A recognition control system and method based on a motorized curtain
Patent Information
- Application Number
- CN202611130097.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-07-28
- Publication Date
- 2026-10-09
AI Technical Summary
误触发率高：传感器仅能探测移动物体，无法区分“途经”与“通行”意图，行人从门帘侧面经过时易触发无效开帘，干扰性强且用户体验差；
本发明全程通过视觉识别自动控制门帘启闭，人员无需接触门帘即可通行，从根源上减少接触式交叉传播，适配医院等感控要求高的场所；可识别目标类别、数量与尺寸，动态计算匹配的开启幅度，既保证轮椅、病床等大尺寸目标的通行安全，又能针对单人通行采用小开度，最大化节能效果；采用量化YOLO+ByteTrack算法方案，适配嵌入式端有限算力，在保证跟踪连续性的同时实现高实时性，降低硬件成本。
Smart Images
Figure CN122883699A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of machine vision target tracking technology, specifically to a recognition control system and method based on an electric door curtain. Background Technology
[0002] Currently, most passageway curtains in public places such as hospitals, hotels, and commercial centers are non-automatic, requiring people to manually lift them to pass through. This has several drawbacks: First, in places with high hygiene requirements such as hospitals, hand contact with the curtain can easily lead to cross-infection of bacteria and viruses, making infection control difficult. Second, it is strenuous for people carrying heavy objects, pushing wheelchairs / beds, as well as the elderly, children, and disabled people to pass through, resulting in poor passage efficiency and accessibility.
[0003] The few existing industrial automatic door curtains mostly use passive infrared sensing or microwave radar sensing solutions, which also have obvious technical defects: High false trigger rate: The sensor can only detect moving objects and cannot distinguish between the intention of "passing by" and "passing through". When pedestrians pass by the side of the curtain, it is easy to trigger invalid curtain opening, which is highly interfering and results in a poor user experience. Serious energy waste: Frequent and ineffective opening and closing of door curtains in air-conditioned spaces can lead to a large amount of cold / warm air leakage, resulting in continuous energy loss; The passage strategy is limited: it cannot identify the type and size of the target to be passed, and can only perform a fixed opening of the curtain. It cannot meet the passage needs of large targets such as wheelchairs and hospital beds, nor can it achieve small opening and energy saving for small targets. Inefficient operation and maintenance management: Equipment operates independently, forming management silos, and does not support remote status monitoring and parameter adjustment. Operation and maintenance personnel need to be on-site for debugging, resulting in high labor costs. Poor scene adaptability: The recognition logic and curtain opening rules are fixed at the factory, and the management cannot customize the recognition target type according to the needs of the venue, making it difficult to adapt to the differentiated management requirements of diverse scenarios such as hospitals and shopping malls. Summary of the Invention
[0004] The present invention proposes an identification control system and method based on an electric door curtain, which can at least solve one of the technical problems in the background art.
[0005] To achieve the above objectives, the present invention adopts the following technical solution: A recognition and control system based on an electric door curtain includes: a visual recognition and tracking module, an embedded main control motherboard, a motor drive actuator, and a configuration management platform; The visual recognition and tracking module is electrically connected to the embedded main control motherboard. It is used to acquire real-time images of the channel, run a lightweight quantized target detection model trained with a private dataset, and combine it with the ByteTrack multi-target tracking algorithm to perform category recognition, size estimation and continuous motion trajectory tracking of passing targets entering the field of view, and output target recognition and tracking data to the embedded main control motherboard. The embedded main control motherboard is electrically connected to the motor drive actuator and has a built-in preset forced opening area and passage intention judgment logic. It is used to receive target recognition and tracking data. When a target enters the forced opening area, it directly triggers the curtain opening command. When the target is outside the forced opening area, the passage intention is determined based on continuous trajectory data. The curtain opening range is calculated according to the type, quantity and size of the passing target, and a PWM speed control signal with the corresponding duty cycle is output to the motor drive actuator to control the curtain to open at the set speed. After detecting that the target has completely passed through the curtain and left the forced opening area, a curtain closing command is output after a delay. At the same time, the target type, opening data and passage time are uploaded to the configuration management platform. The configuration management platform is connected to the embedded main control motherboard for remote monitoring of the curtain's operating status, issuing parameter configuration commands and firmware update packages, and supporting custom settings for target categories and access control strategies.
[0006] As a preferred embodiment of the identification and control system based on electric door curtain described in this invention, during the system power-on initialization phase, the embedded main control motherboard drives the motor to drive the actuator to complete the full stroke of the door curtain, collects and records the number of motor pulses corresponding to the fully open and fully closed positions of the door curtain, and establishes a mapping relationship between the door curtain opening degree and the motor running pulses for precise opening degree control.
[0007] As a preferred embodiment of the identification and control system based on electric door curtains described in this invention, the ByteTrack multi-target tracking algorithm employs a confidence leveling and two-stage data association mechanism. The detection boxes in each frame are divided into high-scoring detection boxes and low-scoring detection boxes according to the confidence threshold; Kalman filtering is used to predict the position of the existing trajectory in the next frame, and a matching cost matrix is constructed based on the IoU distance. In the first stage, the Hungarian algorithm is used to match high-scoring detection boxes with all trajectories to obtain the first matching set, unmatched high-scoring detection boxes, and unmatched trajectories. The second stage involves a secondary matching of low-scoring detection boxes with unmatched trajectories to supplement the tracking continuity of occluded targets, and finally updates the trajectory IDs and location information of all targets.
[0008] As a preferred embodiment of the identification control system based on electric door curtain described in this invention, the opening and closing speeds of the door curtain controlled by the motor drive actuator are independently adjustable, with an adjustment range of 0.1m / s to 1.0m / s; the delay time for closing the door curtain is configurable, with a default value of 3 seconds.
[0009] As a preferred embodiment of the recognition and control system based on electric door curtain described in this invention, the forced opening area is a preset rectangular area adjacent to the door curtain installation surface, used to cover the scenario where a target suddenly enters the passageway from the side, skipping the trajectory intent judgment step and directly triggering the curtain opening, thus eliminating the response delay of the target entering from the side.
[0010] As a preferred embodiment of the electric curtain-based identification control system of the present invention, the configuration management platform supports custom access strategies including: enabling or disabling identification permissions for any type of target such as pedestrians, wheelchairs, hospital beds, and trolleys, and setting differentiated curtain opening coefficients and opening / closing speed parameters for different types of targets.
[0011] A recognition and control method based on an electric door curtain includes the following steps: S1. After the embedded main control motherboard completes the gate width self-learning calibration, it drives the vision recognition and tracking module to start the target detection mode and loads the quantized YOLO detection model and ByteTrack tracking configuration. S2, the visual recognition and tracking module acquires channel images in real time, performs category recognition and size estimation on passing targets entering the field of view, and assigns a trajectory ID to each target through the ByteTrack algorithm to continuously track its motion trajectory; S3. The embedded main control motherboard determines the target position in real time. If the target enters the preset forced opening area, it immediately outputs the curtain opening command. If the target is outside the forced opening area, it determines whether the target has the intention to pass towards the curtain based on continuous multi-frame trajectory data. If it is determined that the target has the intention to pass, it outputs the curtain opening command. S4. The embedded main control motherboard calculates the matching curtain opening range based on the type, quantity, and size of the current target, and outputs the corresponding PWM signal to control the motor drive actuator to drive the curtain to open to the target opening at a set speed. S5, the visual recognition and tracking module continuously tracks the target. When it detects that all targets have completely passed through the curtain and left the forced opening area, it controls the curtain to close after a preset delay. At the same time, it uploads the passage record to the configuration management platform.
[0012] As a preferred embodiment of the identification and control method based on electric door curtains described in this invention, the tracking process of the ByteTrack algorithm in step S2 specifically includes: For each frame of image, all detection results are divided into a high-scoring detection set and a low-scoring detection set according to a confidence threshold. The existing trajectory states are predicted by Kalman filtering to obtain the predicted bounding boxes of each trajectory in the next frame. A cost matrix is constructed using the IoU distance between the predicted bounding box and the detection box as the matching cost, and the Hungarian algorithm is used to complete the matching of the high-resolution detection set and the full trajectory in the first stage. For trajectories that are not matched in the first stage, a second stage matching is performed with the low-scoring detection set to restore the tracking link of the occluded target and filter out background interference detection boxes. Update the status and ID information of all trajectories based on the two-stage matching results.
[0013] As a preferred embodiment of the identification and control method based on electric door curtain described in this invention, the calculation method of the door curtain opening range in step S4 is as follows: based on the baseline opening range for single-pedestrian passage, the opening range is superimposed according to the proportion of the horizontal size of the target and the number of targets, and the superimposed opening range does not exceed the maximum full opening range of the door curtain; different types of targets correspond to different baseline opening range coefficients.
[0014] As a preferred embodiment of the identification and control method based on electric door curtains described in this invention, the method further includes a remote configuration step: the configuration management platform sends configuration instructions to the embedded main control motherboard, and the parameters that can be modified include the range of the forced opening area, the opening and closing speed, the curtain closing delay time, the target category for enabling identification, and the opening coefficient corresponding to each type of target. The parameters take effect immediately after modification.
[0015] The beneficial effects of this invention are: This invention uses visual recognition to automatically control the opening and closing of a door curtain, allowing people to pass through without touching the curtain, thus reducing contact-based cross-infection at the source and making it suitable for places with high infection control requirements such as hospitals. It can identify the type, quantity, and size of targets and dynamically calculate the matching opening range, ensuring the safety of large targets such as wheelchairs and hospital beds while using a small opening for single-person passage to maximize energy saving. It adopts a quantized YOLO+ByteTrack algorithm scheme, which is adapted to the limited computing power of embedded devices, achieving high real-time performance while ensuring tracking continuity and reducing hardware costs. Attached Figure Description
[0016] Figure 1 This is an overall system block diagram of the identification control system and method based on electric door curtains according to the present invention.
[0017] Figure 2 This is a flowchart illustrating the steps of the identification control system and method based on an electric door curtain according to the present invention. Detailed Implementation
[0018] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.
[0019] like Figure 1 As shown, this embodiment of the identification and control system based on electric door curtain consists of two parts: field terminal equipment and cloud configuration management platform. The field terminal includes a visual recognition and tracking module, an embedded main control motherboard, a motor drive actuator and the door curtain body. The embedded main control motherboard integrates a 4G / 5G mobile network communication module to achieve bidirectional data interaction with the cloud configuration management platform.
[0020] The system modules are as follows: 1. Visual recognition and tracking module: Employing a high-definition wide-angle camera, the module is centrally mounted above the curtain, covering the passageway area in front of it. It integrates a lightweight AI inference engine, deploying a quantized YOLOv8n detection model trained on a private dataset, capable of recognizing various passing targets such as pedestrians, wheelchairs, hospital beds, and trolleys. The module also integrates the ByteTrack tracking algorithm, assigning a unique tracking ID to each target and continuously outputting the target's bounding box coordinates, category confidence, size information, and motion trajectory.
[0021] In this scenario, the target moves continuously within the channel and there is no need for re-identification after a long period of loss. Therefore, the ByteTrack tracking algorithm with lower computational load is selected. It does not require the extraction of appearance feature vectors, which greatly reduces the computing load on the embedded end and ensures that the detection and tracking frame rate is not less than 25fps, meeting the real-time requirements.
[0022] 2. Embedded main control motherboard: It adopts an ARM architecture embedded processor, integrating a motor control interface, a video input interface, and a network communication module. Its core functions include: Door width self-learning: After the first power-on or triggering of the calibration command, the main control board outputs a control signal to drive the motor to move the door curtain from the fully closed position to the fully open position, reads the total number of pulses of the motor encoder, and establishes a linear mapping table of "opening percentage - number of pulses"; in subsequent opening control, it converts the target opening degree into the corresponding number of pulses to achieve precise closed-loop control.
[0023] Region and Intent Determination: A forced opening area adjacent to the door curtain is preset within the field of view coordinate system, and the area range is configurable. When the target's center point enters this area, the curtain is directly triggered to open, handling scenarios where the target suddenly cuts in from the side, avoiding response lag; when the target is outside the forced opening area, it is checked whether the target enters the pre-triggered area / near the door area, and whether it moves downwards as a whole in the historical trajectory, whether the frame area increases, whether it continues to move towards the door in the most recent step, and whether it crosses into the pre-triggered area.
[0024] Opening width calculation and speed control: Based on the target category, quantity and size, the target opening width is calculated using the opening width formula and converted into the corresponding motor pulse quantity; using PID to adjust the PWM output, the curtain is controlled to accelerate, maintain constant speed and decelerate to reach the target opening width. The opening speed and closing speed can be configured independently, with an adjustment range of 0.1m / s~1.0m / s.
[0025] Curtain closing delay control: When the center point of all targets has crossed the curtain baseline and left the forced opening area, the delay timer is started. The default delay is 3 seconds before the curtain closing command is output. The delay duration can be customized within the range of 0 to 10 seconds.
[0026] Data interaction: After each passage is completed, the passage record is packaged and uploaded to the configuration management platform; at the same time, the parameter configuration and firmware update package issued by the platform are received in real time, and the changes take effect after local updates.
[0027] 3. Motor-driven actuator: It includes a brushless DC motor, a motor drive board, and a transmission mechanism. It receives PWM speed control signals and direction signals from the embedded main control motherboard and drives the door curtain to complete the opening and closing action through the transmission belt / transmission rod. The motor has a built-in encoder that feeds back the position pulses to the main control board to form a closed-loop control.
[0028] 4. Configuration Management Platform: Deployed on a cloud server, it provides web and mobile management interfaces, with functions including: online device status monitoring, access data statistics and query, remote parameter configuration (forced opening area, opening and closing speed, delay time, recognition category, opening coefficient), remote firmware upgrade, and multi-device group management. Administrators can flexibly adjust strategies according to site requirements; for example, in a hospital setting, bed and wheelchair recognition can be enabled, while in a shopping mall setting, cart recognition can be disabled.
[0029] The ByteTrack algorithm used in this invention ensures the continuity of target tracking under low computing power conditions through confidence level classification and two-stage matching. The specific implementation steps are as follows: (1) Confidence rating: For all detection boxes output by the YOLO model in the current frame, the confidence scores are ranked. Perform a binary search:
[0030] In this embodiment, the high confidence threshold Low confidence threshold Detection boxes with scores below the low confidence threshold are directly identified as noise and filtered out; low confidence detection boxes are retained for secondary matching to effectively restore the tracking link of occluded targets. (2) Kalman filter trajectory prediction: A uniform motion model is used to describe the inter-frame motion of the target, and each trajectory maintains an 8-dimensional state vector:
[0031] State prediction is achieved through the state transition matrix:
[0032] State transition matrix For an 8th-order policy, corresponding to a position + velocity model of uniform motion, the form is as follows when the inter-frame sampling interval is 1:
[0033] After prediction is completed, bounding box parameters are extracted from the prediction state vector to obtain the prediction box for each trajectory in the current frame.
[0034] (3) Matching cost and two-stage correlation: using IoU distance as the matching cost, the first... The detection box and the first The cost of predicting a trajectory bounding box is:
[0035] in, The area of the intersection of the two frames. The area of the union of the two frames; First-stage matching: Input high-resolution detection set With all trajectories The optimal assignment is solved using the Hungarian algorithm, resulting in a set of matching pairs. Unmatched detection Unmatched trajectory Among them, the high-scoring detection boxes that did not match successfully The Kalman filter state will be initialized as the new target to generate a new tracking trajectory.
[0036] Second-stage matching: Input low-score detection set Unmatched trajectory The matching set is obtained by matching again using the Hungarian algorithm. Low-scoring detection boxes that fail to match are considered noise and discarded. The number of frames lost due to unmatched trajectories accumulates, and the trajectory is deleted once the threshold is exceeded.
[0037] For a successfully matched trajectory and detection box pair, the detection box is used as an observation. The posterior state vector and covariance matrix of the trajectory are updated using Kalman gain to complete the trajectory correction.
[0038] All final matching pairs The Kalman filter posterior state is used to update the trajectory while keeping the trajectory ID unchanged, thus enabling continuous tracking.
[0039] The target opening width of the door curtain is calculated by summing the dimensions of all targets waiting to pass, and an upper limit and safety margin are set, as shown in the following formula:
[0040] In this embodiment, the safety margin The default setting is 0.2m; pedestrian / human target opening coefficient. 0, Wheelchairs Hospital beds Target horizontal width The value is calculated by the vision module based on the bounding box pixel width and the pixel-to-actual size ratio calibrated on-site. If the calculated value exceeds the total width of the curtain... When using the full opening width of the door.
[0041] The complete system operation process is as follows: When the device is powered on, the embedded main control motherboard performs initialization, completes gate width self-learning calibration, loads the detection model and tracking parameters, and the vision module enters the real-time acquisition state. When there is no target in the channel, the curtain remains closed, and the system continuously performs low-power detection. When a target enters the camera's field of view, the YOLO model outputs the detection result, the ByteTrack algorithm starts tracking and assigns an ID, and continuously outputs trajectory data; The embedded main control motherboard determines the target location in real time: if it enters the forced opening area, it immediately triggers the curtain to open; if it is outside the area, it determines the intention to pass, and triggers the curtain to open after the determination is successful. The main control board calculates the target opening and outputs a PWM signal to drive the motor, so that the curtain opens to the target width at a set speed. During the passage of the target, the vision module continuously tracks it. If a new target is added, the opening is recalculated and dynamically adjusted. Once all targets have completely passed through the curtain and left the forced-open area, a 3-second delay is initiated; if no new targets enter during the delay, the curtain is controlled to close at the set speed. Once this passage is completed, the main control board will upload the passage data to the configuration management platform, awaiting the next trigger.
[0042] like Figure 2As shown, a recognition and control method based on an electric door curtain is provided, including the following steps: S1. After the embedded main control motherboard completes the gate width self-learning calibration, it drives the vision recognition and tracking module to start the target detection mode and loads the quantized YOLO detection model and ByteTrack tracking configuration. S2, the visual recognition and tracking module acquires channel images in real time, performs category recognition and size estimation on passing targets entering the field of view, and assigns a trajectory ID to each target through the ByteTrack algorithm to continuously track its motion trajectory; S3. The embedded main control motherboard determines the target position in real time. If the target enters the preset forced opening area, it immediately outputs the curtain opening command. If the target is outside the forced opening area, it determines whether the target has the intention to pass towards the curtain based on continuous multi-frame trajectory data. If it is determined that the target has the intention to pass, it outputs the curtain opening command. S4. The embedded main control motherboard calculates the matching curtain opening range based on the type, quantity, and size of the current target, and outputs the corresponding PWM signal to control the motor drive actuator to drive the curtain to open to the target opening at a set speed. S5, the visual recognition and tracking module continuously tracks the target. When it detects that all targets have completely passed through the curtain and left the forced opening area, it controls the curtain to close after a preset delay. At the same time, it uploads the passage record to the configuration management platform.
[0043] The ByteTrack algorithm's tracking process in step S2 is specifically as follows: For each frame of image, all detection results are divided into a high-scoring detection set and a low-scoring detection set according to a confidence threshold. The existing trajectory states are predicted by Kalman filtering to obtain the predicted bounding boxes of each trajectory in the next frame. A cost matrix is constructed using the IoU distance between the predicted bounding box and the detection box as the matching cost, and the Hungarian algorithm is used to complete the matching of the high-resolution detection set and the full trajectory in the first stage. For trajectories that are not matched in the first stage, a second stage matching is performed with the low-scoring detection set to restore the tracking link of the occluded target and filter out background interference detection boxes. Update the status and ID information of all trajectories based on the two-stage matching results.
[0044] Furthermore, in step S4, the calculation method for the opening range of the door curtain is as follows: based on the baseline opening for single-pedestrian passage, the opening is superimposed according to the proportion of the horizontal size of the target and the number of targets, and the superimposed opening does not exceed the maximum full opening range of the door curtain; different types of targets correspond to different baseline opening coefficients.
[0045] In addition, it also includes a remote configuration step: the configuration management platform sends configuration instructions to the embedded main control motherboard. The parameters that can be modified include the range of the forced opening area, the opening and closing speed, the curtain closing delay time, the target category for enabled recognition, and the opening coefficient corresponding to each type of target. The parameters take effect immediately after modification.
[0046] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.
[0047] The various embodiments in this specification are described in a related manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.
[0048] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A recognition and control system based on an electric door curtain, characterized in that, include: Visual recognition and tracking module, embedded main control motherboard, motor drive actuator, and configuration management platform; The visual recognition and tracking module is electrically connected to the embedded main control motherboard. It is used to acquire real-time images of the channel, run a lightweight quantized target detection model trained with a private dataset, and combine it with the ByteTrack multi-target tracking algorithm to perform category recognition, size estimation and continuous motion trajectory tracking of passing targets entering the field of view, and output target recognition and tracking data to the embedded main control motherboard. The embedded main control motherboard is electrically connected to the motor drive actuator and has a built-in preset forced opening area and passage intention judgment logic. It is used to receive target recognition and tracking data. When the target enters the forced opening area, the curtain opening command is directly triggered. When the target is outside the forced opening area, the passage intention is determined based on continuous trajectory data. The opening range of the door curtain is calculated based on the type, quantity, and size of the target, and a PWM speed control signal with the corresponding duty cycle is output to the motor drive actuator to control the door curtain to open at the set speed. After detecting that the target has completely passed through the door curtain and left the forced opening area, a closing command is output after a delay. At the same time, the target type, opening data, and passage time are uploaded to the configuration management platform. The configuration management platform is connected to the embedded main control motherboard for remote monitoring of the curtain's operating status, issuing parameter configuration commands and firmware update packages, and supporting custom settings for target categories and access control strategies.
2. The identification control system based on an electric door curtain according to claim 1, characterized in that: During the system power-on initialization phase, the embedded main control motherboard drives the motor to drive the actuator to complete the full stroke of the door curtain. It collects and records the number of motor pulses corresponding to the fully open and fully closed positions of the door curtain, and establishes a mapping relationship between the door curtain opening degree and the motor running pulses for precise opening degree control.
3. The identification control system based on an electric door curtain according to claim 1, characterized in that: The ByteTrack multi-target tracking algorithm employs a confidence leveling and two-stage data association mechanism: The detection boxes in each frame are divided into high-scoring detection boxes and low-scoring detection boxes according to the confidence threshold; Kalman filtering is used to predict the position of the existing trajectory in the next frame, and a matching cost matrix is constructed based on the IoU distance. In the first stage, the Hungarian algorithm is used to match high-scoring detection boxes with all trajectories to obtain the first matching set, unmatched high-scoring detection boxes, and unmatched trajectories. The second stage involves a secondary matching of low-scoring detection boxes with unmatched trajectories to supplement the tracking continuity of occluded targets, and finally updates the trajectory IDs and location information of all targets.
4. The identification control system based on an electric door curtain according to claim 1, characterized in that: The motor-driven actuator controls the opening and closing speeds of the curtain to be independently adjustable, with an adjustment range of 0.1m / s to 1.0m / s; the curtain closing delay time is configurable, with a default value of 3 seconds.
5. The identification control system based on an electric door curtain according to claim 1, characterized in that: The forced opening area is a preset rectangular area adjacent to the curtain installation surface, used to cover the scenario where a target suddenly enters the passageway from the side, skipping the trajectory intent judgment step and directly triggering the curtain to open, thus eliminating the response delay of a target entering from the side.
6. The identification control system based on an electric door curtain according to claim 1, characterized in that: The configuration management platform supports custom access policies including: enabling or disabling recognition permissions for any type of target such as pedestrians, wheelchairs, hospital beds, and trolleys, and setting differentiated curtain opening coefficients and opening / closing speed parameters for different types of targets.
7. A recognition and control method based on an electric door curtain, characterized in that, Includes the following steps: S1. After the embedded main control motherboard completes the gate width self-learning calibration, it drives the vision recognition and tracking module to start the target detection mode and loads the quantized YOLO detection model and ByteTrack tracking configuration. S2, the visual recognition and tracking module acquires channel images in real time, performs category recognition and size estimation on passing targets entering the field of view, and assigns a trajectory ID to each target through the ByteTrack algorithm to continuously track its motion trajectory; S3. The embedded main control motherboard determines the target position in real time. If the target enters the preset forced opening area, it immediately outputs the curtain opening command. If the target is outside the forced opening area, it determines whether the target has the intention to pass towards the curtain based on continuous multi-frame trajectory data. If it is determined that the target has the intention to pass, it outputs the curtain opening command. S4. The embedded main control motherboard calculates the matching curtain opening range based on the type, quantity, and size of the current target, and outputs the corresponding PWM signal to control the motor drive actuator to drive the curtain to open to the target opening at a set speed. S5, the visual recognition and tracking module continuously tracks the target. When it detects that all targets have completely passed through the curtain and left the forced opening area, it controls the curtain to close after a preset delay. At the same time, it uploads the passage record to the configuration management platform.
8. The identification and control method based on an electric door curtain according to claim 7, characterized in that: The ByteTrack algorithm's tracking process in step S2 is as follows: For each frame of image, all detection results are divided into a high-scoring detection set and a low-scoring detection set according to a confidence threshold. The existing trajectory states are predicted by Kalman filtering to obtain the predicted bounding boxes of each trajectory in the next frame. A cost matrix is constructed using the IoU distance between the predicted bounding box and the detection box as the matching cost, and the Hungarian algorithm is used to complete the matching of the high-resolution detection set and the full trajectory in the first stage. For trajectories that are not matched in the first stage, a second stage matching is performed with the low-scoring detection set to restore the tracking link of the occluded target and filter out background interference detection boxes. Update the status and ID information of all trajectories based on the two-stage matching results.
9. The identification and control method based on an electric door curtain according to claim 7, characterized in that: The calculation method for the opening range of the door curtain in step S4 is as follows: based on the baseline opening for single-pedestrian passage, the opening is superimposed according to the proportion of the horizontal size of the target and the number of targets. The superimposed opening does not exceed the maximum full opening range of the door curtain; different types of targets correspond to different baseline opening coefficients.
10. The identification and control method based on an electric door curtain according to claim 7, characterized in that: It also includes a remote configuration step: the configuration management platform sends configuration commands to the embedded main control motherboard. The parameters that can be modified include the range of the forced opening area, the opening and closing speed, the curtain closing delay time, the target category for enabled recognition, and the opening coefficient corresponding to each type of target. The parameters take effect immediately after modification.