An elevator door self-adaptive anti-pinch method and system

Abstract

This invention provides an adaptive anti-pinch method and system for elevator doors. The method involves acquiring three-dimensional point cloud data of the elevator door's operating area; obtaining obstacle point cloud data from the three-dimensional point cloud data through foreground segmentation and extracting obstacle feature information; identifying the type of obstacle based on the feature information and obtaining the corresponding closing threshold; generating an adaptive closing speed curve based on the obstacle type and the minimum distance from the obstacle to the side face of the elevator door; controlling the closing process of the elevator door according to the closing speed curve; monitoring and acquiring the closing torque in real time during the closing process; and stopping the closing and reversing the direction when the closing torque exceeds the corresponding closing threshold. This adaptive anti-pinch method does not rely on visible light cameras or image data, does not perform door state detection, and does not trigger reverse opening actions based on the clamping event confirmation result. It can accurately distinguish obstacle types, reduce false triggering rates, and significantly improve elevator door operation safety and passenger experience.

Description

Technical Field

[0001] This invention relates to the field of elevator control, and in particular to an adaptive anti-pinch method and system for elevator doors. Background Technology

[0002] Elevator door anti-pinch protection is a core component of elevator safety. Existing elevator door anti-pinch systems generally have the following shortcomings: (1) Most of them use infrared light curtains, mechanical touch plates or single vision images for detection. They have poor detection capabilities for small, flexible and thin objects and are prone to missed detection. (2) The inability to distinguish the type of obstacle and the uniform use of the same door closing control strategy can easily cause pinching or frequent false triggering, affecting operating efficiency; (3) Voice prompts rely on the elevator main control board for forwarding, resulting in high response delay, lack of differentiated prompts, and poor user experience; they also lack a false trigger filtering mechanism and model self-learning ability, leading to poor long-term stability.

[0003] (4) In the prior art, elevator door control schemes based on the fusion of visible light cameras and depth sensors (such as CN120622277A) need to detect the door status and confirm the occurrence of the clamping event before performing the anti-pinch action. Its control response is an emergency reverse opening after confirming the clamping, which is essentially a passive safety intervention after the fact. In addition, this type of scheme is highly dependent on the quality of visible light images, and its performance is severely degraded in strong light and weak light environments. Summary of the Invention

[0004] To address the problems existing in the background art, the present invention proposes an adaptive anti-pinch method and system for elevator doors.

[0005] An adaptive anti-pinch method for elevator doors includes the following steps: S100: Obtain the 3D point cloud data of the elevator door operating area; S200: Obtain obstacle point cloud data from the three-dimensional point cloud data through foreground segmentation, and extract the feature information of the obstacles; S300. Identify and obtain the type of obstacle based on the feature information, and obtain the door closing threshold corresponding to the type of obstacle; S400: Generate an adaptive door closing speed curve based on the obstacle type and the minimum distance from the obstacle to the side end of the elevator door; S500. Control the closing process of the elevator door according to the closing speed curve, monitor and acquire the closing torque in real time during the closing process, and stop closing and reverse when the closing torque is greater than the corresponding closing threshold.

[0006] Based on the above, a depth sensor installed at the elevator door head collects three-dimensional point cloud data of the elevator door operating area. The depth sensor is a time-of-flight sensor or a lidar.

[0007] Based on the above, the pre-established 3D point cloud data of the empty door background in the elevator door operating area is used as background point cloud data. The acquired real-time 3D point cloud data is then used to perform foreground segmentation with the background point cloud data to obtain obstacle point cloud data.

[0008] Based on the above, in step S200, the feature information of the obstacle includes at least geometric features, the displacement vector of the target point cloud centroid between consecutive frames, the amount of projection change in the normal direction of the elevator door side end face, and reflection features.

[0009] Based on the above, in step S400, the door closing control curve is as follows: when the minimum distance is greater than D1, the normal door closing speed is maintained; when the minimum distance decreases from D1 to D2, the door closing speed smoothly decreases to 50% of the normal speed; when the minimum distance is less than D2, the door closing speed approaches zero but reverse door opening is not performed.

[0010] Based on the above, in step S200, obstacle types are identified by a pre-trained random forest classifier according to obstacle feature information, wherein obstacle types include at least human body parts, rigid objects, living objects, thin objects, and false trigger sources.

[0011] Based on the above, the geometric features include at least one of the following: three-dimensional bounding box volume, thin length, flatness, and point cloud surface curvature variation; the displacement vector of the target point cloud centroid between consecutive frames and the amount of projection change in the normal direction of the gate side end face are used to characterize the motion trend of the obstacle and are only used as input features of the classifier; the reflection features include the average reflection intensity and the reflection intensity distribution variance obtained based on the reflection intensity information output by the sensor.

[0012] Based on the above, step S400 also includes a false trigger filtering step: when the obstacle volume is less than the threshold, the number of consecutive frames is less than the threshold, or / and the trajectory tortuosity is greater than the threshold, it is determined to be a false trigger source; when it is identified as a false trigger source type, it is configured as the default door closing parameter and the anti-pinch process is not triggered.

[0013] Based on the above, in step S500, during the door closing process, corresponding voice prompts are also given according to the type of obstacle.

[0014] An elevator door adaptive anti-pinch system includes an elevator main control unit, a storage device, and a computer program stored on the storage device and executable in the elevator main control unit. When the elevator main control unit executes the computer program, it implements any of the above-mentioned elevator door adaptive anti-pinch methods.

[0015] This invention has outstanding substantive features and significant progress compared to the prior art, specifically: (1) The present invention adopts three-dimensional point cloud difference + clustering recognition, which can identify small and flexible obstacles, with high detection accuracy and strong environmental adaptability; based on random forest, it can accurately distinguish multiple types of obstacles, match differentiated door closing torque and speed, and take into account both safety and operating efficiency. (2) This invention completely abandons the visible light camera and relies solely on the active depth sensor to obtain three-dimensional point cloud. The data quality is not affected by the ambient light. Moreover, there is no clamping confirmation step. The post-event response of "judging clamping → reversing opening" is transformed into the pre-event prevention of "identifying type → adaptive speed and torque adjustment". During the entire closing process, the closing torque and speed are actively and continuously adaptively adjusted according to the type and distance of obstacles. This realizes the paradigm shift from post-event passive response to pre-event active prevention, resulting in higher operating efficiency. (3) The closing process of the present invention is controlled in stages according to distance and is filtered for false triggering, resulting in a lower false triggering rate. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart illustrating the present invention. Detailed Implementation

[0018] 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 only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] like Figure 1 As shown, an adaptive anti-pinch method for elevator doors includes the following steps: S100, acquiring three-dimensional point cloud data of the elevator door operating area; S200, acquiring obstacle point cloud data in the three-dimensional point cloud data through foreground segmentation, and extracting feature information of the obstacles; S300, identifying the type of obstacle based on the feature information, and acquiring the closing threshold corresponding to the obstacle type; S400, generating an adaptive closing speed curve based on the obstacle type and the minimum distance from the obstacle to the side end face of the elevator door; S500, controlling the closing process of the elevator door according to the closing speed curve, monitoring and acquiring the closing torque in real time during the closing process, and stopping the closing and reversing when the closing torque is greater than the corresponding closing threshold.

[0020] In practice, a depth sensor installed above the elevator door collects 3D point cloud data of the elevator door's operating area. In this embodiment, the depth sensor is a time-of-flight sensor or a lidar sensor, and it is communicatively connected to the elevator's main control unit. During operation, 3D point cloud data of the empty door background in the elevator door's operating area is first collected and pre-stored as background point cloud data. After acquiring the 3D point cloud data of the elevator door's operating area in real time, a difference operation is performed between the real-time 3D point cloud data and the background point cloud data to segment the foreground, thereby obtaining the obstacle point cloud data.

[0021] After acquiring obstacle point cloud data, obstacle feature information is extracted to identify obstacles. In this embodiment, obstacle feature information includes at least geometric features, the displacement vector of the target point cloud centroid between consecutive frames, the projection change in the normal direction of the elevator door side face, and reflection features. Specifically, the geometric features include at least one of the following: three-dimensional bounding box volume, thinness, flatness, and point cloud surface curvature change; the displacement vector of the target point cloud centroid between consecutive frames and the projection change in the normal direction of the door side face are used to characterize the obstacle's motion trend and are only used as input features for the classifier; the reflection features include the average reflection intensity and reflection intensity distribution variance obtained based on the reflection intensity information output by the sensor, that is, the reflection of sensor detection signals (such as laser pulses) on the obstacle surface. In this embodiment, a pre-trained random forest classifier is used to identify obstacle types. The random forest classifier and its training method adopt existing technologies, are trained using existing obstacle datasets, and new obstacles and their features and corresponding door closing thresholds are collected in reality through cloud sharing, etc. In practice, if a new obstacle appears that is difficult to identify, the closing threshold is controlled according to the existing lowest value, that is, treated as the highest safety level. At the same time, the detection information and identification results are sent to the host computer to facilitate the staff to debug and improve the classifier. The acquired obstacle feature information is used as input, and the classifier identifies and determines the type of obstacle.

[0022] In this embodiment, the obstacle types include at least human body parts, rigid objects, living beings, thin objects, and accidental triggering sources. Different closing thresholds are set according to different obstacle types. In this embodiment, for human body parts obstacles, a first closing torque threshold T1 is configured; for living beings (such as pets) obstacles, a third closing torque threshold T2 is configured; for rigid objects obstacles, a second closing torque threshold T3 is configured; for thin objects obstacles, a fourth closing torque threshold T4 is configured; and for accidental triggering sources obstacles, default closing parameters are configured, and the anti-pinch function is not triggered. Wherein, T4 > T3 > T2 > T1, meaning the safety level for human bodies is the highest. In this embodiment, T1 = 10 N·m, T2 = 15 N·m, T3 = 30 N·m, and T4 = 50 N·m.

[0023] After obtaining the obstacle type and its corresponding closing threshold, the distance between the obstacle and the side end of the elevator door is obtained based on the point cloud information. Since there are elevator doors on both sides of the obstacle (or one side is a door frame), the minimum distance is selected. Based on the minimum distance between the obstacle and the side end of the elevator door, a corresponding closing curve is configured. In this embodiment, the closing control curve is as follows: when the minimum distance is greater than D1, the normal closing speed is maintained; when the minimum distance decreases from D1 to D2, the closing speed smoothly decreases to 50% of the normal speed; when the minimum distance is less than D2, the closing speed approaches zero but reverse opening is not performed. The values ​​of D1 and D2 can be set according to the actual size of the elevator door in reality. In this embodiment, D1 = 30cm, D2 = 10cm. That is, distance > 30cm: normal closing speed; 10cm~30cm: speed smoothly decreases along the curve to 50% of the normal speed; < 10cm: speed further decreases to an extremely low speed approaching zero, and the door operator maintains the closing trend with minimum kinetic energy but does not actively reverse open. During the closing process, the closing torque is detected in real time. Only when the real-time closing torque exceeds the threshold corresponding to the obstacle type (such as human body part T1=10N·m), the door controller will stop closing and reverse the operation regardless of the closing position or speed. Otherwise, the door will always maintain the current speed curve and close slowly.

[0024] In practice, during the door closing process, corresponding voice prompts are given according to the type of obstacle. For example, in this embodiment, when the obstacle is identified as a human body part, the "Please be careful" prompt is triggered; when the obstacle is identified as a rigid object, the "Please move the obstacle" prompt is triggered; when the obstacle is identified as a living object, the "Please take care of your pet" prompt is triggered; and when the obstacle is identified as a false trigger source, no voice prompt is triggered.

[0025] Preferably, to avoid false triggering, this embodiment further includes a false triggering filtering step: when the obstacle volume is less than a threshold, the number of consecutive frames is less than a threshold, or / and the trajectory tortuosity is greater than a threshold, it is determined to be a false triggering source; for example, in this embodiment, a false triggering is determined if any of the following conditions are met: obstacle volume < 5cm³; number of consecutive frames < 3 frames; trajectory tortuosity > 0.8. When identified as a false triggering source type, the default door closing parameters are configured and the anti-pinch process is not triggered.

[0026] This embodiment also provides an elevator door adaptive anti-pinch system, including an elevator main control unit, a storage device, and a computer program stored on the storage device and run in the elevator main control unit. When the elevator main control unit executes the computer program, it implements any of the above-mentioned elevator door adaptive anti-pinch methods.

[0027] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A self-adaptive anti-pinch method for elevator doors, characterized in that, Including the following steps: S100: Obtain the 3D point cloud data of the elevator door operating area; S200: Obtain obstacle point cloud data from the three-dimensional point cloud data through foreground segmentation, and extract the feature information of the obstacles; S300. Identify and obtain the type of obstacle based on the feature information, and obtain the door closing threshold corresponding to the type of obstacle; S400: Generate an adaptive door closing speed curve based on the obstacle type and the minimum distance from the obstacle to the side end of the elevator door; S500. Control the closing process of the elevator door according to the closing speed curve, monitor and acquire the closing torque in real time during the closing process, and stop closing and reverse when the closing torque is greater than the corresponding closing threshold.

2. The elevator door adaptive anti-pinch method according to claim 1, characterized in that: Three-dimensional point cloud data of the elevator door operating area is collected by a depth sensor installed at the elevator door head. The depth sensor is either a time-of-flight sensor or a lidar.

3. The elevator door adaptive anti-pinch method according to claim 1, characterized in that: The pre-established 3D point cloud data of the empty door background in the elevator door operating area is used as background point cloud data. The acquired real-time 3D point cloud data is then used to perform foreground segmentation with the background point cloud data to obtain obstacle point cloud data.

4. The elevator door adaptive anti-pinch method according to claim 1, characterized in that: In step S200, the feature information of the obstacle includes at least geometric features, the displacement vector of the target point cloud centroid between consecutive frames, the amount of projection change in the normal direction of the elevator door side end face, and reflection features.

5. The elevator door adaptive anti-pinch method according to claim 1, characterized in that: In step S400, the door closing control curve is as follows: when the minimum distance is greater than D1, the normal door closing speed is maintained; when the minimum distance decreases from D1 to D2, the door closing speed smoothly decreases to 50% of the normal speed; when the minimum distance is less than D2, the door closing speed approaches zero but reverse door opening is not performed.

6. The elevator door adaptive anti-pinch method according to claim 4, characterized in that: In step S200, obstacle types are identified using a pre-trained random forest classifier based on obstacle feature information. The obstacle types include at least human body parts, rigid objects, living objects, thin objects, and false trigger sources.

7. The elevator door adaptive anti-pinch method according to claim 6, characterized in that: The geometric features include at least one of the following: three-dimensional bounding box volume, thinness, flatness, and point cloud surface curvature variation; the displacement vector of the target point cloud centroid between consecutive frames and the amount of projection change in the normal direction of the gate side face are used to characterize the motion trend of the obstacle and are only used as input features of the classifier; the reflection features include the average reflection intensity and the reflection intensity distribution variance obtained based on the reflection intensity information output by the sensor.

8. The elevator door adaptive anti-pinch method according to claim 1, characterized in that, Step S400 also includes a false trigger filtering step: when the obstacle volume is less than the threshold, the number of consecutive frames is less than the threshold, or / and the trajectory tortuosity is greater than the threshold, it is determined to be a false trigger source; when it is identified as a false trigger source type, the default door closing parameters are configured and the anti-pinch process is not triggered.

9. The elevator door adaptive anti-pinch method according to claim 1, characterized in that: In step S500, during the door closing process, corresponding voice prompts are also given according to the type of obstacle.

10. An elevator door adaptive anti-pinch system, comprising an elevator main control unit, a storage device, and a computer program stored on the storage device and executable in the elevator main control unit, characterized in that: When the elevator main control unit executes the computer program, it implements any one of the elevator door adaptive anti-pinch methods according to claims 1 to 9.

Images