Implementation method and system for multi-protocol camera shooting distributed human shape detection

Abstract

The invention discloses a multi-protocol camera shooting distributed human shape detection realization method and system, and relates to the technical field of image detection, and the method comprises the steps: designing a unified management platform supporting the access of multiple protocols, so as to achieve the integration and management of cameras of different brands and models; a Python programming language is combined with a YOLOv8 target detection model to realize human shape detection, and a detection result is provided for the human shape detection control module according to a preset format; a Java programming language is adopted to develop a human shape detection control module, and the human shape detection control module utilizes a MyBatis framework to realize storage and modification functions of control parameters, utilizes cache optimization configuration reading, acquires a frame image by means of an FFmpeg tool, and performs human shape detection and alarm marking in combination with a YOLOv8 target detection model; and the time difference value of the alarm mark is calculated and compared with a preset control parameter, and whether a new alarm is triggered or not is intelligently judged. The method is used for realizing human shape detection and alarm.

Description

Technical Field

[0001] The present invention relates to the technical field of image detection, and in particular to a method and system for realizing distributed human figure detection by multi-protocol camera. Background Art

[0002] With the advancement and development of science and technology, cameras with different protocols, specifications and brands have appeared on the market, and various old and new cameras coexist. However, many old cameras do not integrate the basic function of human detection, and can only provide video streams for viewing and video saving; the human detection function of new cameras also has different processing results when processing the same environmental information; at the same time, the protocols used by various cameras are also different and cannot be shared with each other. If multiple cameras are managed at the same time, it is necessary to switch to different management platforms for viewing. It is also inconvenient to view alarm information and recordings. Users can only purchase products of the same brand for use, which seriously limits the diversity of user choices.

[0003] Based on the above, if the viewing and alarm functions of different types of cameras can be unified and standardized, it will greatly remove the restrictions on users when purchasing cameras, allowing users to only focus on the hardware when selecting cameras without considering future software usage issues. In addition, a unified external API interface can be provided to achieve resource sharing and remote control. Summary of the invention

[0004] In view of the needs and shortcomings of current technological development, the present invention provides a method and system for implementing distributed human detection using multi-protocol cameras to support the access of cameras with multiple protocols, and to perform real-time monitoring of people entering a designated electronic fence area, thereby generating real-time alarms.

[0005] In the first aspect, the present invention provides a method for implementing distributed human detection using multi-protocol cameras. The technical solution used to solve the above technical problems is as follows:

[0006] A method for implementing distributed human figure detection using multi-protocol cameras comprises the following steps:

[0007] S1. Design a unified management platform that supports access to ONVIF, RTSTP and GB28181 protocols to achieve integration and management of cameras of different brands and models;

[0008] S2, using Python programming language combined with YOLOv8 target detection model to realize human detection, and provide detection results to the human detection control module in a predetermined format;

[0009] S3. Use Java programming language to develop human detection control module. The human detection control module uses MyBatis framework to realize the storage and modification functions of control parameters, uses cache optimization configuration reading, uses FFmpeg tool to obtain frame images, and combines YOLOv8 target detection model to perform human detection and alarm marking;

[0010] S4. Calculate the time difference of the alarm mark and compare it with the preset control parameters to intelligently determine whether to trigger a new alarm.

[0011] Optionally, the unified management platform involved integrates camera devices from different manufacturers through a unified interface and protocol, provides a unified management and control interface, and includes cameras that are discovered and successfully connected through ONVIF, RTSTP and GB28181 protocols into a management list, while assigning a unique identifier to each camera; the unified management platform also supports device status monitoring function for real-time monitoring of the online and offline status of the camera.

[0012] Optionally, the human detection control module involved uses the MyBatis framework to implement the storage and modification functions of control parameters. The control parameters are stored in the local cache of the camera, specifically including: turning the human detection function on and off, the human detection alarm suppression time, the human detection alarm dwell time, the recording function on and off, and the electronic fence configuration of human detection.

[0013] Further optionally, step S3 is executed, the human detection control module uses the MyBatis framework to implement the storage and modification functions of the control parameters, uses the cache optimization configuration to read, obtains the frame image with the help of the FFmpeg tool, and combines the YOLOv8 target detection model to perform human detection and alarm marking. This process specifically includes:

[0014] Obtain all connected valid cameras and filter out valid cameras with human detection function enabled; wherein: the valid cameras refer to cameras in online state;

[0015] Check whether the human detection function is enabled on the current camera and whether the human intrusion detection range of the electronic fence is configured;

[0016] When the check results are all yes, use the FFmpeg tool to extract frame images from the video stream to prepare for subsequent human detection;

[0017] The YOLOv8 target detection model is used to process the extracted frame images and identify the human point information;

[0018] Get the stored electronic fence information and determine whether the human point is within the electronic fence: if the electronic fence point intersects with the identified human point, it is determined that a human has intruded. At this time, an alarm mark is set for this determination result and the relevant information is saved in the database, but the intrusion alarm will not be triggered immediately.

[0019] Optionally, the step S4 involved specifically includes:

[0020] S4.1. extracting the time information of each alarm mark from the alarm mark data;

[0021] S4.2, calculating the maximum time difference between two valid alarm marks, wherein the maximum time difference represents the actual residence time of the personnel in the monitoring area;

[0022] S4.3, extract the time when the last alarm was sent, compare it with the current time, and calculate the alarm elapsed time since the last alarm;

[0023] S4.4. Compare the calculated maximum time difference with the configured human detection alarm dwell time:

[0024] a) If the maximum time difference exceeds the human detection alarm residence time, it is considered that the person's residence time in the monitoring area is abnormal, and step S4.5 is continued.

[0025] b) If the maximum time difference does not exceed the human detection alarm dwell time, the person's behavior is in line with expectations and no alarm needs to be triggered;

[0026] S4.5. Compare the calculated alarm elapsed time with the configured human detection alarm suppression time:

[0027] c) If the alarm elapsed time has not reached the human detection alarm suppression time, even if the maximum time difference exceeds the human detection alarm dwell time, no new alarm will be triggered.

[0028] d) If the alarm elapsed time reaches the human detection alarm suppression time, a new alarm is triggered.

[0029] In a second aspect, the present invention provides a system for implementing distributed human detection using multi-protocol cameras. The technical solutions adopted to solve the above technical problems are as follows:

[0030] A multi-protocol camera distributed human detection implementation system, the structure of which includes:

[0031] Unified management module supports access to ONVIF, RTSTP and GB28181 protocols to achieve integration and management of cameras of different brands and models;

[0032] A human figure detection module is used to implement human figure detection by combining Python programming language with YOLOv8 target detection model, and provide detection results to the human figure detection control module in a predetermined format;

[0033] The human detection control module is developed in Java programming language, which uses MyBatis framework to realize the storage and modification functions of control parameters, optimizes configuration reading by cache, obtains frame images with the help of FFmpeg tool, and performs human detection and alarm marking in combination with YOLOv8 target detection model;

[0034] The alarm judgment module is used to calculate the time difference of the alarm mark and compare it with the preset control parameters to intelligently determine whether a new alarm is triggered.

[0035] Optionally, the unified management module involved integrates camera devices from different manufacturers through a unified interface and protocol, provides a unified management and control interface, and includes cameras that are discovered and successfully connected through ONVIF, RTSTP and GB28181 protocols into a management list, while assigning a unique identifier to each camera; the unified management module also supports device status monitoring function for real-time monitoring of the online and offline status of the camera.

[0036] Optionally, the human detection control module involved uses the MyBatis framework to implement the storage and modification functions of control parameters. The control parameters are stored in the local cache of the camera, specifically including: turning the human detection function on and off, the human detection alarm suppression time, the human detection alarm dwell time, the recording function on and off, and the electronic fence configuration of human detection.

[0037] Further optionally, the human figure detection control module involved specifically includes:

[0038] An acquisition and screening unit is used to acquire all connected valid cameras and screen out valid cameras with a human detection function enabled; wherein: the valid camera refers to an online camera;

[0039] The inspection unit is used to check whether the human detection function of the current camera is turned on and whether the human intrusion judgment range of the electronic fence is configured;

[0040] A frame image extraction unit is used to extract frame images from the video stream using the FFmpeg tool when the inspection results of the inspection unit are all yes, so as to prepare for subsequent human detection;

[0041] A calling unit is used to call the YOLOv8 target detection model of the human figure detection module, process the extracted frame image, and identify the human figure point information;

[0042] The judgment marking unit is used to obtain the stored electronic fence information and judge whether the human point is located within the electronic fence: if the electronic fence point intersects with the identified human point, it is judged that a human has intruded. At this time, an alarm mark is set for this judgment result and the relevant information is saved in the database, but the intrusion alarm will not be triggered immediately.

[0043] Further optionally, the alarm judgment module involved specifically includes:

[0044] A time extraction unit, used to extract the time information of each alarm mark from the alarm mark data;

[0045] A difference calculation unit, used to calculate the maximum time difference between two valid alarm marks, wherein the maximum time difference represents the actual residence time of the personnel in the monitoring area;

[0046] An extraction calculation unit, used to extract the time when the alarm was last sent, and compare it with the current time to calculate the alarm elapsed time since the last alarm;

[0047] The judgment processing unit 1 is used to compare the calculated maximum time difference with the configured human detection alarm residence time: a) if the maximum time difference exceeds the human detection alarm residence time, it is considered that the person's residence time in the monitoring area is abnormal, and the judgment processing unit 2 is continued to be executed; b) if the maximum time difference does not exceed the human detection alarm residence time, the person's behavior is in line with expectations, and there is no need to trigger an alarm at this time;

[0048] Judgment processing unit 2 is used to compare the calculated alarm elapsed time with the configured human detection alarm suppression time: c) if the alarm elapsed time has not reached the human detection alarm suppression time, even if the maximum time difference exceeds the human detection alarm dwell time, no new alarm is triggered, d) if the alarm elapsed time reaches the human detection alarm suppression time, a new alarm is triggered.

[0049] The method and system for implementing distributed human figure detection by multi-protocol camera of the present invention have the following beneficial effects compared with the prior art:

[0050] The present invention unifies the management of cameras with different protocols, making it convenient for managers to conduct comprehensive monitoring and scheduling, reducing the complexity and workload of managing multiple camera systems separately; it can configure human detection and alarms according to the usage scenarios of different cameras, avoiding singleness and better meeting business needs; it can perform automated human detection and alarms, replacing part of manual monitoring, reducing labor costs, effectively reducing false alarms and repeated alarms, and improving the intelligence level of the monitoring system and user satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] AttachedFigure 1 is a flow chart of a method according to Embodiment 1 of the present invention;

[0052] Attached Figure 2 is a diagram of a human figure detection control process according to a first embodiment of the present invention;

[0053] Attached Figure 3 is a flowchart of alarm generation according to the first embodiment of the present invention;

[0054] Attached Figure 4 This is a module connection block diagram of the second embodiment of the present invention. DETAILED DESCRIPTION

[0055] In order to make the technical solution, the technical problem solved and the technical effect of the present invention more clearly understood, the technical solution of the present invention is clearly and completely described below in conjunction with specific embodiments.

[0056] Embodiment 1:

[0057] Reference Figure 1 This embodiment proposes a method for implementing distributed human detection using multi-protocol cameras, which includes the following steps:

[0058] S1. Design a unified management platform that supports access to ONVIF, RTSTP and GB28181 protocols to achieve integration and management of cameras of different brands and models.

[0059] In this step, the unified management platform integrates camera devices from different manufacturers through a unified interface and protocol, provides a unified management and control interface, and includes cameras that are discovered and successfully connected through ONVIF, RTSTP and GB28181 protocols into a management list, while assigning a unique identifier to each camera. The unified management platform also supports device status monitoring for real-time monitoring of the online and offline status of the camera.

[0060] S2. Use Python programming language in combination with the YOLOv8 target detection model to implement human detection and provide detection results to the human detection control module in a predetermined format.

[0061] The specific implementation of this step includes the following two steps:

[0062] S2.1. Environment construction and model preparation: In the Python environment, install the relevant dependency libraries required by the YOLOv8 model, such as torch, numpy, opencv-python, etc., to ensure that the environment can support the loading and running of the model; download the pre-trained YOLOv8 human detection model weight file. According to the needs of the actual application scenario, you can choose model versions with different precisions (such as lightweight, high precision, etc.) to balance the detection speed and accuracy;

[0063] S2.2, video stream processing and detection implementation: obtain the video stream data from the camera from the unified management platform, and use Python's video processing library (such as cv2.VideoCapture, etc.) to read the video stream frame by frame; input each frame of image data into the YOLOv8 target detection model, which will perform feature extraction, target classification and positioning according to its internal neural network structure, and finally output whether there is a human figure in the image and the location coordinates of the human figure; provide the detection results to the human detection control module in a predetermined format (for example, it can be defined as a JSON format data structure containing fields such as detection timestamp, camera identifier, human figure location coordinates, confidence, etc.), so that subsequent modules can further process and analyze the detection results.

[0064] S3. Use Java programming language to develop human detection control module. The human detection control module uses MyBatis framework to realize the storage and modification functions of control parameters, uses cache optimization configuration reading, uses FFmpeg tool to obtain frame images, and combines YOLOv8 target detection model to perform human detection and alarm marking.

[0065] This step uses the MyBatis framework to implement the storage and modification functions of the control parameters. Specifically, it requires: integrating the MyBatis framework in the Java project, configuring the connection parameters with the database (such as MySQL, PostgreSQL and other commonly used relational databases), and creating a database table structure for storing human detection control parameters. The control parameters are stored in the local cache of the camera, including: turning on and off the human detection function, the human detection alarm suppression time, the human detection alarm stay time, the opening and closing of the recording function, and the electronic fence configuration of human detection. It should be added that this implementation requires the development of the corresponding DAO (Data Access Object) layer interface and implementation class. Through the SQL mapping file of MyBatis, the control parameters in the database can be added, deleted, modified, and checked, which is convenient for the operation and maintenance personnel or the system to dynamically adjust the relevant configuration parameters of human detection according to actual needs.

[0066] Combined with Figure 2 The specific implementation process of this step includes:

[0067] S3.1. Obtain all connected valid cameras and select valid cameras with human detection function enabled; wherein: the valid cameras refer to cameras in online state;

[0068] S3.2. Check whether the human detection function is enabled on the current camera and whether the human intrusion detection range of the electronic fence is configured;

[0069] S3.3. When all the check results are yes, use the FFmpeg tool to extract frame images from the video stream obtained from the unified management platform and convert them into image formats suitable for YOLOv8 target detection model input (such as RGB format, etc.) to prepare for subsequent human detection;

[0070] S3.4, use the YOLOv8 target detection model to process the extracted frame image and identify the human point information;

[0071] S3.5. Obtain the stored electronic fence information and determine whether the human point is located within the electronic fence: If the electronic fence point intersects with the identified human point, it is determined that a human has intruded. At this time, an alarm mark is set for this determination result, and the relevant information is saved in the database, but the intrusion alarm will not be triggered immediately.

[0072] S4, calculate the time difference of the alarm mark, and compare it with the preset control parameters, intelligently determine whether to trigger a new alarm, and combine the attached Figure 3 , this process specifically includes:

[0073] S4.1. extracting the time information of each alarm mark from the alarm mark data;

[0074] S4.2, calculating the maximum time difference between two valid alarm marks, wherein the maximum time difference represents the actual residence time of the personnel in the monitoring area;

[0075] S4.3, extract the time when the last alarm was sent, compare it with the current time, and calculate the alarm elapsed time since the last alarm;

[0076] S4.4. Compare the calculated maximum time difference with the configured human detection alarm dwell time:

[0077] a) If the maximum time difference exceeds the human detection alarm residence time, it is considered that the person's residence time in the monitoring area is abnormal, and step S4.5 is continued.

[0078] b) If the maximum time difference does not exceed the human detection alarm dwell time, the person's behavior is in line with expectations and no alarm needs to be triggered;

[0079] S4.5. Compare the calculated alarm elapsed time with the configured human detection alarm suppression time:

[0080] c) If the alarm elapsed time has not reached the human detection alarm suppression time, even if the maximum time difference exceeds the human detection alarm dwell time, no new alarm will be triggered.

[0081] d) If the alarm elapsed time reaches the human detection alarm suppression time, a new alarm is triggered.

[0082] Embodiment 2:

[0083] Combined with Figure 4 This embodiment proposes a multi-protocol camera distributed human detection implementation system, whose structure includes:

[0084] Unified management module supports access to ONVIF, RTSTP and GB28181 protocols to achieve integration and management of cameras of different brands and models;

[0085] A human figure detection module is used to implement human figure detection by combining Python programming language with YOLOv8 target detection model, and provide detection results to the human figure detection control module in a predetermined format;

[0086] The human detection control module is developed in Java programming language, which uses MyBatis framework to realize the storage and modification functions of control parameters, optimizes configuration reading by cache, obtains frame images with the help of FFmpeg tool, and performs human detection and alarm marking in combination with YOLOv8 target detection model;

[0087] The alarm judgment module is used to calculate the time difference of the alarm mark and compare it with the preset control parameters to intelligently determine whether a new alarm is triggered.

[0088] In this embodiment, the unified management module involved integrates camera devices from different manufacturers through a unified interface and protocol, provides a unified management and control interface, and incorporates cameras that are discovered and successfully connected through ONVIF, RTSTP and GB28181 protocols into a management list, while assigning a unique identifier to each camera; the unified management module also supports device status monitoring function, which is used to monitor the online and offline status of the camera in real time.

[0089] In this embodiment, the human detection control module involved uses the MyBatis framework to implement the storage and modification functions of control parameters. The control parameters are stored in the local cache of the camera, including: turning on and off the human detection function, the human detection alarm suppression time, the human detection alarm dwell time, turning on and off the recording function, and the electronic fence configuration of human detection.

[0090] The humanoid detection control modules involved specifically include:

[0091] An acquisition and screening unit is used to acquire all connected valid cameras and screen out valid cameras with a human detection function enabled; wherein: the valid camera refers to an online camera;

[0092] The inspection unit is used to check whether the human detection function of the current camera is turned on and whether the human intrusion judgment range of the electronic fence is configured;

[0093] A frame image extraction unit is used to extract frame images from the video stream using the FFmpeg tool when the inspection results of the inspection unit are all yes, so as to prepare for subsequent human detection;

[0094] A calling unit is used to call the YOLOv8 target detection model of the human figure detection module, process the extracted frame image, and identify the human figure point information;

[0095] The judgment marking unit is used to obtain the stored electronic fence information and judge whether the human point is located within the electronic fence: if the electronic fence point intersects with the identified human point, it is judged that a human has intruded. At this time, an alarm mark is set for this judgment result and the relevant information is saved in the database, but the intrusion alarm will not be triggered immediately.

[0096] In this embodiment, the alarm judgment module involved specifically includes:

[0097] A time extraction unit, used to extract the time information of each alarm mark from the alarm mark data;

[0098] A difference calculation unit, used to calculate the maximum time difference between two valid alarm marks, wherein the maximum time difference represents the actual residence time of the personnel in the monitoring area;

[0099] An extraction calculation unit, used to extract the time when the alarm was last sent, and compare it with the current time to calculate the alarm elapsed time since the last alarm;

[0100] The judgment processing unit 1 is used to compare the calculated maximum time difference with the configured human detection alarm residence time: a) if the maximum time difference exceeds the human detection alarm residence time, it is considered that the person's residence time in the monitoring area is abnormal, and the judgment processing unit 2 is continued to be executed; b) if the maximum time difference does not exceed the human detection alarm residence time, the person's behavior is in line with expectations, and there is no need to trigger an alarm at this time;

[0101] Judgment processing unit 2 is used to compare the calculated alarm elapsed time with the configured human detection alarm suppression time: c) if the alarm elapsed time has not reached the human detection alarm suppression time, even if the maximum time difference exceeds the human detection alarm dwell time, no new alarm is triggered, d) if the alarm elapsed time reaches the human detection alarm suppression time, a new alarm is triggered.

[0102] In summary, the implementation method and system of multi-protocol camera distributed human detection of the present invention can support the access of cameras with multiple protocols, and can perform real-time monitoring of people entering the designated electronic fence area, and then generate real-time alarms.

[0103] The above specific examples are used to explain the principles and implementation methods of the present invention in detail. These examples are only used to help understand the core technical content of the present invention. Based on the above specific embodiments of the present invention, any improvements and modifications made by technicians in this technical field without departing from the principles of the present invention should fall within the scope of patent protection of the present invention.

Claims

1. A method for implementing distributed human detection using multi-protocol cameras, characterized in that: The steps include: S1. Design a unified management platform that supports access to ONVIF, RTSTP and GB28181 protocols to achieve integration and management of cameras of different brands and models; S2, using Python programming language combined with YOLOv8 target detection model to realize human detection, and provide detection results to the human detection control module in a predetermined format; S3. Use Java programming language to develop human detection control module. The human detection control module uses MyBatis framework to realize the storage and modification functions of control parameters, uses cache optimization configuration reading, uses FFmpeg tool to obtain frame images, and combines YOLOv8 target detection model to perform human detection and alarm marking; S4. Calculate the time difference of the alarm mark and compare it with the preset control parameters to intelligently determine whether to trigger a new alarm.

2. According to claim 1, a method for implementing distributed human detection using multi-protocol cameras is characterized in that: The unified management platform integrates camera devices from different manufacturers through a unified interface and protocol, provides a unified management and control interface, and integrates cameras discovered and successfully connected through ONVIF, RTSTP and GB28181 protocols into a management list, while assigning a unique identifier to each camera; The unified management platform also supports a device status monitoring function for real-time monitoring of the online and offline status of the camera.

3. The method for implementing distributed human detection using multi-protocol camera according to claim 1, characterized in that: The human detection control module uses the MyBatis framework to implement the storage and modification functions of control parameters. The control parameters are stored in the local cache of the camera, specifically including: turning the human detection function on and off, the human detection alarm suppression time, the human detection alarm stay time, the turning on and off of the recording function, and the electronic fence configuration of human detection.

4. The method for implementing distributed human figure detection using multi-protocol camera according to claim 3, characterized in that: Execute step S3. The human detection control module uses the MyBatis framework to implement the storage and modification functions of the control parameters, uses the cache to optimize the configuration reading, obtains the frame image with the help of the FFmpeg tool, and combines the YOLOv8 target detection model to perform human detection and alarm marking. This process specifically includes: Obtain all connected valid cameras and filter out valid cameras with human detection function enabled; wherein: the valid cameras refer to cameras in online state; Check whether the human detection function is enabled on the current camera and whether the human intrusion detection range of the electronic fence is configured; When the check results are all yes, use the FFmpeg tool to extract frame images from the video stream to prepare for subsequent human detection; The YOLOv8 target detection model is used to process the extracted frame images and identify the human point information; Get the stored electronic fence information and determine whether the human point is within the electronic fence: if the electronic fence point intersects with the identified human point, it is determined that a human has intruded. At this time, an alarm mark is set for this determination result and the relevant information is saved in the database, but the intrusion alarm will not be triggered immediately.

5. The method for implementing distributed human figure detection using multi-protocol camera according to claim 4, characterized in that: The step S4 specifically includes: S4.

1. extracting the time information of each alarm mark from the alarm mark data; S4.2, calculating the maximum time difference between two valid alarm marks, wherein the maximum time difference represents the actual residence time of the personnel in the monitoring area; S4.3, extract the time when the last alarm was sent, compare it with the current time, and calculate the alarm elapsed time since the last alarm; S4.

4. Compare the calculated maximum time difference with the configured human detection alarm dwell time: a) If the maximum time difference exceeds the human detection alarm residence time, it is considered that the person's residence time in the monitoring area is abnormal, and step S4.5 is continued. b) If the maximum time difference does not exceed the human detection alarm dwell time, the person's behavior is in line with expectations and no alarm needs to be triggered; S4.

5. Compare the calculated alarm elapsed time with the configured human detection alarm suppression time: c) If the alarm elapsed time has not reached the human detection alarm suppression time, even if the maximum time difference exceeds the human detection alarm dwell time, no new alarm will be triggered. d) If the alarm elapsed time reaches the human detection alarm suppression time, a new alarm is triggered.

6. A multi-protocol camera distributed human detection implementation system, characterized in that: Its structure includes: Unified management module supports access to ONVIF, RTSTP and GB28181 protocols to achieve integration and management of cameras of different brands and models; A human figure detection module is used to implement human figure detection by combining Python programming language with YOLOv8 target detection model, and provide detection results to the human figure detection control module in a predetermined format; The human detection control module is developed in Java programming language, which uses MyBatis framework to realize the storage and modification functions of control parameters, optimizes configuration reading by cache, obtains frame images with the help of FFmpeg tool, and performs human detection and alarm marking in combination with YOLOv8 target detection model; The alarm judgment module is used to calculate the time difference of the alarm mark and compare it with the preset control parameters to intelligently determine whether a new alarm is triggered.

7. A system for implementing distributed human detection using multi-protocol camera according to claim 6, characterized in that: The unified management module integrates camera devices from different manufacturers through a unified interface and protocol, provides a unified management and control interface, and integrates cameras discovered and successfully connected through ONVIF, RTSTP and GB28181 protocols into a management list, while assigning a unique identifier to each camera; The unified management module also supports a device status monitoring function for real-time monitoring of the online and offline status of the camera.

8. The system for implementing distributed human detection using multi-protocol camera according to claim 6, characterized in that: The human detection control module uses the MyBatis framework to implement the storage and modification functions of control parameters. The control parameters are stored in the local cache of the camera, specifically including: turning the human detection function on and off, the human detection alarm suppression time, the human detection alarm stay time, the turning on and off of the recording function, and the electronic fence configuration of human detection.

9. A system for implementing distributed human detection using multi-protocol camera according to claim 8, characterized in that: The human figure detection control module specifically includes: An acquisition and screening unit is used to acquire all connected valid cameras and screen out valid cameras with a human detection function enabled; wherein: the valid camera refers to an online camera; The inspection unit is used to check whether the human detection function of the current camera is turned on and whether the human intrusion judgment range of the electronic fence is configured; A frame image extraction unit is used to extract frame images from the video stream using the FFmpeg tool when the inspection results of the inspection unit are all yes, so as to prepare for subsequent human detection; A calling unit is used to call the YOLOv8 target detection model of the human figure detection module, process the extracted frame image, and identify the human figure point information; The judgment marking unit is used to obtain the stored electronic fence information and judge whether the human point is located within the electronic fence: if the electronic fence point intersects with the identified human point, it is judged that a human has intruded. At this time, an alarm mark is set for this judgment result and the relevant information is saved in the database, but the intrusion alarm will not be triggered immediately.

10. A system for implementing distributed human detection using multi-protocol cameras according to claim 9, characterized in that: The alarm judgment module specifically includes: A time extraction unit, used to extract the time information of each alarm mark from the alarm mark data; A difference calculation unit, used to calculate the maximum time difference between two valid alarm marks, wherein the maximum time difference represents the actual residence time of the personnel in the monitoring area; An extraction calculation unit, used to extract the time when the alarm was last sent, and compare it with the current time to calculate the alarm elapsed time since the last alarm; The judgment processing unit 1 is used to compare the calculated maximum time difference with the configured human detection alarm residence time: a) if the maximum time difference exceeds the human detection alarm residence time, it is considered that the person's residence time in the monitoring area is abnormal, and the judgment processing unit 2 is continued to be executed; b) if the maximum time difference does not exceed the human detection alarm residence time, the person's behavior is in line with expectations, and there is no need to trigger an alarm at this time; Judgment processing unit 2 is used to compare the calculated alarm elapsed time with the configured human detection alarm suppression time: c) if the alarm elapsed time has not reached the human detection alarm suppression time, even if the maximum time difference exceeds the human detection alarm dwell time, no new alarm is triggered, d) if the alarm elapsed time reaches the human detection alarm suppression time, a new alarm is triggered.

Images