Regional intrusion detection system and method

Abstract

The invention provides a region intrusion detection system and method, and the system comprises a video collection module which is used for collecting the current image data of a detection region in real time; the AI analysis module is used for performing primary screening on the human body area according to the image data, identifying preset key points of the human body area and generating a judgment result of the intrusion behavior according to the identified preset key points; the PLC control module is used for executing the sudden stop operation of the production equipment when the AI analysis module judges that the intrusion behavior occurs in the detection area; the management platform module is used for providing a visual configuration interface of a dynamic detection area, can accurately identify human body invasion behaviors of a dangerous operation area in a factory, greatly reduces the safety accident rate of the factory, can accurately identify human body invasion behaviors by identifying preset key points of a human body area, and improves the safety of the factory. False alarm and unnecessary shutdown operation are avoided, and the production efficiency is further improved.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial safety protection and intelligent monitoring, and in particular to a regional intrusion detection system and method. Background Art

[0002] With the rapid development of industrial automation technology, robotic systems have been widely used in rubber tire manufacturing. High-speed swinging industrial robots have significantly improved production efficiency and process precision, particularly in critical processes such as tire blank molding and vulcanization. However, the high overlap between the dynamic danger zones of the robotic workcell (such as the arm's swing radius and the clamping space) and human activity areas creates a persistent risk of mechanical injury. While safety guards are widely used to isolate areas, traditional safety technologies still exhibit significant shortcomings in practical applications.

[0003] Current safety protection solutions primarily rely on non-contact detection technology, typically millimeter-wave radar scanners and infrared safety barrier systems. These technologies operate by emitting a detection beam to form an electronic safety fence. Interference from the beam triggers an emergency stop signal when it is blocked. However, in the complex industrial environment of tire production, this technology faces two key technical bottlenecks. First, the persistent presence of interfering media such as steam, dust particles, and hot oil mist in the production environment significantly degrades the propagation quality of the detection beam, leading to signal attenuation or false triggering. This is particularly true in the high-temperature and high-humidity environment of the vulcanization workshop, where false alarm rates are high. Second, existing detection devices can only determine whether there are obstructions within the detection area and cannot perform biometric recognition. Consequently, dropped tools, slipped materials, or intrusions from equipment components can trigger protection mechanisms, causing unplanned downtime. Furthermore, existing technologies have detection blind spots when a person intrudes into the danger zone. For example, when a robot performs three-dimensional complex motion, the real-time boundary of the dynamic danger zone is difficult for fixed-mounted detection devices to accurately capture. This can result in parts of a person (such as an arm) entering the danger zone without being detected in time.

[0004] Therefore, the existing technology needs to be further developed. Summary of the Invention

[0005] The purpose of the present invention is to overcome the above technical deficiencies and provide a regional intrusion detection system and method to solve the problems existing in the prior art.

[0006] To achieve the above technical objectives, according to a first aspect of the present invention, the present invention provides a regional intrusion detection system, comprising: Video acquisition module, used to collect the current image data of the detection area in real time; An AI analysis module, in communication with the video acquisition module, is configured to perform a primary screening of a human body region based on the image data, identify preset key points of the human body region, and generate a determination result of an intrusion behavior based on the identified preset key points; A PLC control module is in communication with the AI analysis module and is configured to execute an emergency stop operation of the production equipment when the AI analysis module determines that an intrusion has occurred in the detection area; The management platform module is communicated with the video acquisition module, AI analysis module, and PLC control module to provide a visual configuration interface for the dynamic detection area.

[0007] Specifically, the video acquisition module includes a dual-spectrum camera group for synchronously acquiring visible light images, infrared thermal imaging, and depth perception data, and the video acquisition module includes a temperature compensation algorithm module for eliminating environmental thermal noise; The image data includes visible light images, infrared thermal images, and depth perception data.

[0008] Specifically, the AI analysis module includes a two-stage detection network constructed based on YOLOv7 and HRNet, the YOLOv7 detection network is used to perform primary screening of human body areas in the image data, and the HRNet detection network is used to identify preset key points of human body areas.

[0009] Specifically, the AI analysis module is in communication with the production equipment and is configured to execute an emergency stop operation of the production equipment when the AI analysis module determines that an intrusion has occurred in the detection area.

[0010] Specifically, the area intrusion detection system further includes an alarm module, which is communicatively connected to the AI analysis module. The alarm module is configured to execute an alarm operation when the AI analysis module determines that an intrusion has occurred in the detection area.

[0011] According to a second aspect of the present invention, there is provided a method for detecting an area intrusion, comprising: S100, real-time acquisition of current image data of the detection area; S200, performing a primary screening of a human body region according to the image data, identifying preset key points of the human body region, and generating a determination result of an intrusion behavior according to the identified preset key points; S300: When it is determined that an intrusion occurs in the detection area, an instruction to shut down the production equipment is output, thereby controlling the production equipment to shut down.

[0012] Specifically, the method for real-time acquisition of current image data of the detection area includes: The dual-spectral camera group synchronously collects the current visible light image, infrared thermal imaging and depth perception data of the detection area, and uses a temperature compensation algorithm to eliminate environmental thermal noise.

[0013] Specifically, the step S200 includes: Use the YOLOv7 detection model to perform primary screening of the human body area in the image data, generate a target detection frame, and determine whether the intersection-over-union ratio of the target detection frame and a preset activity restricted area is greater than or equal to a first preset threshold, and the confidence level is greater than or equal to a second preset threshold. If so, detect preset key points in the human body area; If not, continue to monitor the current image data of the detection area.

[0014] Specifically, the method for detecting preset key points in a human body region includes: The HRNet detection model is used to locate the coordinates of the preset key points in the human body area, record the number of frames in which each preset key point continuously invades the preset activity restricted area, and calculate the motion vector of each preset key point; When the number of frames in which any preset key point continuously intrudes into the preset restricted area reaches a third preset threshold, and the direction of the motion vector points to the production equipment, it is determined that an intrusion has occurred in the current detection area, and a command to shut down the production equipment is synchronously output to control the shutdown of the production equipment; Otherwise, it is determined that no intrusion has occurred in the current detection area, that is, the production equipment is operating normally.

[0015] Specifically, the method further includes: When it is determined that an intrusion occurs in the current detection area, an alarm prompt is issued, and an intrusion prompt box pops up through the visual configuration interface of the management platform module. The content of the intrusion prompt box includes the coordinates of the preset key points that intrude into the preset activity restricted area, the intrusion time, and the human body posture.

[0016] Beneficial effects: The present invention provides an area intrusion detection system and method. The area intrusion detection system includes a video acquisition module, an AI analysis module, a PLC control module and a management platform module. The AI analysis module can accurately identify human intrusion behavior in dangerous working areas in a factory. When intrusion behavior is detected, the PLC control module responds quickly and synchronously controls the corresponding production equipment to stop running, greatly reducing the incidence of safety accidents in the factory. At the same time, by identifying preset key points in the human body area, human intrusion behavior can be accurately identified, avoiding false alarms and unnecessary shutdown operations, and achieving further improvement in production efficiency. Compared with manual inspections, it saves manpower and greatly improves the intelligence, usability and reliability of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 Schematic diagram of the composition of the regional intrusion detection system provided in a specific embodiment of the present invention; Figure 2 is a flow chart of a regional intrusion detection method provided in a specific embodiment of the present invention; Figure 3 It is a flow chart of the regional intrusion detection steps provided in a specific embodiment of the present invention. DETAILED DESCRIPTION

[0018] In order to enable those skilled in the art to better understand the technical solution of the present invention, the technical solution of the present invention is clearly and completely described below in conjunction with the drawings of the present invention. Based on the embodiments in this application, other similar embodiments obtained by ordinary technicians in this field without making creative work should fall within the scope of protection of this application. In addition, the directional words mentioned in the following embodiments, such as "up", "down", "left", "right", etc., are only reference to the directions of the drawings. Therefore, the directional words used are used to illustrate rather than limit the invention.

[0019] The present invention will be further described below with reference to the accompanying drawings and preferred embodiments.

[0020] See also Figure 1 This embodiment provides an area intrusion detection system, including a video acquisition module, an AI analysis module, a PLC control module, and a management platform module. It can accurately identify intrusion behaviors in dangerous operating areas in a factory and synchronously control the corresponding production equipment to stop operating, thereby significantly reducing the occurrence rate of safety accidents in the factory.

[0021] See also Figure 1 In the area intrusion detection system of this embodiment, a video acquisition module is included for real-time acquisition of current image data of the detection area. The video acquisition module includes a dual-spectrum camera group for synchronously acquiring visible light images, infrared thermal imaging, and depth perception data. The video acquisition module includes a temperature compensation algorithm module for eliminating environmental thermal noise. The image data includes visible light images, infrared thermal imaging, and depth perception data.

[0022] It is understandable that the built-in temperature compensation algorithm of the video acquisition module can eliminate the interference of high-temperature steam, capture multimodal data such as visible light, infrared thermal imaging, depth perception, etc. in real time, and adapt to complex industrial scenes such as high temperature, low illumination, and dynamic background. Through the cross-modal feature alignment network, the key point detection of visible light images and the human body contour of infrared thermal imaging are integrated, further improving the detection robustness in complex scenes.

[0023] Furthermore, this embodiment utilizes a dual-spectral camera system with a temperature compensation algorithm to simultaneously fuse visible light, infrared thermal imaging, and depth sensing data. This enables reliable detection in complex working conditions, including strong and weak light conditions and fog. This addresses the issue of traditional monocular cameras failing due to interference from media such as steam, dust particles, and high-temperature oil mist. The combination of infrared thermal imaging and depth data allows the system to penetrate common work uniforms and identify human outlines, further mitigating the risk of intrusion through disguise.

[0024] See also Figure 1 In the regional intrusion detection system of this embodiment, an AI analysis module is further included, which is in communication with the video acquisition module and is used to perform a primary screening of the human body area based on the image data, identify preset key points of the human body area, and generate a determination result of the intrusion behavior based on the identified preset key points; Specifically, the AI analysis module includes a two-stage detection network built based on YOLOv7 and HRNet. The YOLOv7 detection network is used to perform primary screening of human body areas in image data, and the HRNet detection network is used to identify preset key points in the human body area.

[0025] Preferably, in this embodiment, the number of preset key points can be set to 16. These 16 preset key points can cover key parts of the human body. Through multiple preset key points, it is possible to distinguish between the intrusion of the entire human body and the intrusion of a partial limb (such as a hand or head), which greatly reduces false alarms. The preset key points are set as follows: (1) Head: top of the head, nose, chin; (2) Torso: neck, left shoulder, right shoulder, left hip, right hip; (3) Upper limbs: left elbow, right elbow, left wrist, right wrist; (4) Lower limbs: left knee, right knee, left ankle, right ankle.

[0026] Furthermore, the AI analysis module can realize human posture estimation, intrusion behavior determination and risk level classification based on lightweight AI algorithms. This embodiment adopts a two-stage detection network of YOLOv7+HRNet. The YOLOv7 detection network is responsible for rough screening of the human body, and the HRNet detection network is responsible for accurately locating 16 preset key points. At the same time, it can eliminate false detections caused by equipment movement based on the algorithm. Infrared thermal imaging detects abnormal heat sources (human body temperature range of 30~42℃), triggering visible light focus capture. When the preset key points are detected to intrude into the restricted area for three consecutive frames and the motion trajectory meets the "active intrusion mode", the PLC is triggered to shut down.

[0027] See also Figure 1In the area intrusion detection system of this embodiment, a PLC control module is also included, which is communicated with the AI analysis module and is used to execute an emergency stop operation of the production equipment when the AI analysis module determines that an intrusion has occurred in the detection area. The AI analysis module is communicated with the production equipment and is used to execute an emergency stop operation of the production equipment when the AI analysis module determines that an intrusion has occurred in the detection area.

[0028] It is understandable that the PLC control module supports PLC-related protocols and can be seamlessly connected with Siemens, Mitsubishi PLCs, etc. It is compatible with hard-wired signal (24V DC) emergency stop circuits. The PLC receives system status in real time. If there is no response within the timeout, it will be forced to shut down and record the fault log.

[0029] See also Figure 1 In the regional intrusion detection system of this embodiment, a management platform module is also included, which is communicatively connected to the video acquisition module, the AI analysis module, and the PLC control module. The management platform module is used to provide a visual configuration interface for the dynamic detection area.

[0030] Furthermore, the management platform module provides a polygonal drawing interface, supports dynamic restricted areas (such as following the movement of the device), and can perform multiple judgments. For example, intrusion into the dynamic restricted area is detected in 3 frames out of 5 consecutive frames, thereby achieving the technical effect of reducing false alarms. Users can delineate multiple irregular polygonal restricted areas in the visual configuration interface of the management platform module. The management platform module has multiple error-proof judgment functions. In addition, multiple cameras can be set up at the same time for all-weather detection, and support switching monitoring and display of any camera, and provide alarm logs and alarm capture functions. The alarm log can automatically associate video clips, shortening the accident tracing time from hours to minutes, and supports dynamic adjustment of electronic fences to adapt to changes in production line layout, thereby reducing downtime and transformation time.

[0031] In some specific embodiments, the regional intrusion detection system may further include an alarm module, which is in communication with the AI analysis module. The alarm module is configured to execute an alarm operation when the AI analysis module determines that an intrusion has occurred in the detection area.

[0032] It can be understood that through the combination of the above-mentioned video acquisition module, AI analysis module, PLC control module, and management platform module, the present invention realizes accurate identification of local human intrusion behavior, millisecond-level alarm and emergency shutdown of equipment, and realizes millisecond-level response of the entire process of "detection-alarm-shutdown", further improving the safety of dangerous working areas, and solving the problems of high cost, high false alarm rate, and large response delay of traditional security solutions. At the same time, by optimizing the AI algorithm, accurately identifying the preset key points of the human body, and performing real-time alarm and PLC control shutdown when human intrusion is detected, efficient monitoring and safety protection of dangerous areas are realized. It has high real-time performance, high accuracy and easy operation, and can be widely used in the field of industrial safety.

[0033] It should be noted here that this embodiment provides an area intrusion detection system, including a video acquisition module, an AI analysis module, a PLC control module and a management platform module. The AI analysis module can accurately identify human intrusion behavior in dangerous working areas in the factory. When intrusion behavior is detected, the PLC control module responds quickly and synchronously controls the corresponding production equipment to stop running, greatly reducing the incidence of safety accidents in the factory. At the same time, by identifying the preset key points of the human area, it can accurately identify human intrusion behavior, avoid false alarms and unnecessary shutdown operations, and achieve further improvement in production efficiency. Compared with manual inspections, it saves manpower and greatly improves the intelligence, usability and reliability of the present invention.

[0034] See also Figure 2 This embodiment provides a method for detecting an area intrusion, the method comprising: S100, real-time acquisition of current image data of the detection area; Specifically, the method for collecting current image data of the detection area includes: The dual-spectral camera group synchronously collects the current visible light image, infrared thermal imaging and depth perception data of the detection area, and uses a temperature compensation algorithm to eliminate environmental thermal noise.

[0035] It should be further explained that by synchronously collecting visible light images, infrared thermal imaging and depth perception data, multi-dimensional information of the detection area can be obtained. This multimodal data fusion method significantly improves the accuracy and reliability of detection.

[0036] Furthermore, visible light images can provide high-resolution visual information, helping to identify detailed features of objects. Infrared thermal imaging can capture the thermal radiation of objects, effectively operating even in low-light or complete darkness. It can also distinguish objects of different temperatures, and thus determine whether a person is intruding. Depth perception data can provide three-dimensional spatial information, helping to accurately measure the distance and position of objects. By simultaneously utilizing visible light and infrared thermal imaging technologies, the present invention can maintain efficient operation under various lighting conditions (including daytime, nighttime, and inclement weather). By employing a temperature compensation algorithm to compensate the collected infrared thermal imaging data, errors caused by ambient temperature fluctuations can be corrected, ensuring the authenticity and consistency of the thermal imaging data. This effectively eliminates the impact of environmental thermal noise on infrared thermal imaging, ensuring stable thermal imaging data even in complex ambient temperature fluctuations. The comprehensive analysis of multimodal data enables more accurate distinction between actual intrusions and environmental interference (such as animal activity, wind-blown leaves, etc.), significantly reducing false alarm rates.

[0037] It should be further explained that the temperature compensation algorithm in this embodiment is based on identifying real-time temperature and comparing it with infrared thermal imaging data. It uses polynomial fitting (such as a quadratic curve) or a neural network to solve complex temperature distribution problems. It implements segmented compensation by collecting data from multiple temperature points. In combination with a real-time temperature sensor to monitor the ambient temperature, it dynamically updates the compensation parameters to account for thermal radiation attenuation errors that may occur due to distance changes. The specific implementation process is as follows: (1) Data acquisition and preprocessing: The original output data of the digital temperature sensor at different temperatures (e.g., -20°C to 60°C) are collected in a constant temperature box. Calibration points are set at intervals of 5°C to remove abnormal values (e.g., sensor transient noise). The sliding standard deviation method (with a window width of N = 50) can be used. That is, when three consecutive sampling points exceed the ±3σ range, they are judged as transient noise, and the sliding average filter is applied to smooth the data. (2) Feature extraction and modeling: Integrate thermistors or digital temperature sensors to build compensation circuits. NTC thermistors (MF52-103F3435) can be used to build temperature compensation networks. Real-time ambient temperature is collected, and a temperature-compensation coefficient table is constructed based on calibration data. Adjacent temperature points are queried in real time, and interpolation calculations are performed. (3) Real-time temperature compensation and optimization: Automatically adjust the human body recognition threshold according to the ambient temperature (such as increasing sensitivity in low-temperature environments), divide the thermal imaging screen into multiple areas, and independently calculate the compensation coefficient for each area (such as the center area has a higher weight). Based on the target and distance information, dynamically compensate for the thermal radiation attenuation error caused by distance changes.

[0038] S200, performing a primary screening of a human body region according to the image data, identifying preset key points of the human body region, and generating a determination result of an intrusion behavior according to the identified preset key points; S300: When it is determined that an intrusion occurs in the detection area, an instruction to shut down the production equipment is output, thereby controlling the production equipment to shut down.

[0039] Specifically, the method for generating the determination result of the intrusion behavior based on the identified preset key points is as follows: First, the YOLOv7 detection model is used to perform a preliminary screening of the human body area in the collected image data, generate a target detection frame, and determine whether the intersection-over-union ratio of the target detection frame and the preset activity restricted area is greater than or equal to a first preset threshold, and the confidence level is greater than or equal to a second preset threshold. If so, the preset key points of the human body area are detected; If not, continue to monitor the current image data of the detection area.

[0040] It should be noted that the target detection box refers to the object boundary predicted by the YOLOv7 detection model, including the object's location (coordinates) and size (width and height). The preset restricted area refers to a closed polygonal area pre-defined by the user on the visual configuration interface of the management platform module (such as the no-entry zone or no-parking zone in monitoring). This area can be described by a sequence of coordinate points. The intersection over union (IoU) between the target detection box and the preset restricted area is used to measure the degree of overlap between the two areas. The specific formula is: ; Preferably, in this embodiment, the first preset threshold is set to 0.5. IoU ≥ 0.5 indicates that the overlapping area of the target detection frame and the preset activity restricted area accounts for at least 50% of the total combined area of the two. At this time, the object has significantly invaded the preset activity restricted area, such as half of the body entering the restricted area.

[0041] Furthermore, the confidence level indicates the YOLOv7 detection model's degree of confidence in the category and location of objects within the target detection frame. Preferably, in this embodiment, the second preset threshold is set to 0.9. A confidence level ≥ 0.9 indicates that the YOLOv7 detection model believes with a probability of more than 90% that the object within the target detection frame is real and the category is correctly identified, thereby filtering out low-quality detection results (such as false detections and blurred objects) and retaining only high-confidence alerts. A judgment is triggered only when both high confidence level (≥ 0.9) and significant overlap (IoU ≥ 0.5) are met, thus ensuring reliable detection results.

[0042] See also Figure 2 In the area intrusion detection method of this embodiment, the method for detecting preset key points in the human body area includes: The HRNet detection model is used to locate the coordinates of the preset key points in the human body area, the number of frames in which each preset key point continuously intrudes into the preset activity restricted area is recorded, and the motion vector of each preset key point is calculated; When the number of frames in which any preset key point continuously intrudes into the preset restricted area reaches a third preset threshold, and the direction of the motion vector points to the production equipment, it is determined that an intrusion has occurred in the current detection area, and a command to shut down the production equipment is synchronously output to control the shutdown of the production equipment; Otherwise, it is determined that no intrusion has occurred in the current detection area, that is, the production equipment is operating normally.

[0043] Preferably, this embodiment sets the third preset threshold to 3, that is, when any preset key point continuously invades the preset active restricted area for 3 frames and the direction of the motion vector points to the production equipment, it is determined that an intrusion has occurred in the current detection area, and the instruction to shut down the production equipment is output synchronously to control the shutdown of the production equipment. This method not only improves the detection accuracy, but also realizes efficient automatic control to ensure production safety.

[0044] Furthermore, HRNet is an advanced deep learning model that can perform multi-level feature fusion while maintaining high-resolution feature maps, thereby improving the positioning accuracy of preset key points on the human body.

[0045] HRNet more accurately detects and tracks the positional changes of preset key points on the human body, operating effectively even in complex environments. By recording the number of frames in which each preset key point continuously intrudes into a pre-set restricted area, it can quickly determine whether a potential intrusion has occurred, further reducing the possibility of single-frame misjudgments and improving system reliability. Calculating the motion vectors of each preset key point dynamically analyzes the direction and speed of movement, further confirming actual intrusion. When the motion vector points toward production equipment, it indicates a potential security threat, enabling the system to respond quickly. This automated control mechanism prevents dangerous situations from occurring immediately, significantly reducing the risk of safety incidents. By combining continuous frame monitoring and motion vector analysis, the system can more accurately identify true intrusions, avoiding false alarms caused by environmental interference or accidental factors. A shutdown command is triggered only when a preset key point continuously intrudes and the motion vector points toward production equipment, ensuring the efficiency and accuracy of the detection system.

[0046] See also Figure 2 In the area intrusion detection method of this embodiment, when it is determined that an intrusion occurs in the current detection area, an alarm prompt is issued, and an intrusion prompt box pops up through the visual configuration interface of the management platform module. The content of the intrusion prompt box includes the coordinates of the preset key points that intrude into the preset activity restricted area, the intrusion time, and the human body posture.

[0047] As you can understand, when an intrusion is detected, the system immediately issues an alarm, notifying relevant personnel to take prompt action. This instant feedback mechanism significantly improves the speed of responding to emergencies and reduces potential security risks. The intrusion alert box includes detailed information such as the coordinates of the preset key points that intruded into the preset restricted area, the time of intrusion, and the body's posture, providing comprehensive data support for managers. The management platform module provides an intuitive visual configuration interface, allowing managers to view and handle intrusion events. Through graphical display, users can more intuitively understand the on-site situation, facilitating subsequent operations and adjustments. Alarms can be communicated to relevant personnel through various means, such as sound and light alarms, text message notifications, or application push messages, making the regional intrusion detection system more intelligent and reliable, suitable for various industrial scenarios.

[0048] See also Figure 3 The following is a specific example of an AI-based area intrusion alarm and detection system to illustrate the implementation steps of the present invention: Step 1: Hardware configuration and parameter setting in the video acquisition module; Infrared dual-spectrum camera model: DS-2TD2628T-3 / QA (Hikvision); 4-megapixel visible light sensor (1 / 2.8" CMOS), thermal imaging resolution 384×288, GPU computing equipment: 4 x NVIDIA Ampere A40, connected via PCIe Gen4 x16 slots, NVLink enabled for 4-card interconnection, memory bandwidth increased to 112GB / s, support for direct access to Hikvision cameras, video streaming latency ≤ 30ms, the primary link uses the Modbus TCP protocol, and the backup link is a 24V hard-wired emergency stop signal. Step 2: Algorithm deployment in the AI analysis module; (1) Data collection and annotation Data collection: Collect images and videos of various parts of the human body, covering different scenes (such as factories and warehouses), different lighting conditions, and different human postures (standing, bending, squatting, etc.); Data annotation: Use annotation tools to annotate 16 key points of the human body. The annotation data includes the coordinates of key points such as the head, neck, shoulder, elbow, wrist, hip, knee, and ankle. (2) Model selection Model selection: Select human posture estimation models such as YOLOv7+HRNet as the basic model; Level 1 verification: The IoU between the YOLOv7 target detection box and the preset restricted area polygon is ≥ 0.5, and the confidence level is ≥ 0.9; Level 2 verification: The wrist key points located by HRNet enter the preset restricted area for three consecutive frames, and the motion vector points toward the device (speed ≥ 0.5 m / s); (3) Model training and deployment Use GPU servers (such as NVIDIA Ampere A40) for training; Segment the background based on the ViBe++ algorithm and insert samples of floating objects (such as cloth and plastic film) to improve anti-interference capabilities; Model integration: Integrate the trained model into the regional intrusion detection system to process video stream data in real time.

[0049] Step 3: Identify preset key points of the human body (1) Preset key points: Head: top of the head, nose, chin; Torso: neck, left shoulder, right shoulder, left hip, right hip; Upper limbs: left elbow, right elbow, left wrist, right wrist; Lower limbs: left knee, right knee, left ankle, right ankle; (2) Identification steps Obtain real-time video streams from surveillance cameras, extract each frame of the image, use the optimized model to detect human body areas in the image, input the detected human body areas into the optimized AI model, and the model outputs the coordinates (x, y) of 16 preset key points. Based on the preset key point coordinates, the model calculates the human body posture (such as standing, bending over, squatting), determines whether the human body enters the preset activity restricted area, and passes the preset key point coordinates and posture information to the alarm module.

[0050] Step 4: Trigger the alarm module to alarm; Users can manage the platform module, namely Figure 3 Preset activity restricted areas (such as robot operation areas) are delineated in the monitoring screen of the management platform. When any preset key point of the human body is detected entering the restricted area, an alarm is triggered. An audible and visual alarm is installed near the dangerous area. When an intrusion is detected, the alarm emits a high-decibel alarm and flashes lights. An intrusion prompt box pops up on the screen of the management platform module, displaying information such as the intrusion location, time, and human posture. At the same time, real-time video images are played for security personnel to view. The system automatically saves video clips of the intrusion event and key point detection results for subsequent analysis.

[0051] Step 5: PLC control module controls shutdown; When the AI analysis module determines that an intrusion has occurred, it will simultaneously trigger an alarm and send a shutdown signal to the PLC. After receiving the shutdown signal, the PLC will immediately execute the preset shutdown program, cut off the power supply to the equipment in the dangerous area or send an emergency stop command. The PLC will feedback the equipment shutdown status to the management platform module to ensure that the detection system receives the instruction that the shutdown operation has been completed.

[0052] It should be noted here that the present embodiment provides a method for regional intrusion detection, which collects the current image data of the detection area in real time; performs primary screening of the human body area based on the image data, identifies preset key points of the human body area, and generates a judgment result of the intrusion behavior based on the identified preset key points; when it is determined that an intrusion behavior has occurred in the detection area, an instruction to shut down the production equipment is output, thereby controlling the shutdown of the production equipment. The present invention can accurately identify human intrusion behavior in dangerous working areas in the factory. When an intrusion behavior is detected, the PLC control module responds quickly and synchronously controls the corresponding production equipment to stop running, greatly reducing the incidence of safety accidents in the factory. At the same time, by identifying the preset key points of the human body area, it can accurately identify human intrusion behavior, avoid false alarms and unnecessary shutdown operations, and achieve further improvement in production efficiency. Compared with manual inspections, it saves manpower and greatly improves the intelligence, usability and reliability of the present invention.

[0053] In a preferred embodiment, the present application further provides an electronic device, comprising: A memory; and a processor, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the area intrusion detection method is implemented. The computer device can be broadly defined as a server, a terminal, or any other electronic device with the necessary computing and / or processing capabilities. In one embodiment, the computer device may include a processor, a memory, a network interface, a communication interface, etc. connected via a system bus. The processor of the computer device can be used to provide the necessary computing, processing, and / or control capabilities. The memory of the computer device may include a non-volatile storage medium and an internal memory. An operating system, a computer program, etc. may be stored in or on the non-volatile storage medium. The internal memory can provide an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface and the communication interface of the computer device can be used to connect to and communicate with external devices via a network. When the computer program is executed by the processor, the steps of the method of the present invention are performed.

[0054] The present invention can be implemented as a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, causes the steps of the method of an embodiment of the present invention to be performed. In one embodiment, the computer program is distributed on a plurality of computer devices or processors coupled to a network so that the computer program is stored, accessed, and executed in a distributed manner by one or more computer devices or processors. A single method step / operation, or two or more method steps / operations, can be performed by a single computer device or processor or by two or more computer devices or processors. One or more method steps / operations can be performed by one or more computer devices or processors, and one or more other method steps / operations can be performed by one or more other computer devices or processors. One or more computer devices or processors can perform a single method step / operation, or perform two or more method steps / operations.

[0055] It should be noted here that the present invention provides an area intrusion detection system and method, including a video acquisition module, an AI analysis module, a PLC control module and a management platform module. The AI analysis module can accurately identify human intrusion behavior in dangerous working areas in the factory. When intrusion behavior is detected, the PLC control module responds quickly and synchronously controls the corresponding production equipment to stop running, greatly reducing the incidence of safety accidents in the factory. At the same time, by identifying preset key points in the human area, it can accurately identify human intrusion behavior, avoid false alarms and unnecessary shutdown operations, and achieve further improvement in production efficiency. Compared with manual inspections, it saves manpower and greatly improves the intelligence, usability and reliability of the present invention.

[0056] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in a sequence other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0057] The various technical features described above can be combined arbitrarily. Although not all possible combinations of these technical features are described, any combination of these technical features should be considered to be covered by this specification as long as such combination does not conflict.

[0058] The specific embodiments of the present invention described above do not limit the scope of protection of the present invention. Any other corresponding changes and modifications made based on the technical concept of the present invention should be included in the scope of protection of the claims of the present invention.

Claims

1. A regional intrusion detection system, characterized in that: include: Video acquisition module, used to collect the current image data of the detection area in real time; An AI analysis module, in communication with the video acquisition module, is configured to perform a primary screening of a human body region based on the image data, identify preset key points of the human body region, and generate a determination result of an intrusion behavior based on the identified preset key points; A PLC control module is in communication with the AI analysis module and is configured to execute an emergency stop operation of the production equipment when the AI analysis module determines that an intrusion has occurred in the detection area; The management platform module is communicated with the video acquisition module, AI analysis module, and PLC control module to provide a visual configuration interface for the dynamic detection area.

2. The regional intrusion detection system according to claim 1, characterized in that: The video acquisition module includes a dual-spectrum camera group for synchronously acquiring visible light images, infrared thermal imaging and depth perception data, and the video acquisition module includes a temperature compensation algorithm module for eliminating environmental thermal noise; The image data includes visible light images, infrared thermal images, and depth perception data.

3. The regional intrusion detection system according to claim 1, characterized in that: The AI analysis module includes a two-stage detection network constructed based on YOLOv7 and HRNet, the YOLOv7 detection network is used to perform primary screening of human body areas in the image data, and the HRNet detection network is used to identify preset key points of human body areas.

4. The regional intrusion detection system according to claim 3, characterized in that: The AI analysis module is in communication with the production equipment and is used to execute an emergency stop operation of the production equipment when the AI analysis module determines that an intrusion occurs in the detection area.

5. The regional intrusion detection system according to claim 3, characterized in that: The regional intrusion detection system also includes an alarm module, which is in communication with the AI analysis module. The alarm module is used to execute an alarm operation when the AI analysis module determines that an intrusion behavior occurs in the detection area.

6. A method for regional intrusion detection, characterized in that: include: S100, real-time acquisition of current image data of the detection area; S200, performing a primary screening of a human body region according to the image data, identifying preset key points of the human body region, and generating a determination result of an intrusion behavior according to the identified preset key points; S300: When it is determined that an intrusion occurs in the detection area, an instruction to shut down the production equipment is output, thereby controlling the production equipment to shut down.

7. The area intrusion detection method according to claim 6, characterized in that: The method for real-time acquisition of current image data of the detection area includes: The dual-spectral camera group synchronously collects the current visible light image, infrared thermal imaging and depth perception data of the detection area, and uses a temperature compensation algorithm to eliminate environmental thermal noise.

8. The area intrusion detection method according to claim 7, characterized in that: The S200 includes: Use the YOLOv7 detection model to perform primary screening of the human body area in the image data, generate a target detection frame, and determine whether the intersection-over-union ratio of the target detection frame and a preset activity restricted area is greater than or equal to a first preset threshold, and the confidence level is greater than or equal to a second preset threshold. If so, detect preset key points in the human body area; If not, continue to monitor the current image data of the detection area.

9. The area intrusion detection method according to claim 8, characterized in that: The method for detecting preset key points in a human body region includes: The HRNet detection model is used to locate the coordinates of the preset key points in the human body area, the number of frames in which each preset key point continuously intrudes into the preset activity restricted area is recorded, and the motion vector of each preset key point is calculated; When the number of frames in which any preset key point continuously intrudes into the preset restricted area reaches a third preset threshold, and the direction of the motion vector points to the production equipment, it is determined that an intrusion has occurred in the current detection area, and a command to shut down the production equipment is synchronously output to control the shutdown of the production equipment; Otherwise, it is determined that no intrusion has occurred in the current detection area, that is, the production equipment is operating normally.

10. The area intrusion detection method according to claim 9, characterized in that: The method further comprises: When it is determined that an intrusion occurs in the current detection area, an alarm prompt is issued, and an intrusion prompt box pops up through the visual configuration interface of the management platform module. The content of the intrusion prompt box includes the coordinates of the preset key points that intrude into the preset activity restricted area, the intrusion time, and the human body posture.

Images