A method for monitoring and controlling the illegal use of mobile phones

Abstract

This invention relates to a domestically produced method for monitoring the unauthorized use of mobile phones, belonging to the field of information security protection technology. The system of this invention includes network cameras, a domestically produced intelligent server, access control, client terminals, a video wall, and a switch. The network cameras and access control are deployed in the monitoring location; the switch and the domestically produced intelligent server are deployed in the server room; the video wall is deployed in the monitoring room; and the client terminals are deployed in the office. The method of this invention includes: video acquisition, video analysis, display of analysis results, and access control linkage control. This invention uses domestically produced CPUs and domestically produced accelerator cards, making it more secure and eliminating the risk of sanctions or backdoors; it can automatically control access, promptly preventing unauthorized mobile phone users from escaping the scene, thus buying valuable time to prevent information leaks.

Description

Technical Field

[0001] This invention belongs to the field of information security protection technology, specifically relating to a domestically produced method for monitoring the unauthorized use of mobile phones. Background Technology

[0002] With the increasing power of mobile phones, which can transmit information via calls, voice messages, videos, and photos, preventing information leaks due to unauthorized mobile phone use has become a key concern for many organizations. In real-world scenarios, personnel monitoring multiple screens simultaneously face monotonous and tiring viewing, making it easy to miss violations. If personnel are not promptly detected and prevented from leaving the leak location, it becomes difficult to control the scope of the leak within a short period.

[0003] Existing methods for detecting unauthorized mobile phone use mainly rely on non-domestic accelerator cards and merely return analysis results, lacking both the processing of unauthorized mobile phone use and statistical analysis of such behavior. Summary of the Invention

[0004] (a) Technical problems to be solved

[0005] The technical problem this invention aims to solve is how to provide a domestically produced method for monitoring the unauthorized use of mobile phones, in order to address the issues of easy omissions and inability to promptly restrict the actions of violators in manual monitoring, as well as the usage restrictions caused by non-domestic systems.

[0006] (II) Technical Solution

[0007] To address the aforementioned technical problems, this invention proposes a method for monitoring the unauthorized use of mobile phones using domestically produced equipment. This method is applied to a system for monitoring the unauthorized use of mobile phones using domestically produced equipment. The system includes network cameras, a domestically produced intelligent server, access control systems, client terminals, a video wall, and a switch. The network cameras and access control systems are deployed in the monitoring location; the switch and the domestically produced intelligent server are deployed in the computer room; the video wall is deployed in the monitoring room; and the client terminals are deployed in the office. The method includes the following steps:

[0008] S1. Video Acquisition: The video stream acquired by the network camera is transmitted to the domestic intelligent server via a switch;

[0009] S2. Video Analysis: The domestically produced intelligent server parses and analyzes the video stream. The domestically produced intelligent server integrates a domestically produced acceleration card and uses the YOLOv7 model to analyze the situation of illegal use of mobile phones.

[0010] S3. Analysis Results Display: The analysis results are transmitted to the client terminal and video wall and displayed on the monitoring video in the form of a detection frame. It also supports the display of statistical results of illegal mobile phone use.

[0011] S4. Access Control Linkage: After the analysis results are filtered, the access control is closed. Once the access control is closed due to violation, it cannot be opened from the inside by password, card or facial recognition. It can only be reopened from the outside or by the client terminal with administrator privileges.

[0012] Furthermore, step S1 specifically includes: deploying a network camera in the area to be monitored, connecting it to a switch via a network cable, maintaining the camera height between 2 and 2.5 meters and the tilt angle less than 45°, encoding the video stream collected by the network camera in H264 encoding mode, and transmitting the video stream to a domestic intelligent server via the RTSP protocol.

[0013] Further, step S2 specifically includes:

[0014] S21. Video stream parsing: Parse the RTSP video stream using OpenCV, obtain images by frame skipping, convert the color encoding format of the parsed images from BGR to RGB format, and scale the image size to 640*480.

[0015] S22. Establishment of a domestic intelligent analysis environment: The domestic intelligent server adopts the Cambricon MLU 270 accelerator card;

[0016] S23. Training the model for detecting illegal mobile phone use: The YOLOv7 model is trained using the COCO2017 dataset. The detection categories are people and mobile phones. The ratio of training set to test set is 8:2. 16 images are used as a batch. The SGD optimizer method is used and the training is carried out for 200 rounds.

[0017] S24. Image Analysis: Analyze the images after parsing the video stream using the trained YOLOv7 model. Set the confidence threshold to 0.5. If a person or mobile phone target is detected, the model outputs the categories and coordinates of all detected targets.

[0018] Furthermore, in step S21, the aspect ratio is maintained during image scaling until the scaled length or width is exactly equal to the target size, and the image pixels are less than or equal to the target image pixels. The blank portion of the target size is then filled with 0 for each of the RGB three channels.

[0019] Furthermore, step S22 also includes installing Python C, ATeN, and Torch_MLU modules, as well as Cambricon PyTorch, Cambricon Catch, and the binary code compiled from the Cambricon Vision source code in the Phytium 2000+ CPU and Galaxy Kylin V10 operating system environment.

[0020] Furthermore, in step S23, the hyperparameter values ​​are set as follows: network depth parameter is 0.75, network width parameter is 1, learning rate is 0.01, stochastic gradient descent momentum parameter is 0.99, and weight decay is 0.00025.

[0021] Furthermore, step S3 specifically includes the following steps:

[0022] S31. Video overlay of analysis results: OpenCV is used to draw the detection box, and the detection category and confidence level are overlaid on the screen above the detection box. The category of human is represented by a yellow box and text, and the category of mobile phone is represented by a red box and text. FFmpeg is used to transmit the image after the detection box is overlaid to the client terminal and TV screen for display via RTSP in H264 encoding.

[0023] S32. Analysis Results Statistics: If a mobile phone and a person are detected, calculate the Euclidean distance between the center of the mobile phone detection frame and the center of the nearest person's detection frame. If the distance is less than 3 times the width of the person's detection frame, it is considered that there is a violation of mobile phone use behavior, and one violation of mobile phone use is recorded.

[0024] S33. Analysis Result Warning: When the analysis results indicate that there are records of unauthorized mobile phone use, an alarm message, including the location and time, will be displayed in a pop-up window on the client terminal monitoring screen.

[0025] Furthermore, in step S32, the recorded information includes the time, the camera number of the video source, and the camera's deployment location. Statistics are compiled based on time and location, and the locations and time periods where unauthorized mobile phone use is likely to occur are analyzed, which can be viewed on the client terminal.

[0026] Furthermore, in step S32, alarms from the same camera within 5 seconds are not recorded repeatedly.

[0027] Furthermore, step S4 specifically includes:

[0028] S41. Analysis result filtering control: Utilizing the analysis results of each frame, if 10 consecutive frames within 15 frames are judged as unauthorized use of mobile phones, it is considered that there is a very high probability of unauthorized use of mobile phones, and it is necessary to close the access control.

[0029] S42. Access Control Closure: When the system determines that it is necessary to close the access control, the domestic intelligent server automatically sends a closure command to the access control, and the access control cannot be unlocked internally.

[0030] S43. Access Control Restriction Removal: Access is also prohibited to ordinary users outside the access control system. Only administrators can remove access control restrictions and open the access control system through the client terminal or from outside the access control system.

[0031] (III) Beneficial Effects

[0032] This invention proposes a domestically developed method for monitoring unauthorized mobile phone use. Compared with existing technologies, this invention has the following advantages: it utilizes domestically produced CPUs and accelerator cards, making it more secure and eliminating the risk of sanctions or backdoors; it can automatically control access control, promptly preventing unauthorized mobile phone users from escaping the scene, thus buying valuable time to prevent information leaks; and it employs the YOLOv7 model to detect violations, avoiding the problem of omissions that are easily missed by manual monitoring. Attached Figure Description

[0033] Figure 1 This is a schematic diagram of the system structure. Detailed Implementation

[0034] To make the objectives, contents, and advantages of the present invention clearer, the specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples.

[0035] The technical problem to be solved by this invention is to provide a domestically produced system for monitoring the unauthorized use of mobile phones, so as to solve the problems of easy omissions and inability to restrict the actions of unauthorized personnel in a timely manner by manual monitoring, and to solve the problem of usage restrictions caused by non-domestic systems.

[0036] This invention discloses a domestically produced system for monitoring the unauthorized use of mobile phones, comprising a network camera, a domestically produced intelligent server, an access control system, a client terminal, a video wall, and a switch. The network camera and access control system are deployed in the monitoring location; the switch and the domestically produced intelligent server are deployed in the server room; the video wall is deployed in the monitoring room; and the client terminal is deployed in the office.

[0037] The system workflow is as follows: (1) Video acquisition. The video stream acquired by the network camera is transmitted to the domestic intelligent server through the switch; (2) Video analysis. The domestic intelligent server performs parsing and analysis on the video stream. The domestic intelligent server integrates a domestic acceleration card and uses the YOLOv7 model to analyze the situation of illegal use of mobile phones; (3) Analysis result display. The analysis results are transmitted to the client terminal and the video wall and displayed on the monitoring video in the form of a detection frame. It also supports the display of statistical results of illegal use of mobile phones; (4) Access control linkage. The analysis results are filtered and the access control is closed. After the access control is closed due to illegal use, it cannot be opened from the inside by password, card or face recognition. It can only be reopened from the outside or by the client terminal with the administrator's permission.

[0038] To achieve the above objectives, this invention provides a domestically produced system for monitoring the unauthorized use of mobile phones, comprising a network camera, a domestically produced intelligent server, an access control system, a client terminal, a video wall, and a switch. The network camera and access control system are deployed in the monitoring location; the switch and the domestically produced intelligent server are deployed in the server room; the video wall is deployed in the monitoring room; and the client terminal is deployed in the office.

[0039] The system's workflow is as follows:

[0040] S1, Video capture.

[0041] The network camera is deployed in the area to be monitored and connected to the switch via a network cable. The optimal deployment position is to keep the camera height between 2 and 2.5 meters and the tilt angle less than 45 degrees. The video stream captured by the network camera is encoded in H.264 and transmitted to the domestic intelligent server via the RTSP protocol.

[0042] S2, Video Analysis.

[0043] S21. Video Stream Parsing. Parse the RTSP video stream using OpenCV. Frame skipping can be used to acquire images, for example, taking one frame out of every three for analysis. Convert the parsed image color encoding format from BGR to RGB format, and scale the image size to 640*480. Maintain the aspect ratio during scaling until the scaled length or width is exactly equal to the target size, and the image pixel count is less than or equal to the target image pixel count. Fill the blank areas of the target size with 0 for each of the RGB channels.

[0044] S22. Establishment of a domestically developed intelligent analysis environment. The domestically developed intelligent server uses a Cambricon MLU 270 accelerator card. Python C, ATen, and Torch_MLU modules are installed on a Phytium 2000+ CPU and a Kylin V10 operating system environment. Cambricon PyTorch, Cambricon Catch, and binary code compiled from Cambricon Vision source code are also installed.

[0045] S23. Training the model for detecting unauthorized mobile phone use. A YOLOv7 model was trained using the COCO2017 dataset, detecting two categories: people and mobile phones. The training set to test set ratio was 8:2, with 16 images per batch. The SGD optimizer was used, and the training was performed for 200 epochs. The hyperparameters were set as follows: network depth 0.75, network width 1, learning rate 0.01, stochastic gradient descent momentum 0.99, and weight decay 0.00025.

[0046] S24. Image Analysis. The trained YOLOv7 model is used to analyze the images parsed from the video stream. The confidence threshold is set to 0.5. If a person or mobile phone target is detected, the model outputs the categories and coordinates of all detected targets.

[0047] S3. Analysis Results Display. The analysis results are transmitted to the client terminal and video wall and displayed on the surveillance video in the form of a detection frame. It also supports displaying statistical results of illegal mobile phone use.

[0048] S31. Video Overlay Analysis of Results. OpenCV is used to draw the detection bounding boxes. The detected category and confidence level are overlaid on top of the boxes. The category "human" is represented by a yellow box and text, while the category "mobile phone" is represented by a red box and text. FFmpeg is used to transmit the image with the overlaid detection boxes to the client terminal and a large TV screen via RTSP with H.264 encoding.

[0049] S32. Statistical Analysis of Results. If a mobile phone and a person are detected, calculate the Euclidean distance between the center of the mobile phone detection frame and the center of the nearest person's detection frame. If the distance is less than three times the width of the person's detection frame, it is considered a violation of mobile phone use, and this violation is recorded. Alarms from the same camera within 5 seconds are not recorded repeatedly. Recorded information includes time, camera ID of the video source, and camera deployment location. Statistics are compiled by time and location, and the locations and time periods where mobile phone violations are likely to occur are analyzed. This information can be viewed on the client terminal.

[0050] S33. Analysis Result Warning. When the analysis results indicate unauthorized mobile phone use, an alarm message, including location and time, will be displayed as a pop-up on the client terminal monitoring screen.

[0051] S4, Access Control Linkage Control.

[0052] S41. Analysis Result Filtering Control. To minimize the consequences of false alarms leading to erroneous access control closure, the analysis results of each frame are used. If 10 consecutive frames within 15 frames are identified as unauthorized mobile phone use, it is considered that there is a very high probability of unauthorized mobile phone use, and it is necessary to close the access control.

[0053] S42. Access Control Closure. When the system determines that it is necessary to close the access control, the domestic intelligent server automatically sends a closure command to the access control system, and the access control cannot be unlocked internally.

[0054] S43. Access Control Restriction Removed. Access is also prohibited to ordinary users outside the access control system. Only administrators can remove the access control restrictions and open the access control system via the client terminal or from outside the system.

[0055] Example 1:

[0056] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0057] (1) Video capture. The video stream captured by the network camera is transmitted to the domestic intelligent server via a switch:

[0058] The network camera is deployed in the area to be monitored and connected to the switch via a network cable. The optimal deployment position is to keep the camera height between 2 and 2.5 meters and the tilt angle less than 45 degrees. The video stream captured by the network camera is encoded in H.264 and transmitted to the domestic intelligent server via the RTSP protocol.

[0059] (2) Video Analysis. Domestic intelligent servers perform parsing and analysis on the video stream:

[0060] (21) Video Stream Parsing. The RTSP video stream is parsed using OpenCV. To improve efficiency for multi-channel video analysis, a frame skipping method is used, typically taking one frame out of every three for analysis. The parsed color encoding format is converted from BGR to RGB format, and the image size is scaled to 640*480. The aspect ratio is maintained during image scaling until the scaled length or width is exactly equal to the target size, while the image pixel count is less than or equal to the target image pixel count. The blank areas of the target size are then filled with 0s for each of the RGB channels.

[0061] (22) Establishment of a domestic intelligent analysis environment. A Cambricon MLU 270 accelerator card was used. Python C, ATen, and Torch_MLU modules were installed on a Phytium 2000+ CPU and a Kylin V10 operating system environment. Cambricon PyTorch, Cambricon Catch, and binary code compiled from Cambricon Vision source code were also installed.

[0062] (23) Training the model for detecting unauthorized mobile phone use. The YOLOv7 object detection model for detecting mobile phones was trained using the COCO2017 dataset. The detection categories are people and mobile phones. The ratio of training set to test set is 8:2. 16 images are used as a batch. The SGD optimizer method is used and the training is carried out for 200 epochs. The hyperparameters are set as follows: network depth is 0.75, network width is 1, learning rate is 0.01, stochastic gradient descent momentum is 0.99, and weight decay is 0.00025.

[0063] (24) Image Analysis. The model for detecting illegal mobile phone use is used to analyze the images after the video stream is parsed: the confidence threshold is set to 0.5. If a person or mobile phone target is detected, the model outputs the categories and coordinates of all detected targets.

[0064] (3) Analysis Results Display. The analysis results are transmitted to the client terminal and displayed on the video wall as a detection frame on the surveillance video. It also supports displaying statistical results of illegal mobile phone use.

[0065] (31) Video overlay analysis results. OpenCV was used to draw the detection bounding boxes, and the detected category and confidence level were overlaid on the image above the boxes. The category "human" was represented by a yellow box and text, while the category "mobile phone" was represented by a red box and text. FFmpeg was used to transmit the above images via RTSP in H.264 encoding to the client terminal and a large television screen for display.

[0066] (32) Statistical Analysis of Results. If a mobile phone and a person are detected, calculate the Euclidean distance between the center of the mobile phone detection frame and the center of the nearest person detection frame. If the distance is less than three times the width of the person's detection frame, it is considered an unauthorized use of a mobile phone, and this is recorded as one instance of unauthorized use. Alarms from the same camera within 5 seconds are not recorded repeatedly. Recorded information includes time, camera number of the video source, and camera deployment location. Statistics are compiled by time and location, and the locations and time periods where unauthorized use of mobile phones is likely to occur are analyzed. This information can be viewed on the client terminal.

[0067] (33) Analysis and early warning. When the analysis results indicate that there are records of unauthorized mobile phone use, an alarm message, including the location and time, will be displayed in a pop-up window on the client terminal monitoring screen.

[0068] (4) Access control linkage control

[0069] (41) Analysis result filtering control. In order to minimize the consequences of false alarms leading to incorrect closure of the access control, the results of each frame analysis are used. If 10 consecutive frames within 15 frames are judged as unauthorized use of mobile phones, it is considered that there is a very high probability of unauthorized use of mobile phones, and it is necessary to close the access control.

[0070] (42) Control access control to close. When the system determines that it is necessary to close the access control, the domestic intelligent server automatically sends a closing command to the access control, and the access control cannot be unlocked internally.

[0071] (43) Remove access restrictions. Access is also prohibited to ordinary users outside the access control system. Only administrators can remove access restrictions and open the access control system through the client terminal or from outside the access control system.

[0072] Example 2:

[0073] A domestically developed system for monitoring the unauthorized use of mobile phones, the system comprising:

[0074] (1) Network cameras and access control systems are deployed to monitor the location;

[0075] (2) Switches and domestically produced intelligent servers are deployed in the computer room;

[0076] (3) The video wall is deployed in the monitoring room;

[0077] (4) The client terminal is deployed in the office.

[0078] Furthermore, the workflow of the system includes:

[0079] (1) Video capture. The video stream captured by the network camera is transmitted to the domestic intelligent server via a switch;

[0080] (2) Video analysis. Domestically produced intelligent servers perform parsing, analysis, and processing of video streams;

[0081] (3) Presentation of analysis results;

[0082] (4) Access control linkage control.

[0083] Furthermore, in step (2), the video stream captured by the network camera is transmitted to the domestically produced intelligent server via a switch, specifically including:

[0084] The network camera is deployed in the area to be monitored and connected to the switch via a network cable. The optimal deployment position is to keep the camera height between 2 and 2.5 meters and the tilt angle less than 45 degrees. The video stream captured by the network camera is encoded in H.264 and transmitted to the domestic intelligent server via the RTSP protocol.

[0085] Further, in step (2), the Chinese-made intelligent server performs parsing and analysis on the video stream, specifically including:

[0086] (21) Video stream parsing;

[0087] (22) The establishment of a domestic intelligent analysis environment;

[0088] (23) Model training for detecting unauthorized mobile phone use;

[0089] (24) Image analysis. Images obtained after parsing the video stream are analyzed using a model that detects unauthorized mobile phone use.

[0090] Furthermore, the video stream parsing in step (21) specifically includes:

[0091] RTSP video streams are parsed using OpenCV, employing a frame-skipping approach, analyzing one frame out of every three. The parsed color encoding format is converted from BGR to RGB, and the image size is scaled to 640*480. The aspect ratio is maintained during scaling until the scaled length or width exactly matches the target size, while the image pixel count is less than or equal to the target image pixel count. Any blank areas within the target size are then filled with zeros for each of the RGB channels.

[0092] Furthermore, step (22) of establishing a domestic intelligent analysis environment specifically includes:

[0093] Using the Cambricon MLU 270 accelerator card, Python C, ATen, and Torch_MLU modules were installed on a Phytium 2000+ CPU and a Kylin V10 operating system environment. Cambricon PyTorch, Cambricon Catch, and the binary code compiled from the CambriconVision source code were also installed.

[0094] Furthermore, step (23) of the model training for detecting unauthorized mobile phone use specifically includes:

[0095] The target detection model chosen for mobile phones was YOLOv7, trained using the COCO2017 dataset. Two categories were selected: people and mobile phones. The training set to test set ratio was 8:2, with 16 images per batch. The SGD optimizer was used, and the training run was 200 epochs. The hyperparameters were set as follows: network depth 0.75, network width 1, learning rate 0.01, stochastic gradient descent momentum 0.99, and weight decay 0.00025.

[0096] Furthermore, the analysis results of step (3) are presented, specifically including:

[0097] (31) Video overlay of analysis results;

[0098] (32) Statistical analysis of results;

[0099] (33) Analysis and early warning.

[0100] Furthermore, the statistical analysis of the results in step (32) specifically includes:

[0101] If a phone and a person are detected, the Euclidean distance between the center of the phone detection frame and the center of the nearest person detection frame is calculated. If the distance is less than three times the width of the person's detection frame, it is considered unauthorized phone use, and this violation is recorded. Alarms from the same camera within 5 seconds are not recorded repeatedly. Recorded information includes time, camera ID of the video source, and camera deployment location. Statistics are compiled based on time and location, and the locations and time periods where unauthorized phone use is likely to occur are analyzed. This information can be viewed on the client terminal.

[0102] Furthermore, step (4) access control linkage control specifically includes:

[0103] To minimize the consequences of false alarms leading to erroneous access control closures, the analysis of each frame is used. If 10 out of 15 frames are consecutively identified as unauthorized mobile phone use, it is considered that there is a very high probability of unauthorized mobile phone use, and it is necessary to close the access control system.

[0104] Furthermore, step (4) access control linkage control specifically includes:

[0105] The access control system is automatically closed by a domestically produced intelligent server, and cannot be unlocked internally. Access is also restricted to users with ordinary external permissions; only administrators can unlock the access control system via a client terminal or externally.

[0106] Compared with existing technologies, this invention has the following advantages: it uses domestically produced CPUs and accelerator cards, making it more secure and eliminating the risk of sanctions or backdoors; it can automatically control access control, promptly preventing unauthorized mobile phone users from escaping the scene, thus buying valuable time to control information leaks.

[0107] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method of monitoring for homegrown violations of the use of cell phones, characterized by, This method is applied to a domestically produced system for monitoring the unauthorized use of mobile phones. The system includes network cameras, domestically produced smart servers, access control systems, client terminals, a video wall, and switches. The network cameras and access control systems are deployed in the monitoring area; the switches and domestically produced smart servers are deployed in the server room; the video wall is deployed in the monitoring room; and the client terminals are deployed in the office. The method includes the following steps: S1. Video Acquisition: The video stream acquired by the network camera is transmitted to the domestic intelligent server via a switch; S2. Video Analysis: The domestically produced intelligent server parses and analyzes the video stream. The domestically produced intelligent server integrates a domestically produced acceleration card and uses the YOLOv7 model to analyze the situation of illegal use of mobile phones. S3. Analysis Results Display: The analysis results are transmitted to the client terminal and video wall and displayed on the monitoring video in the form of a detection frame. It also supports the display of statistical results of illegal mobile phone use. S4. Access Control Linkage: After the analysis results are filtered, the access control is closed. Once the access control is closed due to violation, it cannot be opened from the inside by password, card or facial recognition. It can only be reopened from the outside or by the client terminal with administrator privileges. in, Step S2 specifically includes: S21. Video Stream Parsing: Parse the RTSP video stream using OpenCV, acquire images using frame skipping, convert the color encoding format of the parsed images from BGR to RGB format, and scale the image size to 640. 480; S22. Establishment of a domestic intelligent analysis environment: The domestic intelligent server adopts the Cambricon MLU 270 accelerator card; S23. Training the model for detecting illegal mobile phone use: The YOLOv7 model is trained using the COCO2017 dataset. The detection categories are people and mobile phones. The ratio of training set to test set is 8:

2. 16 images are used as a batch. The SGD optimizer method is used and the training is carried out for 200 rounds. S24. Image Analysis: Analyze the images after parsing the video stream using the trained YOLOv7 model. Set the confidence threshold to 0.

5. If a person or mobile phone target is detected, the model outputs the categories and coordinates of all detected targets. Step S3 specifically includes the following steps: S31. Video overlay of analysis results: OpenCV is used to draw the detection box, and the detection category and confidence level are overlaid on the screen above the detection box. The category of human is represented by a yellow box and text, and the category of mobile phone is represented by a red box and text. FFmpeg is used to transmit the image after the detection box is overlaid to the client terminal and TV screen for display via RTSP in H264 encoding. S32. Analysis Results Statistics: If a mobile phone and a person are detected, calculate the Euclidean distance between the center of the mobile phone detection frame and the center of the nearest person's detection frame. If the distance is less than 3 times the width of the person's detection frame, it is considered that there is a violation of mobile phone use behavior, and one violation of mobile phone use is recorded. S33. Analysis Result Warning: When the analysis results indicate that there are records of unauthorized mobile phone use, an alarm message, including the location and time, will be displayed in a pop-up window on the client terminal monitoring screen.

2. The method for monitoring the illegal use of mobile phones using domestically produced technology as described in claim 1, characterized in that, Step S1 specifically includes: deploying a network camera in the area to be monitored, connecting it to a switch via a network cable, maintaining the camera height between 2 and 2.5 meters and the tilt angle less than 45°, encoding the video stream collected by the network camera in H264 encoding mode, and transmitting the video stream to a domestic intelligent server via the RTSP protocol.

3. The method for monitoring the illegal use of mobile phones using domestically produced technology as described in claim 2, characterized in that, In step S21, the aspect ratio is maintained when scaling the image until the scaled length or width is exactly equal to the target size and the image pixels are less than or equal to the target image pixels. The blank part of the target size is filled with 0 for each of the RGB three channels.

4. The method for monitoring the illegal use of mobile phones as described in claim 2, characterized in that, Step S22 further includes installing Python C, ATeN, and Torch_MLU modules, as well as Cambricon PyTorch, Cambricon Catch, and the binary code compiled from the Cambricon Vision source code in the Phytium 2000+ CPU and Galaxy Kylin V10 operating system environment.

5. The method for monitoring the illegal use of mobile phones using domestically produced equipment as described in claim 2, characterized in that, In step S23, the hyperparameters are set to the following values: network depth is 0.75, network width is 1, learning rate is 0.01, stochastic gradient descent momentum is 0.99, and weight decay is 0.00025.

6. The method for monitoring the illegal use of mobile phones as described in claim 2, characterized in that, In step S32, the recorded information includes time, camera number of the video source, and camera deployment location. Statistics are compiled according to time and location, and the locations and time periods where unauthorized use of mobile phones is likely to occur are analyzed. This information can be viewed on the client terminal.

7. The method for monitoring the illegal use of mobile phones in China as described in claim 2, characterized in that, In step S32, alarms from the same camera within 5 seconds are not recorded repeatedly.

8. The method for monitoring the illegal use of mobile phones as described in claim 2, characterized in that, Step S4 specifically includes: S41. Analysis result filtering control: Utilizing the analysis results of each frame, if 10 consecutive frames within 15 frames are judged as unauthorized use of mobile phones, it is considered that there is a very high probability of unauthorized use of mobile phones, and it is necessary to close the access control. S42. Access Control Closure: When the system determines that it is necessary to close the access control, the domestic intelligent server automatically sends a closure command to the access control, and the access control cannot be unlocked internally. S43. Access Control Restriction Removal: Access is also prohibited to ordinary users outside the access control system. Only administrators can remove access control restrictions and open the access control system through the client terminal or from outside the access control system.

Images