Mobile phone off-post automatic monitoring method and system based on YOLOv8

Abstract

The invention discloses a mobile phone off-post automatic monitoring method and system based on YOLOv8, and the method comprises the steps: obtaining a video stream data set based on a target supervision place, and carrying out the video stream service process monitoring and frame rate capturing processing, and obtaining a target image data set; based on a YOLOv8 target detection and identification network, introducing a double-branch convolutional network structure, and constructing an improved YOLOv8 target detection and identification network model; performing target identification and detection processing on the target image data set based on the improved YOLOv8 target detection and identification network model to obtain a target identification and detection result; and performing mobile phone target coordinate sorting detection on the target identification detection result to obtain a target monitoring result with the mobile phone. According to the embodiment of the invention, the detail feature extraction of the image can be improved, and the recognition and detection precision of using a mobile phone is improved. The method can be widely applied to the technical field of image real-time monitoring.

Description

Technical Field

[0001] The present application relates to the technical field of real-time image monitoring, and in particular to a method and system for automatically monitoring mobile phone off-duty based on YOLOv8. Background Art

[0002] For existing important supervision sites, personnel are required to work in real time, and it is prohibited to bring or play with mobile phones on the job. The mobile phones of on-the-job personnel are strictly managed. The traditional manual monitoring of whether mobile phones are used on the job is obviously inefficient and cannot meet the needs of supervision methods in real scenarios. In addition, the existing deep convolutional neural networks have been greatly developed. In the field of image recognition, many scholars have emerged to design different network architectures to solve image processing problems in different fields. Their goal is to achieve faster and better automated image recognition, thereby replacing manual work and improving diagnostic efficiency and accuracy. However, related technical identification uses Faster RCNN and other slow reasoning speeds, low recognition accuracy, and requires manual network disconnection or flow recovery.

[0003] In summary, the technical problems existing in the relevant technologies need to be improved. Summary of the invention

[0004] The main purpose of the embodiments of the present application is to propose a method and system for automatically monitoring mobile phone off-duty based on YOLOv8, which can improve the extraction of detailed features of images and thus improve the accuracy of recognition and detection of whether a mobile phone is in use.

[0005] To achieve the above purpose, one aspect of an embodiment of the present application proposes a method for automatically monitoring mobile phone off-duty based on YOLOv8, the method comprising:

[0006] Based on the target supervision location, the video stream data set is obtained and the video stream service process is monitored and the frame rate is captured to obtain the target image data set;

[0007] Based on the YOLOv8 target detection and recognition network, a dual-branch convolutional network structure is introduced to build an improved YOLOv8 target detection and recognition network model;

[0008] Performing target recognition detection processing on the target image dataset based on the improved YOLOv8 target detection and recognition network model to obtain a target recognition detection result;

[0009] The target recognition detection results are subjected to mobile phone target coordinate sorting detection to obtain target monitoring results having mobile phones.

[0010] In some embodiments, the step of acquiring a video stream data set based on a target supervision location and performing video stream service process monitoring and frame rate capture processing to obtain a target image data set includes:

[0011] Collecting data from the target supervision site through a camera to obtain the video stream data set;

[0012] Build a preset video streaming service process detection mechanism;

[0013] Obtaining the stream address of the camera, performing frame rate capture processing on the video stream data set in combination with the preset video stream service process detection mechanism, and obtaining a target image;

[0014] The target image is annotated and format converted to obtain the target image data set.

[0015] In some embodiments, the preset video stream service process detection mechanism includes:

[0016] Set the start mark command and the stop mark command;

[0017] Initialize the local service network and detect whether the network is abnormal;

[0018] If the network is normal, the video stream service network is initialized and the existence of the video service process is detected. If the video service process exists, the video service process is obtained and the stop identification command is triggered. If the video service process does not exist, the start identification command is re-triggered.

[0019] If the network is abnormal, it is detected whether the video service process exists. If it exists, the video service process is obtained and the stop identification command is triggered. If it does not exist, the start identification command is re-triggered.

[0020] In some embodiments, the improved YOLOv8 target detection and recognition network model includes a backbone network module, a neck module and a prediction module, the output end of the backbone network module is connected to the input end of the neck module, and the output end of the neck module is connected to the input end of the prediction module, wherein:

[0021] The backbone network module includes a convolution module, a dual-branch convolution network structure, a splicing module, a C2F module, an optional selection module and a spatial pyramid pooling module, the convolution module includes a first convolution layer, a second convolution layer, a third convolution layer, a fourth convolution layer and a fifth convolution layer, and the C2F module includes a first C2F layer, a second C2F layer, a third C2F layer and a fourth C2F layer;

[0022] The neck module includes an upsampling module, a splicing module and a convolution module;

[0023] The prediction module includes three detection head modules.

[0024] In some embodiments, the dual-branch convolutional network structure includes a first branch convolutional structure and a second branch convolutional structure, the first branch convolutional structure includes a sixth convolutional layer and a seventh convolutional layer, and the second branch convolutional structure includes an eighth convolutional layer, a ninth convolutional layer, a concatenation layer and a tenth convolutional layer, wherein:

[0025] Based on the first branch convolution structure, the output end of the first convolution layer is connected to the input end of the sixth convolution layer and the input end of the seventh convolution layer respectively, and the output end of the sixth convolution layer and the output end of the seventh convolution layer are connected to the input end of the splicing module respectively;

[0026] Based on the second branch convolution structure, the output end of the first C2F layer is respectively connected to the input end of the eighth convolution layer and the input end of the ninth convolution layer, the output end of the eighth convolution layer and the output end of the ninth convolution layer are respectively connected to the first input end of the splicing layer, the output layer of the second C2F layer is connected to the input end of the optional selection module, the output end of the optional selection module is connected to the second input end of the splicing layer, and the output end of the splicing layer is connected to the input end of the tenth convolution layer.

[0027] In some embodiments, the target image dataset is subjected to target recognition detection processing based on the improved YOLOv8 target detection and recognition network model to obtain a target recognition detection result, including:

[0028] Inputting the target image data set into the improved YOLOv8 target detection and recognition network model;

[0029] Based on the backbone network module of the improved YOLOv8 target detection and recognition network model, feature extraction processing is performed on the target image data set to obtain target image features;

[0030] Based on the neck module of the improved YOLOv8 target detection and recognition network model, feature fusion processing is performed on the target image features to obtain target image fusion features;

[0031] Based on the prediction module of the improved YOLOv8 target detection and recognition network model, feature detection processing is performed on the target image fusion features to obtain the target recognition detection result.

[0032] In some embodiments, the backbone network module based on the improved YOLOv8 target detection and recognition network model performs feature extraction processing on the target image data set to obtain target image features, including:

[0033] Inputting the target image data set into the backbone network module;

[0034] Based on the convolution module of the backbone network module, the target image data set is subjected to convolution processing to obtain a convolved target image;

[0035] Based on the dual-branch convolutional network structure of the backbone network module, down-sampling and convolution splicing processing are performed on the convolved target image to obtain a preliminary spliced ​​target image;

[0036] Based on the stitching module of the backbone network module, the preliminary stitching target image is stitched to obtain a stitched target image;

[0037] Based on the C2F module of the backbone network module, multi-scale feature extraction processing is performed on the spliced ​​target image to obtain multi-scale target image features;

[0038] Based on the optional selection module of the backbone network module, selectively operate the multi-scale target image features to obtain pre-processed target image features;

[0039] Based on the spatial pyramid pooling module of the backbone network module, spatial channel feature extraction processing is performed on the preprocessed target image features to obtain the target image features.

[0040] In some embodiments, the optional selection module based on the backbone network module performs selective operation processing on the multi-scale target image features to obtain pre-processed target image features, including:

[0041] Inputting the multi-scale target image features into the optional selection module;

[0042] If the feature resolution of the output result of the eighth convolutional layer, the feature resolution of the output result of the ninth convolutional layer and the feature resolution of the output result of the second C2F layer are consistent, the optional selection module selects a no operation;

[0043] If the feature resolution of the output result of the eighth convolutional layer, the feature resolution of the output result of the ninth convolutional layer and the feature resolution of the output result of the second C2F layer are inconsistent, the optional selection module selects to downsample the multi-scale target image features.

[0044] In some embodiments, performing mobile phone target coordinate sorting detection on the target recognition detection results to obtain target monitoring results with mobile phones includes:

[0045] Performing image segmentation processing according to the size of the target recognition detection result to obtain a segmented target recognition detection image, wherein the segmented target recognition detection image has a plurality of grids, and determining the center coordinates of the plurality of grids;

[0046] Screening the divided target recognition detection images to obtain target recognition detection images with mobile phones;

[0047] The center coordinates of the grid are subjected to coordinate sorting and matching processing with the target recognition detection image with the mobile phone to obtain the target monitoring result with the mobile phone.

[0048] To achieve the above purpose, another aspect of the embodiment of the present application proposes a mobile phone off-duty automatic monitoring system based on YOLOv8, the system comprising:

[0049] The first module is used to obtain a video stream data set based on the target supervision site and perform video stream service process monitoring and frame rate capture processing to obtain a target image data set;

[0050] The second module is used to introduce a dual-branch convolutional network structure based on the YOLOv8 target detection and recognition network to build an improved YOLOv8 target detection and recognition network model;

[0051] The third module is used to perform target recognition detection processing on the target image data set based on the improved YOLOv8 target detection and recognition network model to obtain a target recognition detection result;

[0052] The fourth module is used to perform mobile phone target coordinate sorting detection on the target recognition detection results to obtain target monitoring results with mobile phones.

[0053] The embodiments of the present application include at least the following beneficial effects: The present application provides a method and system for automatically monitoring off-duty mobile phones based on YOLOv8. The solution obtains a video stream data set of a target supervision site and performs video stream service process monitoring and frame rate capture processing to obtain a target image data set, integrates a mobile phone supervision detection algorithm into the network process, and then introduces a dual-branch convolutional network structure to construct an improved YOLOv8 target detection and recognition network model, performs target recognition detection processing on the target image data set, and the introduction of a dual-branch convolutional network structure can extract finer feature details of the image, which is more suitable for the case where the size is smaller than that of mobile phones, and can retain more information flows of target features to a certain extent, thereby enhancing the robustness of target recognition, and finally performs mobile phone target coordinate sorting detection on the target recognition detection results, thereby improving the recognition accuracy of the image. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1It is a flowchart of a method for automatically monitoring mobile phone off-duty based on YOLOv8 provided in an embodiment of the present application;

[0055] Figure 2 It is a structural schematic diagram of a mobile phone off-duty automatic monitoring system based on YOLOv8 provided in an embodiment of the present application;

[0056] Figure 3 It is a recognition schematic diagram of the existing YOLOv8 network provided in an embodiment of the present application;

[0057] Figure 4 : is a recognition schematic diagram of the improved YOLOv8 network provided in an embodiment of the present application;

[0058] Figure 5 is a schematic diagram of a preset video stream service process detection mechanism constructed in an embodiment of the present application;

[0059] Figure 6 It is a structural diagram of the improved YOLOv8 target detection and recognition network model provided in the embodiment of the present application.

[0060] Figure numerals: 1. first convolutional layer; 2. sixth convolutional layer; 3. splicing module; 4. second convolutional layer; 5. first C2F layer; 6. third convolutional layer; 7. second C2F layer; 8. fourth convolutional layer; 9. third C2F layer; 10. fifth convolutional layer; 11. fourth C2F layer; 12. spatial pyramid pooling module; 13. seventh convolutional layer; 14. eighth convolutional layer; 15. ninth convolutional layer; 16. optional selection module; 17. splicing layer; 18. tenth convolutional layer. DETAILED DESCRIPTION

[0061] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application is further described in detail below in conjunction with the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the embodiments of the present application. They are only examples of systems and methods consistent with some aspects of the embodiments of the present application as detailed in the attached claims.

[0062] It is understood that the terms "first", "second", etc. used in this application can be used to describe various concepts in this article, but unless otherwise specified, these concepts are not limited by these terms. These terms are only used to distinguish one concept from another concept. For example, without departing from the scope of the embodiment of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the words "if" and "if" as used herein can be interpreted as "at the time of" or "when" or "in response to determination".

[0063] The terms "at least one", "multiple", "each", "any", etc. used in this application, at least one includes one, two or more, multiple includes two or more, each refers to each of the corresponding multiple, and any refers to any one of the multiple.

[0064] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0065] Reference Figure 1 , Figure 1 A flowchart of a method for automatically monitoring mobile phone off-duty based on YOLOv8 provided by an embodiment of the present invention, referring to Figure 1 , the method comprises the following steps:

[0066] S100, obtaining a video stream data set based on a target supervision location and performing video stream service process monitoring and frame rate capture processing to obtain a target image data set;

[0067] It should be noted that, in some embodiments, step S100 may include: S110, collecting and processing data on the target supervision place through a camera to obtain a video stream data set; S120, building a preset video stream service process detection mechanism; S130, obtaining the stream address of the camera, and performing frame rate capture processing on the video stream data set in combination with the preset video stream service process detection mechanism to obtain a target image; S140, annotating and format converting the target image to obtain a target image data set.

[0068] It should be noted that, Figure 5As shown, the preset video stream service process detection mechanism includes setting a start identification command and a stop identification command; initializing the local service network and detecting whether the network is abnormal; if the network is normal, initializing the video stream service network and detecting whether the video service process exists, if it exists, obtaining the video service process and triggering the stop identification command, if not, re-triggering the start identification command; if the network is abnormal, detecting whether the video service process exists, if it exists, obtaining the video service process and triggering the stop identification command, if not, re-triggering the start identification command.

[0069] In some specific embodiments, data is collected from cameras installed in real scenes, and the LabelImg labeling tool is used to label the data, and the labeling categories are classified into the presence of mobile phones and the absence of mobile phones. The image format is converted to convert the HWC format image into the CHW format image.

[0070] S200, based on the YOLOv8 target detection and recognition network, introduces a dual-branch convolutional network structure to build an improved YOLOv8 target detection and recognition network model;

[0071] It should be noted that in some embodiments, Figure 6 As shown, the improved YOLOv8 target detection and recognition network model includes a backbone network module, a neck module and a prediction module. The output end of the backbone network module is connected to the input end of the neck module, and the output end of the neck module is connected to the input end of the prediction module. The backbone network module includes a convolution module, a dual-branch convolution network structure, a splicing module 3, a C2F module, an optional selection module 16 and a spatial pyramid pooling module 12. The convolution module includes a first convolution layer 1, a second convolution layer 4, a third convolution layer 6, a fourth convolution layer 8 and a fifth convolution layer 10. The C2F module includes a first C2F layer 5, a second C2F layer 7, a third C2F layer 9 and a fourth C2F layer 11. The neck module includes an upsampling module, a splicing module and a convolution module. The prediction module includes three detection head modules.

[0072] Furthermore, it should be noted that the dual-branch convolutional network structure includes a first branch convolutional structure and a second branch convolutional structure, the first branch convolutional structure includes a sixth convolutional layer 2 and a seventh convolutional layer 13, and the second branch convolutional structure includes an eighth convolutional layer 14, a ninth convolutional layer 15, a splicing layer 17 and a tenth convolutional layer 18, wherein, based on the first branch convolutional structure, the output end of the first convolutional layer is respectively connected to the input end of the sixth convolutional layer and the input end of the seventh convolutional layer, and the output end of the sixth convolutional layer and the output end of the seventh convolutional layer are respectively connected to the input end of the splicing module; based on the second branch convolutional structure, the output end of the first C2F layer is respectively connected to the input end of the eighth convolutional layer and the input end of the ninth convolutional layer, the output end of the eighth convolutional layer and the output end of the ninth convolutional layer are respectively connected to the first input end of the splicing layer, the output layer of the second C2F layer is connected to the input end of the optional selection module, the output end of the optional selection module is connected to the second input end of the splicing layer, and the output end of the splicing layer is connected to the input end of the tenth convolutional layer.

[0073] In some specific embodiments, the original YOLOv8 resizes the input image to a certain input size, then downsamples it through a convolution layer, and continues to downsample after passing through a convolution layer. Then it is processed through the c2f layer. For mobile phones that are smaller in size, this method is to more efficiently save the extracted feature information of the target. The feature map output by the first layer of convolution is downsampled by two second layers of convolution and then spliced, and finally a convolution layer is used to restore the feature map that is the same as the original input feature map of the c2f layer. This method can retain more information flow of target features to a certain extent. Enhance the robustness of target recognition.

[0074] S300, performing target recognition detection processing on the target image data set based on the improved YOLOv8 target detection and recognition network model to obtain a target recognition detection result;

[0075] It should be noted that, in some embodiments, step S300 may include: S310, inputting the target image data set into the improved YOLOv8 target detection and recognition network model; S320, based on the backbone network module of the improved YOLOv8 target detection and recognition network model, performing feature extraction processing on the target image data set to obtain target image features; S330, based on the neck module of the improved YOLOv8 target detection and recognition network model, performing feature fusion processing on the target image features to obtain target image fusion features; S340, based on the prediction module of the improved YOLOv8 target detection and recognition network model, performing feature detection processing on the target image fusion features to obtain target recognition detection results.

[0076] Among them, in some specific implementation steps S320, it also includes: S321, inputting the target image data set into the backbone network module; S322, based on the convolution module of the backbone network module, convolution processing is performed on the target image data set to obtain the convolved target image; S323, based on the dual-branch convolution network structure of the backbone network module, downsampling and convolution splicing processing is performed on the convolved target image to obtain a preliminary spliced ​​target image; S324, based on the splicing module of the backbone network module, splicing processing is performed on the preliminary spliced ​​target image to obtain the spliced ​​target image; S325, based on the C2F module of the backbone network module, multi-scale feature extraction processing is performed on the spliced ​​target image to obtain multi-scale target image features; S326, based on the optional selection module of the backbone network module, selective operation processing is performed on the multi-scale target image features to obtain the pre-processed target image features; S327, based on the spatial pyramid pooling module of the backbone network module, spatial channel feature extraction processing is performed on the pre-processed target image features to obtain the target image features.

[0077] In some specific embodiments, based on the optional selection module of the backbone network module, a selective operation is performed on the multi-scale target image features to obtain the preprocessed target image features, including: inputting the multi-scale target image features into the optional selection module; if the feature resolution of the output result of the eighth convolutional layer and the feature resolution of the output result of the ninth convolutional layer are consistent with the feature resolution of the output result of the second C2F layer, the optional selection module selects a no operation; if the feature resolution of the output result of the eighth convolutional layer and the feature resolution of the output result of the ninth convolutional layer are inconsistent with the feature resolution of the output result of the second C2F layer, the optional selection module selects to perform a downsampling operation on the multi-scale target image features.

[0078] In addition, in this embodiment, after the network diagram of the improved part performs convolution processing on the input image, two convolution operations are performed, and then splicing is performed, and then the number of channels and the resolution size are kept consistent with the previous ones through convolution. For example, the convolution operation of the blue part maintains the size of the original image resolution, and the convolution of the orange part restores the number of channels of the original unimproved model while reducing the resolution to be consistent with the previous one. If the convolution output feature map of the blue part is to reduce the resolution to be consistent with the previous resolution. The convolution size of the orange part is set to only change its number of channels to be consistent with the previous one. Similarly, the two convolutions of the purple part are to maintain the original feature map resolution, and the black part can be upsampled. Then splicing. The gray convolution performs downsampling and changes the number of channels. If the purple part reduces the resolution of the original image and keeps the resolution consistent with the spliced ​​C2F, the black part can not perform any operation. After splicing, the gray convolution restores the original number of channels. Then perform subsequent splicing operations.

[0079] S400, performing mobile phone target coordinate sorting detection on the target recognition detection result to obtain a target monitoring result having a mobile phone;

[0080] It should be noted that, in some embodiments, step S400 may include: S410, performing image division processing according to the size of the target recognition detection result to obtain a divided target recognition detection image, wherein the divided target recognition detection image has a number of grids, and determining the center coordinates of the several grids; S420, screening the divided target recognition detection image to obtain a target recognition detection image with a mobile phone; S430, performing coordinate sorting and matching processing on the center coordinates of the grid and the target recognition detection image with a mobile phone to obtain a target monitoring result with a mobile phone.

[0081] In some specific embodiments:

[0082] 1) First define the grid area;

[0083] Set the size and position of the grid. Assuming that a 3*4 shelf is used to place mobile phones, there are 12 grids in total. The model can predict 12 boxes, one with a mobile phone and one without a mobile phone, and determine the coordinates of the upper left corner or center point of each grid as a reference for subsequent sorting.

[0084] 2) Obtain recognition results;

[0085] Perform YOLOv8 target recognition, run the YOLOv8 model, obtain all detected locations and their category information, and extract the location coordinates of the mobile phone (x1, y1, x2, y2).

[0086] 3) Extract mobile phone coordinates;

[0087] Filter the mobile phone detection results, filter out the detection frame with the category of "mobile phone" from the detection results, record its coordinates, and obtain the coordinate value of the mobile phone detection frame. Then filter out the detection frame with the category of "non-mobile phone" from the detection results, record its coordinates, and obtain the coordinate value of the non-mobile phone detection frame.

[0088] 4) Sort by superscript;

[0089] First, sort by y coordinates. Here, we take a 3*4 grid as an example. Each row has four grids, so take out every four and then sort by x axis. Finally, save the data format as a JSON file. Each grid is set to 1 if it exists on a mobile phone, and 0 if it does not exist, so that the platform can call this JSON file for display.

[0090] To sum up, the original YOLOv8 flowchart is as follows Figure 3 As shown, the present invention is implemented as Figure 4 As shown in the figure, the detection and supervision algorithm of mobile phones is integrated into the network process, where network supervision includes local network supervision and flow address supervision. Based on the YOLOv8 target detection and recognition network, a dual-branch convolutional network structure is introduced to build an improved YOLOv8 target detection and recognition network model. It also includes a mobile phone recognition post-processing sorting algorithm, which adds real-name statistical supervision to the platform for the mobile phones of a certain category of people, generates reports, and corresponding warning functions.

[0091] See also Figure 2 The embodiment of the present application also provides a mobile phone off-duty automatic monitoring system based on YOLOv8, which can implement the above-mentioned mobile phone off-duty automatic monitoring method based on YOLOv8, and the system includes:

[0092] The first module 201 is used to obtain a video stream data set based on a target supervision location and perform video stream service process monitoring and frame rate capture processing to obtain a target image data set;

[0093] The second module 202 is used to introduce a dual-branch convolutional network structure based on the YOLOv8 target detection and recognition network to build an improved YOLOv8 target detection and recognition network model;

[0094] The third module 203 is used to perform target recognition detection processing on the target image data set based on the improved YOLOv8 target detection and recognition network model to obtain a target recognition detection result;

[0095] The fourth module 204 is used to perform mobile phone target coordinate sorting detection on the target recognition detection results to obtain target monitoring results with mobile phones.

[0096] It can be understood that the contents of the above method embodiments are all applicable to the present system embodiments, the functions specifically implemented by the present system embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0097] The preferred embodiments of the present application are described above with reference to the accompanying drawings, but the scope of the rights of the present application is not limited thereto. Any modification, equivalent substitution and improvement made by a person skilled in the art without departing from the scope and essence of the present application should be within the scope of the rights of the present application.

Claims

1. A method for automatically monitoring mobile phone off-duty based on YOLOv8, characterized in that: The method comprises the following steps: Based on the target supervision location, the video stream data set is obtained and the video stream service process is monitored and the frame rate is captured to obtain the target image data set; Based on the YOLOv8 target detection and recognition network, a dual-branch convolutional network structure is introduced to build an improved YOLOv8 target detection and recognition network model; Performing target recognition detection processing on the target image dataset based on the improved YOLOv8 target detection and recognition network model to obtain a target recognition detection result; The target recognition detection results are subjected to mobile phone target coordinate sorting detection to obtain target monitoring results having mobile phones.

2. The method according to claim 1, characterized in that The method of obtaining a video stream data set based on the target supervision site and performing video stream service process monitoring and frame rate capture processing to obtain a target image data set includes: Collecting data from the target supervision site through a camera to obtain the video stream data set; Build a preset video streaming service process detection mechanism; Obtaining the stream address of the camera, performing frame rate capture processing on the video stream data set in combination with the preset video stream service process detection mechanism, and obtaining a target image; The target image is annotated and format converted to obtain the target image data set.

3. The method according to claim 2, characterized in that The preset video stream service process detection mechanism includes: Set the start mark command and the stop mark command; Initialize the local service network and detect whether the network is abnormal; If the network is normal, the video stream service network is initialized and the existence of the video service process is detected. If the video service process exists, the video service process is obtained and the stop identification command is triggered. If the video service process does not exist, the start identification command is re-triggered. If the network is abnormal, it is detected whether the video service process exists. If it exists, the video service process is obtained and the stop identification command is triggered. If it does not exist, the start identification command is re-triggered.

4. The method according to claim 1, characterized in that: The improved YOLOv8 target detection and recognition network model includes a backbone network module, a neck module and a prediction module, wherein the output end of the backbone network module is connected to the input end of the neck module, and the output end of the neck module is connected to the input end of the prediction module, wherein: The backbone network module includes a convolution module, a dual-branch convolution network structure, a splicing module, a C2F module, an optional selection module and a spatial pyramid pooling module, the convolution module includes a first convolution layer, a second convolution layer, a third convolution layer, a fourth convolution layer and a fifth convolution layer, and the C2F module includes a first C2F layer, a second C2F layer, a third C2F layer and a fourth C2F layer; The neck module includes an upsampling module, a splicing module and a convolution module; The prediction module includes three detection head modules.

5. The method according to claim 4, characterized in that The dual-branch convolutional network structure includes a first branch convolutional structure and a second branch convolutional structure, the first branch convolutional structure includes a sixth convolutional layer and a seventh convolutional layer, and the second branch convolutional structure includes an eighth convolutional layer, a ninth convolutional layer, a concatenation layer and a tenth convolutional layer, wherein: Based on the first branch convolution structure, the output end of the first convolution layer is connected to the input end of the sixth convolution layer and the input end of the seventh convolution layer respectively, and the output end of the sixth convolution layer and the output end of the seventh convolution layer are connected to the input end of the splicing module respectively; Based on the second branch convolution structure, the output end of the first C2F layer is respectively connected to the input end of the eighth convolution layer and the input end of the ninth convolution layer, the output end of the eighth convolution layer and the output end of the ninth convolution layer are respectively connected to the first input end of the splicing layer, the output layer of the second C2F layer is connected to the input end of the optional selection module, the output end of the optional selection module is connected to the second input end of the splicing layer, and the output end of the splicing layer is connected to the input end of the tenth convolution layer.

6. The method according to claim 1, characterized in that The target recognition detection processing is performed on the target image data set based on the improved YOLOv8 target detection and recognition network model to obtain a target recognition detection result, including: Inputting the target image data set into the improved YOLOv8 target detection and recognition network model; Based on the backbone network module of the improved YOLOv8 target detection and recognition network model, feature extraction processing is performed on the target image data set to obtain target image features; Based on the neck module of the improved YOLOv8 target detection and recognition network model, feature fusion processing is performed on the target image features to obtain target image fusion features; Based on the prediction module of the improved YOLOv8 target detection and recognition network model, feature detection processing is performed on the target image fusion features to obtain the target recognition detection result.

7. The method according to claim 6, characterized in that The backbone network module based on the improved YOLOv8 target detection and recognition network model performs feature extraction processing on the target image data set to obtain target image features, including: Inputting the target image data set into the backbone network module; Based on the convolution module of the backbone network module, the target image data set is subjected to convolution processing to obtain a convolved target image; Based on the dual-branch convolutional network structure of the backbone network module, down-sampling and convolution splicing processing are performed on the convolved target image to obtain a preliminary spliced ​​target image; Based on the stitching module of the backbone network module, the preliminary stitching target image is stitched to obtain a stitched target image; Based on the C2F module of the backbone network module, multi-scale feature extraction processing is performed on the spliced ​​target image to obtain multi-scale target image features; Based on the optional selection module of the backbone network module, selectively operate the multi-scale target image features to obtain pre-processed target image features; Based on the spatial pyramid pooling module of the backbone network module, spatial channel feature extraction processing is performed on the preprocessed target image features to obtain the target image features.

8. The method according to claim 7, characterized in that The optional selection module based on the backbone network module performs selective operation processing on the multi-scale target image features to obtain pre-processed target image features, including: Inputting the multi-scale target image features into the optional selection module; If the feature resolution of the output result of the eighth convolutional layer, the feature resolution of the output result of the ninth convolutional layer and the feature resolution of the output result of the second C2F layer are consistent, the optional selection module selects a no operation; If the feature resolution of the output result of the eighth convolutional layer, the feature resolution of the output result of the ninth convolutional layer and the feature resolution of the output result of the second C2F layer are inconsistent, the optional selection module selects to downsample the multi-scale target image features.

9. The method according to claim 1, characterized in that: The target recognition detection results are subjected to mobile phone target coordinate sorting detection to obtain target monitoring results having mobile phones, including: Performing image segmentation processing according to the size of the target recognition detection result to obtain a segmented target recognition detection image, wherein the segmented target recognition detection image has a plurality of grids, and determining the center coordinates of the plurality of grids; Screening the divided target recognition detection images to obtain target recognition detection images with mobile phones; The center coordinates of the grid are subjected to coordinate sorting and matching processing with the target recognition detection image with the mobile phone to obtain the target monitoring result with the mobile phone.

10. A mobile phone off-duty automatic monitoring system based on YOLOv8, characterized in that: The system comprises: The first module is used to obtain a video stream data set based on the target supervision site and perform video stream service process monitoring and frame rate capture processing to obtain a target image data set; The second module is used to introduce a dual-branch convolutional network structure based on the YOLOv8 target detection and recognition network to build an improved YOLOv8 target detection and recognition network model; The third module is used to perform target recognition detection processing on the target image data set based on the improved YOLOv8 target detection and recognition network model to obtain a target recognition detection result; The fourth module is used to perform mobile phone target coordinate sorting detection on the target recognition detection results to obtain target monitoring results with mobile phones.

Images