Human body behavior recognition early warning system with redundancy function

Abstract

The invention discloses a human body behavior recognition early warning system with a redundancy function, which comprises a controller, a PC (Personal Computer), a network switch, a network camera and behavior recognition early warning modules, and is characterized in that each behavior recognition early warning module is provided with a redundant standby behavior recognition early warning module; the controller is connected with the plurality of behavior recognition early warning modules through 485 buses, each behavior recognition early warning module is connected with the network switch through a network cable, the network switch is connected with the plurality of network cameras at the same time, and at least two network cameras are installed in one application scene. The network switch and all the behavior identification early warning modules and the network cameras are located in the same local area network; a behavior identification early warning module used for the same identification task is matched with the IP address of a network camera, and one network camera can be matched with the IP addresses of a plurality of behavior identification early warning modules. And video data acquired by each network camera is transmitted to the behavior identification early warning module matched with the network camera through the network switch for behavior identification early warning. Through a behavior identification early warning module fault redundancy mechanism and a network camera fault redundancy mechanism, it is ensured that the overall identification effect of the system is not affected by the fault of a single device.

Description

Technical Field

[0001] The invention belongs to the field of monitoring technology, and in particular is a human behavior recognition and early warning system with redundant functions. Background Art

[0002] In traditional workplaces such as automated production lines and the chemical industry where labor cannot be replaced, production safety cannot be ignored, so it is necessary to monitor the abnormal behavior of operators. Electronic monitoring is an important monitoring method, but traditional electronic monitoring equipment only has functions such as video recording, query and playback, and cannot identify abnormal human behavior and issue timely warnings.

[0003] With the rapid development of network monitoring and artificial intelligence, although some intelligent monitoring technologies have emerged that can provide timely warnings for abnormal human behavior, the intelligent recognition algorithms of existing monitoring systems are usually solidified in the program of the main controller, which are designed and implemented for specific application scenarios and are only applicable to single behavior recognition in specific application scenarios. However, their applicability and recognition effects in other application scenarios are limited. When the application scenario changes and the behavior recognition task changes, the abnormal behavior recognition in the new application scenario cannot be achieved unless the main controller is replaced. In addition, when one or more monitoring devices fail, the corresponding video data cannot be obtained, which makes the system's behavior recognition task face the threat of interruption.

[0004] The invention patent with authorization announcement number CN110008867B discloses an early warning method, device and storage medium based on abnormal behavior of people. It tracks and judges the target object, uses a 3D convolutional neural network to analyze and process the judgment data, derives the behavior recognition type, and realizes the behavior recognition of moving targets.

[0005] The invention application with application publication number CN116416683A discloses a human posture and behavior recognition device and a posture and behavior recognition method thereof, which locates feature points of human information based on human video images and tracks and recognizes human postures in video images.

[0006] The invention application with application publication number CN114937308A discloses an intelligent security system based on human dynamic behavior feature recognition and its implementation method. The dynamic feature recognition module is used to identify the behavior in the data. When the behavior warning module detects dangerous behavior, it warns the user and confirms the dangerousness of the behavior.

[0007] The above-mentioned documents all have the following major defects: 1) There is only one core recognition module, which has poor scalability and cannot achieve redundant backup. When the core recognition module fails, the recognition task can only be interrupted; 2) The application scenario is single and can only recognize specific behaviors. When the application scenario changes and the recognition task changes, the behavior in the new application scenario cannot be recognized unless the main controller is replaced or the system is reconstructed; 3) When one or more monitoring devices in the application scenario fail, the recognition task is threatened with interruption, which in turn affects the recognition effect of the system. Summary of the invention

[0008] In view of the deficiencies in the prior art, the technical problem to be solved by the present invention is to provide a human behavior recognition and early warning system with redundant functions.

[0009] The technical solution adopted by the present invention to solve the technical problem is:

[0010] A human behavior recognition and early warning system with redundancy function, comprising a controller, a PC, a network switch and a network camera; characterized in that the system also comprises a behavior recognition and early warning module, each behavior recognition and early warning module is equipped with a redundant standby behavior recognition and early warning module;

[0011] The controller is connected to multiple behavior recognition and warning modules through the 485 bus. Each behavior recognition and warning module is connected to a network switch through a network cable. The network switch is connected to multiple network cameras at the same time. At least two network cameras are installed in one application scenario. The network switch and all behavior recognition and warning modules and network cameras are in the same local area network.

[0012] Before the system is run, identification tasks are assigned to the behavior recognition and warning modules and network cameras, and the behavior recognition and warning modules used for the same identification task are matched with the IP addresses of the network cameras. One network camera can be matched with the IP addresses of multiple behavior recognition and warning modules, so that the video data collected by each network camera is transmitted to the matching behavior recognition and warning module through the network switch for behavior recognition and warning.

[0013] During the operation of the system, the controller polls and monitors the operating status of each behavior recognition and warning module. If a behavior recognition and warning module fails, the controller controls the corresponding redundant standby behavior recognition and warning module to be enabled, so that the redundant standby behavior recognition and warning module replaces the faulty behavior recognition and warning module to continue the recognition task; if the video data of a network camera cannot be obtained or the network switch fails to recognize the IP address, another network camera that is working normally in the same application scenario and has a different recognition task from the faulty network camera is matched with the behavior recognition and warning module corresponding to the faulty network camera recognition task for IP address matching, so that the video data collected by the normally working network camera is transmitted to the behavior recognition and warning module corresponding to the faulty network camera recognition task for behavior recognition and warning, so that the normally working network camera replaces the faulty network camera.

[0014] Compared with the prior art, the present invention has the following beneficial effects:

[0015] (1) The present invention sets up multiple behavior recognition and warning modules and network cameras. The network cameras and behavior recognition and warning modules are matched with IP addresses according to the recognition tasks, so that the video data collected by each network camera can only be transmitted to the matching behavior recognition and warning module for analysis and processing. Each behavior recognition and warning module can only be used for the recognition and warning of one behavior. The behavior recognition and warning modules are independent of each other and do not interfere with each other. Under the premise of ensuring normal monitoring needs, the recognition and warning of multiple abnormal behaviors can be realized to the greatest extent. The controller calls the corresponding behavior recognition and warning module for behavior recognition and warning according to the recognition requirements of the current application scenario. When the application scenario changes and the recognition task changes, it only needs to add, replace or call the corresponding behavior recognition and warning module in the existing system, without replacing the controller or reconstructing the entire system. Therefore, the system can flexibly respond to different application scenarios, thereby improving the adaptability and flexibility of the system, optimizing resource utilization and enhancing the application efficiency of the system in different scenarios.

[0016] (2) The system has a high degree of reliability. Each behavior recognition warning module is equipped with a redundant standby behavior recognition warning module. When a behavior recognition warning module fails, the controller immediately receives the fault information and quickly starts the behavior recognition warning module fault redundancy mechanism. The controller sends a start signal to the corresponding redundant standby behavior recognition warning module through the 485 communication module. After receiving the start signal, the redundant standby behavior recognition warning module will automatically activate and replace the faulty behavior recognition warning module to continue to perform the unfinished recognition task. The behavior recognition warning module fault redundancy mechanism not only ensures the normal progress of the recognition task and maximizes the stability of the system, but also greatly simplifies the replacement process of the faulty behavior recognition warning module, making system maintenance more efficient and convenient. At the same time, it overcomes the defects of the existing monitoring system that the function is single and cannot be quickly recovered after failure.

[0017] Multiple network cameras are grouped according to application scenarios, and each network camera is assigned an independent IP address. When a network camera in an application scenario fails, the network camera failure redundancy mechanism will be activated, that is, the recognition task of the failed network camera will be transferred to another normally working network camera in the same application scenario that has a different recognition task from the failed network camera, and the network camera will be matched with the behavior recognition warning module corresponding to the recognition task of the failed network camera for IP address, and the video data collected by the network camera will be transmitted to the behavior recognition warning module corresponding to the recognition task of the failed network camera for processing. The network camera failure redundancy mechanism allows the system to quickly adjust monitoring resources and reallocate the recognition tasks of other network cameras in the same application scenario when a single network camera fails, ensuring that the system can continue to operate when a network camera fails, so that the corresponding behavior recognition warning will not be interrupted, improving the overall reliability of the system, and maintaining the overall security of the system.

[0018] (3) The behavior recognition and warning modules and network cameras for multiple recognition tasks can be quickly deployed into the existing system based on the IP address. There is no need to install a separate video surveillance system in the application scenario. The simultaneous recognition and warning of multiple behaviors such as human falls, sleeping, drowsiness, and wearing safety clothing can be achieved. The behavior recognition and warning module adopts a modular design, which is convenient for users to perform maintenance operations such as module replacement and function addition. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a structural schematic diagram of the present invention;

[0020] Figure 2 It is a configuration management system interface diagram of a PC of the present invention;

[0021] In the figure, 1-controller, 2-PC, 3-WIFI module, 4-485 communication module, 5-LED warning module, 6-sound alarm module, 7-behavior recognition warning module, 8-network switch, 9-network camera, 10-hard disk recorder. DETAILED DESCRIPTION

[0022] The present invention provides a human behavior recognition and early warning system with redundant functions (referred to as system, see Figures 1-2 ), including a controller 1, a PC 2, a WIFI module 3, a 485 communication module 4, an LED warning module 5, a sound alarm module 6, a behavior recognition warning module 7, a network switch 8, a network camera 9 and a hard disk recorder 10;

[0023] The PC 2 is connected to the controller 1 via the WIFI module 3, and the controller 1 is connected to the WIFI module 3 via the serial port to realize data transmission and reception between the controller 1 and the PC 2; the controller 1 is a communication control center connecting the PC 2 and the behavior recognition and warning module 7, and is connected to multiple behavior recognition and warning modules 7 and a hard disk recorder 10 via the 485 communication module 4, and the controller 1 is connected to the 485 communication module 4 via the I / O interface to realize data transmission and reception between the controller 1 and the behavior recognition and warning modules 7 and the hard disk recorder 10; each behavior recognition and warning module 7 has a network interface and is connected to the network switch 8 via a network cable, and each behavior recognition and warning module 7 is equipped with a redundant standby behavior recognition and warning module, and the redundant standby behavior recognition and warning module is connected to the behavior recognition and warning module. The recognition and warning modules 7 are completely identical, and are connected to the controller 1 and the network switch 8 through the 485 communication module 4 and the network cable respectively. The network switch 8 is connected to the hard disk recorder 10 and multiple network cameras 9 at the same time. Multiple network cameras 9 can be installed in different application scenarios, but it is necessary to ensure that at least two network cameras 9 are installed in the same application scenario, so as to collect video data of human behavior from different angles. The network switch 8, the hard disk recorder 10 and all the behavior recognition and warning modules 7 and network cameras 9 are in the same local area network; the controller 1 is connected to the LED warning module 5 and the sound alarm module 6 through the I / O interface respectively. The LED warning module 5 realizes abnormal behavior warning by flashing the LED light, and the sound alarm module 6 realizes abnormal behavior warning by voice broadcasting;

[0024] Each behavior recognition and early warning module 7 and redundant standby behavior recognition and early warning module stores a behavior recognition model, which can only recognize and warn one behavior; the video data collected by each network camera 9 is used for the recognition and early warning of at least one behavior to meet different monitoring requirements, that is, the video data collected by the network camera 9 is transmitted to the corresponding behavior recognition and early warning module 7 through the network switch 8 for behavior recognition and early warning, and in the behavior recognition and early warning module 7, abnormal behavior is visually warned in the form of wireframe annotation in the video frame; the number of network cameras 9 and behavior recognition and early warning modules 7 can be determined according to specific monitoring and behavior recognition and early warning requirements, and each network camera 9, behavior recognition and early warning module 7 and redundant standby behavior recognition and early warning module are equipped with an independent IP address, and the network cameras 9 and behavior recognition and early warning modules 7 used for the same recognition task are matched according to the IP address, and one network camera 9 can match the IP address with multiple behavior recognition and early warning modules 7, so that the video data collected by one network camera can be used for the recognition and early warning of multiple behaviors. This flexible configuration enables the system to monitor and manage efficiently. The IP addresses of network cameras 9 and behavior recognition and early warning modules 7 are grouped according to the application scenario (for example, in units of room numbers) to facilitate management and control.

[0025] The PC 2 is a general-purpose electronic computer that can run Visual Basic software, which is used for sending IP addresses, displaying status, adding devices and displaying warning information of network cameras and behavior recognition warning modules. The controller 1 has a 485 bus and a TCP / IP network interface, which is used to communicate with the PC 2 and the behavior recognition warning module 7. The core of the behavior recognition warning module is the RK3588 module, which has 8G memory, 32G data storage, an independent GPU and 6Tops computing power. The corresponding behavior recognition model transplantation is performed according to the recognition tasks such as falling, sleeping, drowsiness and wearing safety clothing. The network switch is a POE switch with power supply capacity of more than 16 Gigabit ports. The network camera is a TCP / IP camera with more than 200W pixels that can be connected to the network cable. The hard disk recorder is a TCP / IP network recorder with 485 communication function.

[0026] The behavior recognition model stored in the behavior recognition warning module is obtained based on neural network training. The training images of different recognition tasks are all taken based on actual scenes. There are no less than 1,000 images of each abnormal behavior to ensure the accuracy of model training. The trained behavior recognition model is transplanted into the behavior recognition warning module.

[0027] The working principle and workflow of the present invention are:

[0028] Before the system is run, firstly, a recognition task is assigned to each behavior recognition warning module 7 and network camera 9, that is, the recognition task is determined according to the behavior recognition model stored in the behavior recognition warning module 7, and the recognition task is determined according to the installation position of the network camera 9. The IP address of each behavior recognition warning module 7 and the network camera is set by the PC 2 and sent to the controller 1. The controller 1 matches the IP address of the behavior recognition warning module 7 and the network camera 9 of the same recognition task, so that the network switch 8 transmits the video data collected by the network camera 9 to the matching behavior recognition warning module 7 according to the IP address of the network camera 9 for behavior recognition warning.

[0029] When the behavior recognition and warning module 7 identifies abnormal behavior, the abnormal behavior information is sent to the controller 1 through the 485 communication module 4. The controller 1 controls the LED warning module 5 and the sound alarm module 6 to warn of abnormal behavior in the form of flashing LED lights and voice broadcasts. At the same time, the controller 1 transmits the warning information to the PC 2 through the WIFI module 3 for warning display, and controls the hard disk recorder 10 through the 485 communication module 4 to lock the video data of the abnormal behavior for query. The status of the corresponding network camera 9 and the behavior recognition and warning module 7 is changed to warning.

[0030] During the operation of the system, the controller 1 sets the initial communication mode of the 485 communication module 4 to "receive" to receive the status information of each behavior recognition and warning module, that is, the controller 1 monitors the operation status of each behavior recognition and warning module 7 in a polling manner. During the polling process, if a behavior recognition and warning module 7 fails and cannot communicate, the behavior recognition and warning module fault redundancy mechanism will be started, that is, the faulty behavior recognition and warning module will change its own status information from normal to error. When the controller 1 receives the error status information of the faulty behavior recognition and warning module through the polling mechanism, the communication mode of the 485 communication module 4 is modified to "send", and a startup instruction is sent to the corresponding redundant standby behavior recognition and warning module to enable the corresponding redundant standby behavior recognition and warning module, so that the redundant standby behavior recognition and warning module replaces the faulty behavior recognition and warning module to start working. At the same time, the controller 1 sends the fault information of the faulty behavior recognition and warning module to the PC 2 through the WIFI module 3, and the configuration management system of the PC 2 changes the status of the faulty behavior recognition and warning module from normal to "lost connection", and the status of the redundant standby behavior recognition and warning module from normal to online;

[0031] According to the monitoring requirements, the video data of human behavior is collected through the network camera 9, and transmitted to the matching behavior recognition warning module 7 through the network switch 8 for behavior recognition warning; if a network camera 9 fails and the video data cannot be obtained, or the network switch 8 fails to recognize the IP address of the network camera 9, the network camera fault redundancy mechanism will be started, that is, another network camera that works normally in the same application scenario and has a different recognition task from the faulty network camera is matched with the behavior recognition warning module corresponding to the faulty network camera recognition task, and the video data collected by the normal working network camera is transmitted to the behavior recognition warning module corresponding to the faulty network camera recognition task for behavior recognition warning, so that the normal working network camera replaces the faulty network camera, and the impact of the lack of video data of the faulty network camera on the recognition result is minimized. Specifically, assuming that network camera A is used for the fall recognition task, when network camera A fails, network camera B (network cameras A and B have different recognition tasks) is replaced by A for fall recognition, then network camera B is matched with the behavior recognition warning module corresponding to network camera A for IP address, and the network switch transmits the video data collected by network camera B to the behavior recognition warning module corresponding to network camera A for fall recognition.

[0032] In order to enhance the management and fault tracking capabilities of the system, the corresponding behavior recognition and warning module sends the IP address of the faulty network camera to the controller 1, and the controller 1 sends the fault information to the PC 2 through the WIFI module 3. The configuration management system of the PC 2 changes the status of the faulty network camera from normal to "lost connection", so that the administrator can track and handle the fault problem in time.

[0033] The behavior recognition warning module fault redundancy mechanism ensures that when a behavior recognition warning module fails, the corresponding recognition task is automatically and seamlessly transferred to the redundant standby behavior recognition warning module to continue; the network camera fault redundancy mechanism ensures that when a network camera fails, the corresponding recognition task is automatically and seamlessly transferred to another normally working network camera in the application scenario. These two mechanisms ensure that the overall recognition effect of the system will not be affected by the failure of a single behavior recognition warning module and network camera.

[0034] Any matters not described in the present invention are applicable to the prior art.

Claims

1. A human behavior recognition and early warning system with redundant functions, comprising a controller, a PC, a network switch and a network camera; characterized in that: The system also includes a behavior recognition warning module, each of which is equipped with a redundant backup behavior recognition warning module; The controller is connected to multiple behavior recognition and warning modules through the 485 bus. Each behavior recognition and warning module is connected to a network switch through a network cable. The network switch is connected to multiple network cameras at the same time. At least two network cameras are installed in one application scenario. The network switch and all behavior recognition and warning modules and network cameras are in the same local area network. Before the system is run, identification tasks are assigned to the behavior recognition and warning modules and network cameras, and the behavior recognition and warning modules used for the same identification task are matched with the IP addresses of the network cameras. One network camera can be matched with the IP addresses of multiple behavior recognition and warning modules, so that the video data collected by each network camera is transmitted to the matching behavior recognition and warning module through the network switch for behavior recognition and warning. During the operation of the system, the controller polls and monitors the operating status of each behavior recognition and warning module. If a behavior recognition and warning module fails, the controller controls the corresponding redundant standby behavior recognition and warning module to be enabled, so that the redundant standby behavior recognition and warning module replaces the faulty behavior recognition and warning module to continue the recognition task; if the video data of a network camera cannot be obtained or the network switch fails to recognize the IP address, another network camera that is working normally in the same application scenario and has a different recognition task from the faulty network camera is matched with the behavior recognition and warning module corresponding to the faulty network camera recognition task for IP address matching, so that the video data collected by the normally working network camera is transmitted to the behavior recognition and warning module corresponding to the faulty network camera recognition task for behavior recognition and warning, so that the normally working network camera replaces the faulty network camera.

2. A human behavior recognition and early warning system with redundant functions according to claim 1, characterized in that: The system also includes an LED warning module and a sound alarm module, and the controller is connected to the LED warning module and the sound alarm module respectively through an I / O interface.

3. A human behavior recognition and early warning system with redundant function according to claim 1 or 2, characterized in that: The system also includes a hard disk video recorder, which is connected to the controller and the network switch at the same time and is used to lock the video data of abnormal behavior.

4. The human behavior recognition and early warning system with redundant function according to claim 1, characterized in that: The behavior recognition warning module is transplanted with a behavior recognition model, and the behavior recognition model is constructed based on a neural network.

5. A human behavior recognition and early warning system with redundant function according to claim 1 or 4, characterized in that: In the behavior recognition and warning module, abnormal behaviors are visually warned in the form of wireframe annotations in the video frame.

Images