A barrier gate anti-smashing system and method based on computer vision detection technology

Abstract

The present invention discloses a barrier gate anti-smashing system and method based on computer vision detection technology, the system comprises a visual detection camera, a visual detection processing module, a visual detection control module, a barrier gate controller and a barrier gate body, the visual detection camera is connected to the visual detection processing module via Ethernet; the visual detection processing module is connected to the visual detection control module via Ethernet or a serial port; the visual detection control module is connected to the barrier gate controller via Ethernet or a serial port or an IO, and the barrier gate controller is connected to the barrier gate body via a serial port or an IO. The present invention has high anti-smashing response efficiency, low misidentification rate, is not easily interfered with, actively prevents smashing, and can automatically drop the bar.

Description

Technical Field

[0001] The invention relates to a barrier gate anti-smashing system, in particular to a barrier gate anti-smashing system and a method thereof based on computer vision detection technology. Background Art

[0002] At present, the main anti-smashing technologies for non-motorized vehicle lane barrier machines (hereinafter referred to as "barriers") include: 1. Ground sensing anti-smashing; 2. Pressure wave anti-smashing; 3. Anti-smashing radar anti-smashing and other methods, which have achieved anti-smashing effects to a certain extent, but there are still the following pain points: 1. Low anti-smashing response efficiency; 2. Inaccurate pedestrian detection and high misidentification rate; 3. Susceptible to interference; 4. Passive anti-smashing, not active anti-smashing; 5. Only one pedestrian passes, and the barrier cannot be automatically lowered.

[0003] CN114742885B discloses a method for determining target consistency in a binocular vision system, comprising the following steps: S1: collecting left and right disparity images of a target simultaneously through a binocular vision system; S2: obtaining left and right disparity images containing the target frame through an artificial intelligence target detection model; S3: determining whether the target posture is consistent through the left and right disparity images containing the target frame, if consistent, proceeding to the next step, if inconsistent, determining that the targets in the left and right disparity images are not the same target; S4: determining whether the spatial positions of the targets are consistent, if consistent, determining that the targets in the left and right disparity images are the same target, if inconsistent, determining that they are not the same target. Summary of the invention

[0004] In view of the problems existing in the prior art, the present invention provides a barrier gate anti-smashing system and method based on computer vision detection technology, which has high anti-smashing response efficiency, low misrecognition rate, is not easily interfered with, actively prevents smashing, and can automatically drop the rod.

[0005] The purpose of the present invention is achieved through the following technical solutions.

[0006] A barrier gate anti-smashing system based on computer vision detection technology, comprising a vision detection camera, a vision detection processing module, a vision detection control module, a barrier gate controller and a barrier gate body, wherein the vision detection camera is used to detect people or objects within a visual range; the vision detection camera is connected to the vision detection processing module via Ethernet, and the vision detection processing module is used to process the video stream image transmitted back by the vision detection camera and convert it into image coordinates and control instructions; the vision detection processing module is connected to the vision detection control module via Ethernet or a serial port, and the vision detection control module is used to receive the image or coordinate instructions of the vision detection processing module and convert them into instructions for controlling the barrier gate controller; the vision detection control module is connected to the barrier gate controller via Ethernet or a serial port or an IO, and the barrier gate controller is used to receive the instructions of the vision detection control module and control the action of the barrier gate; the barrier gate controller is connected to the barrier gate body via a serial port or an IO, and the barrier gate body is used to execute the action sent by the barrier gate controller.

[0007] The visual detection camera is a binocular camera.

[0008] A computer vision detection area is provided, which includes an import exit area, an import safety area, an exit safety area, and an exit exit area. The import safety area and the exit safety area are respectively located on the front and rear sides of the railing controlled by the gate body.

[0009] When a pedestrian enters the import / exit safety area from the import / exit exit area, the pedestrian safety mechanism is activated, and the visual detection control module will control the gate controller to prohibit the gate from closing; when a pedestrian enters the import / exit exit area from the import / exit safety area, the pedestrian safety mechanism is closed, and the visual detection control module will allow the gate controller to send a gate closing command; if a pedestrian breaks into the import / exit safety area during the gate closing process, the pedestrian safety mechanism will be activated urgently, and the visual detection control module will control the gate controller to emergency stop or emergency lift the bar.

[0010] The visual detection processing module receives the deep network head detection list, body frame detection list and binocular disparity image of each frame image, and detects the target in three paths:

[0011] The first path: for each frame of the deep network head detection list, filter out the wrong results through the disparity foreground ratio and Haar features, and then extract its position in the three-dimensional coordinates;

[0012] The second path: for each frame of the deep network body box detection list, directly extract its position in 3D coordinates;

[0013] The third path: For binocular disparity images, the foreground image is extracted through the foreground extraction module, and the body frame target of the deep network is used as a mask to remove the duplicate foreground. The connected domain is extracted from the remaining image, and the connected domain is filtered. The connected domain that meets the conditions is selected through its shape and three-dimensional coordinate height as the detection target;

[0014] Finally, the targets obtained from the three paths are merged, and their feature information is uniformly extracted to obtain the final pedestrian detection result.

[0015] A method for preventing gates from being smashed based on computer vision detection technology, comprising the following steps:

[0016] 1) Pedestrians drive non-motor vehicles or walk into the computer vision detection area;

[0017] 2) The system sends safety instructions and prohibition of gate drop instructions to the gate controller. At this time, the gate controller cannot control the gate to drop;

[0018] 3) Pedestrians authorize the gate controller to control the gate lift bar through ID card, IC card, QR code, or face recognition;

[0019] 4) The gate executes the gate controller's lifting command;

[0020] 5) The computer vision detection camera monitors and determines whether the pedestrian has safely left the computer vision detection safety area;

[0021] 6) If the pedestrian does not leave the computer vision detection safety area, the barrier will remain in the raised state;

[0022] 7) If the pedestrian has safely left the computer vision detection safety area, the system will send a command to the gate controller to close the safety command and send a bar drop command, and the gate controller will control the gate to drop the bar; if a pedestrian breaks into the computer vision detection safety area at this time, the system will control the gate controller to emergency stop the bar or emergency lift the bar.

[0023] Compared with the prior art, the advantages of the present invention are: computer vision technology is based on artificial intelligence (AI) algorithms, and can accurately identify and locate pedestrians through pedestrian profile analysis and three-dimensional size measurement, and can accurately identify whether the detected target is a person or a non-motor vehicle through passenger profile analysis and three-dimensional size measurement and training. Therefore, it can accurately and reliably detect and judge whether pedestrians are safe within its field of view. And through Ethernet, serial port, IO, etc., the situation in the field of view, safety instructions, bar drop instructions, etc. are accurately conveyed to the controller in the system, thereby ensuring the safety of pedestrians and alleviating the pain points of existing non-motor vehicle (barrier) anti-smashing technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 Schematic diagram of the system architecture of the present invention.

[0025] Figure 2 Schematic diagram of the computer vision detection area of ​​the present invention.

[0026] Figure 3 This is a schematic diagram of the pedestrian detection principle of the present invention.

[0027] Figure 4 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0028] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0029] like Figure 1 As shown, a barrier gate anti-smashing system based on computer vision detection technology includes a visual detection camera, a visual detection processing module, a visual detection control module, a barrier gate controller and a barrier gate body, wherein the visual detection camera is used to detect people or objects within the visual range; the visual detection camera is connected to the visual detection processing module via Ethernet, and the visual detection processing module is used to process the video stream image transmitted back by the visual detection camera, and convert it into image coordinates and control instructions; the visual detection processing module is connected to the visual detection control module via Ethernet or a serial port, and the visual detection control module is used to receive the image or coordinate instructions of the visual detection processing module, and convert it into instructions for controlling the barrier gate controller; the visual detection control module is connected to the barrier gate controller via Ethernet or a serial port or IO, and the barrier gate controller is used to receive the visual detection control module instructions and control the barrier gate action; the barrier gate controller is connected to the barrier gate body via a serial port or IO, and the barrier gate body is used to execute the action sent by the barrier gate controller.

[0030] The visual detection camera is a binocular camera.

[0031] like Figure 2 As shown, a computer vision detection area is set up, and the computer vision detection area includes an import exit area, an import safety area, an exit safety area, and an exit exit area. The import safety area and the exit safety area are respectively located on the front and rear sides of the railing controlled by the gate body.

[0032] When a pedestrian enters the import / exit safety area from the import / exit exit area, the pedestrian safety mechanism is activated, and the visual detection control module will control the gate controller to prohibit the gate from closing; when a pedestrian enters the import / exit exit area from the import / exit safety area, the pedestrian safety mechanism is closed, and the visual detection control module will allow the gate controller to send a gate closing command; if a pedestrian breaks into the import / exit safety area during the gate closing process, the pedestrian safety mechanism will be activated urgently, and the visual detection control module will control the gate controller to emergency stop the bar or emergency lift the bar (determined by the gate function).

[0033] like Figure 3 As shown, the visual detection processing module receives the deep network head detection list, body frame detection list and binocular disparity image of each frame image, and detects the target in three paths:

[0034] The first path: for each frame of the deep network head detection list, filter out the wrong results through the disparity foreground ratio and Haar features, and then extract its position in the three-dimensional coordinates;

[0035] The second path: for each frame of the deep network body box detection list, directly extract its position in 3D coordinates;

[0036] The third path: For binocular disparity images, the foreground image is extracted through the foreground extraction module, and the body frame target of the deep network is used as a mask to remove the duplicate foreground. The connected domain is extracted from the remaining image, and the connected domain is filtered. The connected domain that meets the conditions is selected through its shape and three-dimensional coordinate height as the detection target;

[0037] Finally, the targets obtained from the three paths are merged, and their feature information is uniformly extracted to obtain the final pedestrian detection result.

[0038] like Figure 4 As shown, a method for preventing gates from being smashed based on computer vision detection technology comprises the following steps:

[0039] 1) Pedestrians drive non-motor vehicles or walk into the computer vision detection area;

[0040] 2) The system sends safety instructions and prohibition of gate drop instructions to the gate controller. At this time, the gate controller cannot control the gate to drop;

[0041] 3) Pedestrians authorize the gate controller to control the gate lift bar through ID card, IC card, QR code, or face recognition;

[0042] 4) The gate executes the gate controller's lifting command;

[0043] 5) The computer vision detection camera monitors and determines whether the pedestrian has safely left the computer vision detection safety area;

[0044] 6) If the pedestrian does not leave the computer vision detection safety area, the barrier will remain in the raised state;

[0045] 7) If the pedestrian has safely left the computer vision detection safety area, the system will send a command to the gate controller to close the safety command and send a bar drop command, and the gate controller will control the gate to drop the bar; if a pedestrian breaks into the computer vision detection safety area at this time, the system will control the gate controller to emergency stop the bar or emergency lift the bar (determined by the gate function).

Claims

1. A barrier gate anti-smashing system based on computer vision detection technology, characterized in that It includes a visual detection camera, a visual detection processing module, a visual detection control module, a gate controller and a gate body, wherein the visual detection camera is used to detect people or objects within the visual range; the visual detection camera is connected to the visual detection processing module via Ethernet, and the visual detection processing module is used to process the video stream image sent back by the visual detection camera and convert it into image coordinates and control instructions; the visual detection processing module is connected to the visual detection control module via Ethernet or a serial port, and the visual detection control module is used to receive the image or coordinate instructions of the visual detection processing module and convert them into instructions for controlling the gate controller; the visual detection control module is connected to the gate controller via Ethernet or a serial port or IO, and the gate controller is used to receive the instructions of the visual detection control module and control the gate action; the gate controller is connected to the gate body via a serial port or IO, and the gate body is used to execute the action sent by the gate controller; The visual detection camera is a binocular camera; A computer vision detection area is provided, and the computer vision detection area includes an entrance exit area, an entrance safety area, an exit safety area, and an exit exit area. The entrance safety area and the exit safety area are respectively located at the front and rear sides of the railing controlled by the gate body; When a pedestrian enters the entrance / exit safety area from the entrance / exit exit area, the pedestrian safety mechanism is activated, and the visual detection control module will control the gate controller to prohibit the gate from closing; when a pedestrian enters the entrance / exit exit area from the entrance / exit safety area, the pedestrian safety mechanism is closed, and the visual detection control module will allow the gate controller to send a gate closing command; if a pedestrian breaks into the entrance / exit safety area during the gate closing process, the pedestrian safety mechanism will be activated urgently, and the visual detection control module will control the gate controller to emergency stop the bar or emergency lift the bar; The visual detection processing module receives the deep network head detection list, body frame detection list and binocular disparity image of each frame image, and detects the target in three paths: The first path: for each frame of the deep network head detection list, filter out the wrong results through the disparity foreground ratio and Haar features, and then extract its position in the three-dimensional coordinates; The second path: for each frame of the deep network body box detection list, directly extract its position in 3D coordinates; The third path: For binocular disparity images, the foreground image is extracted through the foreground extraction module, and the body frame target of the deep network is used as a mask to remove the duplicate foreground. The connected domain is extracted from the remaining image, and the connected domain is filtered. The connected domain that meets the conditions is selected through its shape and three-dimensional coordinate height as the detection target; Finally, the targets obtained from the three paths are merged, and their feature information is uniformly extracted to obtain the final pedestrian detection result.

2. A barrier gate anti-smashing method based on computer vision detection technology, characterized in that The barrier gate anti-smashing system based on computer vision detection technology as described in claim 1 comprises the following steps: 1) Pedestrians drive non-motor vehicles or walk into the computer vision detection area; 2) The system sends safety instructions and prohibition of barrier drop instructions to the barrier controller. At this time, the barrier controller cannot control the barrier drop; 3) Pedestrians authorize the gate controller to control the gate lift bar through ID card, IC card, QR code, or face recognition; 4) The gate executes the gate controller's lifting command; 5) The computer vision detection camera monitors and determines whether the pedestrian has safely left the computer vision detection safety area; 6) If the pedestrian does not leave the computer vision detection safety area, the barrier will remain in the raised state; 7) If the pedestrian has safely left the computer vision detection safety area, the system will send a command to the gate controller to close the safety command and send a bar drop command, and the gate controller will control the gate to drop the bar; if a pedestrian breaks into the computer vision detection safety area at this time, the system will control the gate controller to emergency stop the bar or emergency lift the bar.

Images