vehicle, and belongs to the technical field of target detection and environmental protection, the system comprises an unmanned aerial vehicle monitoring module, a YOLO model training module, a model deployment module, a back-end program module and a user interface module; wherein the unmanned aerial vehicle monitoring module is used for acquiring video data and picture data in real time; the YOLO model training module is used for training a YOLO deep learning model based on the picture data; the model deployment module is used for deploying the trained YOLO deep learning model to a local terminal; the back-end program module is used for carrying out real-time garbage overflow detection on a detection site based on the trained YOLO deep learning model; and the user interface module is used for displaying the garbage overflow detection result to a user and carrying out data analysis and subsequent garbage overflow processing decision.

Description

Technical Field

[0001] The present invention belongs to the technical field of target detection and environmental protection, and in particular relates to a YOLO-based intelligent monitoring system for garbage overflow from the perspective of a drone. Background Art

[0002] With the acceleration of urbanization, the amount of domestic waste has increased dramatically, and waste management has become a severe challenge in urban management. Traditional garbage monitoring methods mainly rely on manual inspections, which are not only inefficient but also easily restricted by human factors and environmental conditions. In recent years, with the development of computer vision and deep learning technology, intelligent monitoring systems based on image recognition have become a research hotspot. As an efficient target detection framework, the YOLO algorithm has been widely used in the field of image processing for its fast and accurate detection capabilities. However, the application of the YOLO algorithm to garbage overflow monitoring from the perspective of drones is still in the exploratory stage, especially in real-time data processing and model deployment, there are still many technical difficulties.

[0003] As a cross-platform graphical user interface application framework, PyQt5 has been widely used in desktop application development with its rich components and good user experience. Combined with the user interface developed with PyQt5, data visualization and interactive operations can be effectively realized, providing users with an intuitive operating environment. However, in the field of environmental protection, especially in the application of garbage overflow monitoring system, how to combine PyQt5 with the YOLO algorithm to build a stable and efficient monitoring platform is a problem worthy of in-depth study. In addition, as an important tool for data collection, the stability and data collection quality of drones in complex environments are also key factors affecting the performance of the monitoring system. Therefore, the development of an intelligent garbage overflow monitoring system from the perspective of drones that integrates the YOLO algorithm and PyQt5 technology has important practical significance for improving garbage disposal efficiency and reducing environmental protection risks. Summary of the invention

[0004] In order to solve the above technical problems, the present invention proposes a YOLO-based intelligent garbage overflow monitoring system from the perspective of a drone to solve the problems existing in the above-mentioned prior art.

[0005] To achieve the above object, the present invention provides a YOLO-based intelligent monitoring system for garbage overflow from the perspective of a drone, comprising: a drone monitoring module, a YOLO model training module, a model deployment module, a backend program module and a user interface module;

[0006] Wherein, the drone monitoring module is used to collect video data and image data in real time;

[0007] The YOLO model training module is used to train a YOLO deep learning model based on the image data;

[0008] The model deployment module is used to deploy the trained YOLO deep learning model to the local terminal;

[0009] The backend program module is used to perform real-time garbage overflow detection on the detection location based on the trained YOLO deep learning model;

[0010] The user interface module is used to display the garbage overflow detection results to the user, and to perform data analysis and subsequent garbage overflow processing decisions.

[0011] Optionally, the drone monitoring module includes a first unit and a second unit;

[0012] Wherein, the first unit is used to collect picture data;

[0013] The second unit is used to collect real-time video data.

[0014] Optionally, the YOLO model training module includes a picture preprocessing unit and a model training unit;

[0015] The image preprocessing unit is used to label and preprocess the image data for target detection, and construct a training set and a test set for garbage overflow detection;

[0016] The model training unit is used to build a YOLO garbage overflow detection model, and train the YOLO garbage overflow detection model based on the training set to obtain a trained YOLO garbage overflow detection model.

[0017] Optionally, the YOLO garbage overflow detection model includes a SimAM attention mechanism with slicing operations and several large object detection heads.

[0018] Optionally, the expression of the SimAM attention mechanism is:

[0019]

[0020] In the formula, ω i is the attention weight of the ith pixel, k is a normalization constant, N i is the set of neighboring pixels of the ith pixel, s(f i ,f j ) is the similarity between the i-th pixel and the j-th pixel.

[0021] Optionally, the loss calculation formula of the YOLO garbage overflow detection model is:

[0022] L total=L bce +L cls +L dfl

[0023] in,

[0024] L bce = -ylog(p)-(1-y)log(1-p)

[0025]

[0026] L dfl =-(i+1-y)log(y i )-(yi)log(y i+1 )

[0027] Where, L total is the total loss value of the model, L bce is the standard binary cross entropy loss, L cls To calculate the intersection-over-union ratio between the predicted box and the target real box, L dfl is the distribution focus loss, y is the real target category, p is the predicted target category, c is the diagonal length of the minimum enclosing rectangular box, ρ is the Euclidean distance between the two, b is the coordinate of the center point of the predicted box, and b gt is the coordinate of the center point of the target real box, αν is the aspect ratio hyperparameter of the predicted box and the target real box, IoU is the intersection over union ratio of the predicted box and the target real box, y i is the predicted sequence value.

[0028] Optionally, the user interface module includes a decision support tool, which includes: a route planning unit, a system decision unit, a model configuration unit and a detection result unit.

[0029] The present invention also provides a computer terminal device, comprising:

[0030] one or more processors;

[0031] A memory, coupled to the processor, for storing one or more programs;

[0032] When the one or more programs are executed by the one or more processors, the one or more processors implement a YOLO-based intelligent garbage overflow monitoring system from the perspective of a drone.

[0033] The present invention also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, a YOLO-based intelligent garbage overflow monitoring system from the perspective of a drone is implemented.

[0034] Compared with the prior art, the present invention has the following advantages and technical effects:

[0035] The present invention successfully built an efficient real-time monitoring and identification system for garbage overflow based on the YOLO algorithm. With its excellent target detection and real-time processing capabilities, the system can capture garbage information instantly, providing a basis for decision-making for rapid response and precise cleaning for relevant departments. This not only significantly improves the efficiency of garbage detection, but also greatly reduces the investment of manpower and time, bringing unprecedented convenience and efficiency to the maintenance of the street environment, and improving the overall effect of the cleaning work. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0037] Figure 1 It is a schematic diagram of the process of a YOLO-based intelligent monitoring system for garbage overflow from the perspective of a drone in an embodiment of the present invention;

[0038] Figure 2 A schematic diagram of a YOLOv8 network framework of a YOLO-based garbage overflow intelligent monitoring system from the perspective of a drone in an embodiment of the present invention;

[0039] Figure 3 A schematic diagram of the structure of a large target detection head according to an embodiment of the present invention;

[0040] Figure 4 Schematic diagram of the network structure of the SimAM attention mechanism of an embodiment of the present invention;

[0041] Figure 5 This is a YOLOv8 training evaluation curve diagram of an embodiment of the present invention. DETAILED DESCRIPTION

[0042] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0043] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0044] Embodiment 1

[0045] like Figure 1 As shown, this embodiment provides a YOLO-based intelligent monitoring system for garbage overflow from the perspective of a drone, including:

[0046] The real-time video data and image data of garbage overflow are obtained through drones; the image data is used for training the YOLO target detection model, and the real-time video data is used for garbage overflow detection;

[0047] The target detection data is annotated and preprocessed based on the image data taken by the drone, and the training set and test set for garbage overflow detection are constructed. The training set images have two labels: "overflow" and "no overflow".

[0048] Using the training set to train the pre-built YOLO garbage overflow detection model, to obtain a trained garbage overflow detection model;

[0049] Deploy the trained YOLO garbage overflow detection model to the local terminal;

[0050] Based on the backend program, the real-time video data of the drone at the required detection location is obtained, and the video stream of the garbage overflow to be detected is input into the target YOLO garbage overflow detection model to obtain the real-time prediction and recognition results of the garbage overflow to be detected;

[0051] Finally, the recognition results are transmitted to the visual user interface for data analysis and subsequent garbage overflow processing decisions based on user configuration.

[0052] The present invention proposes a YOLO-based intelligent monitoring system for garbage overflow from the perspective of a drone. The system collects video data in real time through a drone monitoring module, and combines a YOLO model training module and a model deployment module to achieve efficient detection of garbage overflow. The system uses a YOLOv8 network as the basis, integrates a SimAM attention mechanism and a large target detection head, and optimizes the recognition ability of multi-scale garbage overflow from the top-down perspective of a drone. During the training process, the detection accuracy and robustness of the model are further improved through a positive and negative sample allocation strategy and an improved loss function. The back-end program module is responsible for processing real-time video data and transmitting the detection results to a user interface module developed based on PyQt5 to achieve data visualization and rapid decision support. In addition, the present invention also includes an electronic device that integrates a drone, a memory, a processor, and a computer program to provide hardware support for the intelligent monitoring system. Overall, the present invention not only improves the automation level and efficiency of garbage monitoring, reduces manpower and time costs, but also provides an innovative technical solution for smart city environmental management.

[0053] Specifically, the YOLO-based garbage overflow intelligent monitoring system from the perspective of drones includes a drone monitoring module, a YOLO model training module, a model deployment module, a backend program module, and a user interface module. Among them, the drone monitoring module is used to collect video data and image data in real time; the YOLO model training module is used to train the YOLO deep learning model based on image data; the model deployment module is used to deploy the trained YOLO deep learning model to the local terminal; the backend program module is used to perform real-time garbage overflow detection on the detection location based on the trained YOLO deep learning model; the user interface module is used to display the garbage overflow detection results to the user, and perform data analysis and subsequent garbage overflow treatment decisions.

[0054] Furthermore, the drone monitoring module includes a first unit and a second unit; wherein the first unit is used to collect image data; and the second unit is used to collect real-time video data.

[0055] In this embodiment, the drone in the drone monitoring module uses a DJI Mini 4Pro shooting system. DJI Mini 4Pro is a high-performance drone launched by DJI Innovations. It is equipped with omnidirectional obstacle avoidance technology for the first time in the Mini series, providing users with a safer and more convenient flight experience. This drone is known for its lightweight body design, which is easy to carry and operate, and has strong endurance.

[0056] DJI Mini 4Pro, with its 1 / 1.3-inch image sensor and three-axis mechanical gimbal, provides 4K / 60fps HDR video recording and 4K / 100fps slow-motion video, as well as up to 48-megapixel Raw photo shooting, ensuring excellent and stable image quality. Its omnidirectional binocular vision system and 3D infrared sensor enhance environmental perception, while smart features such as omnidirectional smart follow, master lens, and one-click short film simplify the shooting process and make single-person operation more convenient. O4 image transmission technology supports 20 kilometers of high-definition image transmission, expanding the control range.

[0057] In this embodiment, the YOLO model training module, model deployment module and backend program module are all implemented based on the YOLOv8 network structure, and the SimAM attention mechanism with slicing operation and the large target detection head are added, so that the YOLOv8 network can adapt to the top-down perspective, multi-target, and multi-scale garbage overflow detection environment under the perspective of the drone.

[0058] Furthermore, the YOLO model training module includes an image preprocessing unit and a model training unit; wherein the image preprocessing unit is used to label and preprocess the image data for target detection data, and construct a training set and a test set for garbage overflow detection; the model training unit is used to build a YOLO garbage overflow detection model, and train the YOLO garbage overflow detection model based on the training set to obtain a trained YOLO garbage overflow detection model.

[0059] Specifically, the image data obtained by the drone is used to train the YOLO garbage overflow detection model. First, the target detection data is annotated and preprocessed according to the image data taken by the drone, and the training set and test set for garbage overflow detection are constructed. Among them, the training set images have two labels: "overflow" and "no overflow". Multiple labeled garbage overflow image datasets are Input image data is x i And the corresponding TXT format truth label y i .

[0060] Based on the YOLOv8 basic network structure, the SimAM attention mechanism with slicing operation and the large target detection head are added to build the YOLO garbage overflow detection model, so that the YOLOv8 network can adapt to the top-down perspective, multi-target, and multi-scale garbage overflow detection environment under the perspective of drones. The overall framework of the YOLO garbage overflow detection model is shown in the figure below. Figure 2 shown.

[0061] In standard target detection, when there are multiple targets in the data set, overly large targets may be missed or have poor detection results. YOLO has three detection heads by default, which can detect targets at multiple scales, and are used to detect targets larger than 8X8, 16X16, and 32X32. In order to solve the problem of large trash can overflow and garbage dump missed detection in garbage overflow detection, a large target detection head is added. The Backbone part remains unchanged, and the Neck and Header model structures are adjusted to detect targets larger than 64X64. The entire network structure includes four detection heads: P3, P4, P5, and P6. The improved large target detection head network structure is shown in the figure below. Figure 3 shown.

[0062] In a further embodiment, by adding the SimAM attention mechanism to the Backbone of the YOLOv8 network, the network has efficient feature extraction capabilities and anti-deformation capabilities when facing drone bird's-eye view pictures. At the same time, in order to enhance the feature extraction capabilities of small targets, the slicing operation is introduced to improve the SimAM attention mechanism. Unlike the existing channel attention and spatial attention modules, the SimAM attention module infers 3D attention weights for feature maps without adding parameters to the original network. SimAM is a simple and parameter-free attention module that optimizes the energy function to find the importance of each neuron by inferring the 3D attention weights of the feature map, thereby improving performance on various visual tasks.

[0063] In one embodiment, the calculation formula of the SimAM attention mechanism is:

[0064]

[0065] In the formula, ω i is the attention weight of the i-th pixel, k is a normalization constant, N i is the set of neighboring pixels of the ith pixel, s(f i ,f j ) is the similarity between the i-th pixel and the j-th pixel.

[0066] In order to enhance the feature extraction capability of small targets, a slicing operation is introduced. When the feature map is cut into different blocks, the texture features of large targets will obviously affect the average value of the block, resulting in a reduction in the additional weighting they receive. After merging the feature maps, large targets can still maintain high recognizability or even be further enhanced; while the features of small targets are more different from the local average, thus obtaining more weighting, and the features of small targets are enhanced. That is, the improved SimAM attention module ensures that both large and small targets receive fair attention and enhancement. The network structure of the SimAM attention mechanism with slicing operation, such as Figure 4 shown.

[0067] The YOLO garbage overflow detection model is trained using the constructed garbage overflow training set to obtain the trained garbage overflow detection model. During the training process, YOLOv8n is selected as the pre-training model, which not only ensures the lightweight of the model but also the real-time performance of the model detection. After the trained YOLO garbage overflow detection model is packaged into an executable program and deployed to the local terminal, the real-time video data of the drone at the required detection location is obtained based on the back-end program to perform real-time intelligent monitoring of garbage overflow.

[0068] The calculation formula for the positive and negative sample allocation strategy is:

[0069] t=s α +u β;

[0070] Where s is the predicted score corresponding to the labeled category, u is the IOU of the predicted box and the target box, and α / β are weight hyperparameters.

[0071] The calculation formula of YOLOv8 loss function is:

[0072] L total =L bce +L cls +L dfl ;

[0073] L bce = -ylog(p)-(1-y)log(1-p);

[0074]

[0075] L dfl =-(i+1-y)log(y i )-(yi)log(y i+1 );

[0076]

[0077] Where, L bce is the standard binary cross entropy loss, L cls To calculate the intersection-over-union ratio between the predicted box and the target real box, L dfl is the distribution focus loss, y is the real target category, p is the predicted target category, c is the diagonal length of the minimum enclosing rectangular box, ρ is the Euclidean distance between the two, b is the coordinate of the center point of the predicted box, and b gt is the coordinate of the center point of the target real box, αν is the aspect ratio hyperparameter of the predicted box and the target real box, IoU is the intersection over union ratio of the predicted box and the target real box, y i is the predicted sequence value, (w gt ,h gt ), (w,h) are the width and height of the target real box and the predicted box respectively.

[0078] In this embodiment, the YOLO-based garbage overflow intelligent monitoring system from the perspective of the drone transmits the recognition results to the visual user interface through real-time video data capture and target detection algorithm. The interface not only displays the real-time detection of garbage overflow, but also provides a series of functions to assist users in making decisions. Users can set the detection threshold of garbage overflow, video data source, detection method and detection model, etc. In addition, the system also supports the recording and analysis of historical data to help users identify the patterns and causes of garbage overflow, thereby optimizing garbage management strategies.

[0079] Furthermore, the user interface provides decision support tools, such as garbage removal route planning and resource allocation suggestions, to improve response efficiency. Users can also manually confirm or correct the detection results to ensure the accuracy of the system. The system is designed with regular report generation functions, including key indicator statistics and trend analysis, to provide data support for policy making and resource planning. In addition, the system integrates a user feedback mechanism to ensure that the monitoring system can be continuously iterated and optimized to meet the actual needs of users and ultimately achieve automation and intelligence of the garbage disposal process. The specific visualization interface is implemented through PyQt5 under the Qt framework combined with the Python language. PyQt5 is a Python language implementation of the Qt framework, developed by Riverbank Computing, and is one of the most powerful GUI libraries. PyQt5 provides a well-designed window control set that can run cross-platform on Linux, Window and Mac OS systems, using IDE for interface design and automatically generating executable Python code.

[0080] PyQt5 consists of a series of Python modules, with more than 620 classes, 6,000 functions and methods. The main modules include Qt Core, Qt GUI and Qt Widgets.

[0081] The present invention also proposes an electronic device, including a drone, a memory, a processor, and a computer program stored in the memory and executable on the processor, to implement a YOLO-based intelligent garbage overflow monitoring system from the perspective of any of the above-mentioned drones.

[0082] The effect of the present invention can be further illustrated by the following tests.

[0083] 1. Test conditions

[0084] The test hardware platform parameters are shown in Table 1, and the software platform parameters are shown in Table 2.

[0085] Table 1

[0086] Drones DJI Mini4 Pro System CPU Inteli7-12700F Memory 32GB Graphics card model NVIDIA GeForce RTX 3060 Graphics card memory 12GB

[0087] Table 2

[0088] operating system Windows 10 64 bit Compiler Pycharm2023.1 CUDA Version 12.1 QT Version 13.0.2

[0089] In this embodiment, YOLOv8n is used as a pre-trained model loaded in the training phase. The model is trained for 100 rounds using a stochastic gradient descent optimizer with a momentum of 0.937 and a weight decay of 0.0005. The batch size is 16, and the learning rate is initially set to 0.01. All input images are standardized and cropped to a size of 1024×1024 pixels. In terms of data enhancement, all training images are randomly rotated at a small angle to expand their diversity.

[0090] 2. Test system

[0091] The system of the present invention is a YOLO-based intelligent monitoring system for garbage overflow from the perspective of a drone.

[0092] 3. Test content and test results

[0093] The test experiment used 4062 garbage overflow pictures taken by drones for model training, and 1017 garbage overflow pictures for model testing. In order to verify the performance of the proposed YOLO garbage overflow detection model, this embodiment uses precision, recall, average precision mean (0.5) and average precision mean (0.5:0.95) to evaluate the performance of the model. In addition, the model training process is visualized through the training evaluation curve. The specific model performance of garbage overflow detection in this embodiment and the performance improvement results after adding the SimAM attention mechanism with slicing operation and the large target detection head are shown in Table 3. The YOLO garbage overflow detection model training evaluation curve visualization diagram is shown in Table 3. Figure 5 shown.

[0094] Table 3

[0095]

[0096] It can be seen that compared with the traditional garbage overflow detection network, the YOLO garbage overflow detection model of the present invention can quickly perform garbage overflow detection under drone aerial photography conditions, with a garbage overflow detection accuracy of 80.93%, and the overall average accuracy is stable at around 83%, which is more than 10% higher than the basic YOLOv8 detection performance. At the same time, the YOLO garbage overflow detection model was tested for real-time video detection, and the model response time was extremely short, achieving the optimal detection accuracy and the optimal real-time detection performance.

[0097] from Figure 5 It can be seen that the system of the present invention has significant advantages in garbage overflow detection. The system uses the YOLOv8 network as the basis, integrates the SimAM attention mechanism and the large target detection head, and optimizes the recognition ability of multi-scale garbage overflow under the top-down perspective of the drone. During the training process, the detection accuracy and robustness of the model are further improved through the positive and negative sample allocation strategy and the improved loss function.

[0098] In the present invention, the proposed YOLO-based garbage overflow intelligent monitoring system from the perspective of a drone is intended to solve the challenges of garbage management in the process of urbanization. The system integrates a drone monitoring module, a YOLO model training module, a model deployment module, a backend program module, and a user interface module developed based on PyQt5 to achieve full process automation from real-time video data acquisition to target detection, data analysis, and processing decisions. The system uses YOLOv8 as the basic network, combined with the SimAM attention mechanism and a large target detection head, to optimize the detection capability of multi-scale garbage overflow from a bird's-eye view. During the training process, the detection accuracy and robustness of the model are improved through the positive and negative sample allocation strategy and the improved loss function. Finally, the system can feed back the recognition results to the user interface in real time, assisting users in making fast and effective garbage disposal decisions. In addition, the present invention also includes an electronic device that integrates a drone, a memory, a processor, and a corresponding computer program to provide hardware support for the intelligent monitoring system. Overall, the present invention not only improves the efficiency of garbage monitoring, reduces manpower and time costs, but also provides an innovative technical path for the environmental management of smart cities.

[0099] Embodiment 2

[0100] This embodiment also provides a computer terminal device, including:

[0101] one or more processors;

[0102] A memory, coupled to the processor, for storing one or more programs;

[0103] When the one or more programs are executed by the one or more processors, the one or more processors implement a YOLO-based intelligent garbage overflow monitoring system from the perspective of a drone.

[0104] Embodiment 3

[0105] This embodiment also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, a YOLO-based intelligent garbage overflow monitoring system from the perspective of a drone is implemented.

[0106] The above are only preferred specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A YOLO-based intelligent garbage overflow monitoring system from the perspective of a drone, characterized in that: include: Drone monitoring module, YOLO model training module, model deployment module, backend program module and user interface module; Wherein, the drone monitoring module is used to collect video data and image data in real time; The YOLO model training module is used to train a YOLO deep learning model based on the image data; The model deployment module is used to deploy the trained YOLO deep learning model to the local terminal; The backend program module is used to perform real-time garbage overflow detection on the detection location based on the trained YOLO deep learning model; The user interface module is used to display the garbage overflow detection results to the user, and to perform data analysis and subsequent garbage overflow processing decisions.

2. The YOLO-based intelligent garbage overflow monitoring system from the perspective of a drone according to claim 1 is characterized in that: The drone monitoring module includes a first unit and a second unit; Wherein, the first unit is used to collect picture data; The second unit is used to collect real-time video data.

3. The YOLO-based intelligent garbage overflow monitoring system from the perspective of a drone according to claim 1 is characterized in that: The YOLO model training module includes an image preprocessing unit and a model training unit; The image preprocessing unit is used to label and preprocess the image data for target detection, and construct a training set and a test set for garbage overflow detection; The model training unit is used to build a YOLO garbage overflow detection model, and train the YOLO garbage overflow detection model based on the training set to obtain a trained YOLO garbage overflow detection model.

4. The YOLO-based intelligent garbage overflow monitoring system from the perspective of a drone according to claim 3 is characterized in that: The YOLO garbage spill detection model includes a SimAM attention mechanism with slicing operations and several large object detection heads.

5. The YOLO-based intelligent garbage overflow monitoring system from the perspective of a drone according to claim 4 is characterized in that: The expression of the SimAM attention mechanism is: In the formula, ω i is the attention weight of the ith pixel, k is a normalization constant, N i is the set of neighboring pixels of the ith pixel, s(f i ,f j ) is the similarity between the i-th pixel and the j-th pixel.

6. The YOLO-based intelligent garbage overflow monitoring system from the perspective of a drone according to claim 4 is characterized in that: The loss calculation formula of the YOLO garbage overflow detection model is: L total =L bce +L cls +L dfl in, L bce =-ylog(p)-(1-y)log(1-p) L dfl =-(i+1-y)log(y i )-(yi)log(y i+1 ) Where, L total is the total loss value of the model, L bce is the standard binary cross entropy loss, L cls To calculate the intersection-over-union ratio between the predicted box and the target real box, L dfl is the distribution focus loss, y is the real target category, p is the predicted target category, c is the diagonal length of the minimum enclosing rectangular box, ρ is the Euclidean distance between the two, b is the coordinate of the center point of the predicted box, and b gt is the coordinate of the center point of the target real box, αν is the aspect ratio hyperparameter of the predicted box and the target real box, IoU is the intersection over union ratio of the predicted box and the target real box, y i is the predicted sequence value.

7. The YOLO-based intelligent garbage overflow monitoring system from the perspective of a drone according to claim 1, characterized in that: The user interface module includes a decision support tool, which includes: a route planning unit, a system decision unit, a model configuration unit and a detection result unit.

8. A computer terminal device, characterized in that: include: one or more processors; A memory, coupled to the processor, for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the YOLO-based intelligent garbage overflow monitoring system from the perspective of the drone as described in any one of claims 1-7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, the YOLO-based intelligent garbage overflow monitoring system from the perspective of the drone as described in any one of claims 1 to 7 is implemented.

Images