A subway door safety control method and device integrated with a TOF camera

Abstract

The application belongs to the technical field of safety interlocking control, and discloses a subway door safety control method and device integrated with a TOF camera, which comprises the following steps: when a door controller confirms that a subway door is in a locked closed state, a TOF camera is controlled to collect multiple frames of depth images, the door frame reference area is aligned and filtered, the door gap width is extracted, and a closed reference door gap width distribution is formed. In the door closing termination stage, a depth image is collected and a real-time door gap width distribution is obtained. Based on the door gap determination threshold, a door gap deviation feature is obtained, and a motor driving current end fluctuation feature is collected, and the door closing state is comprehensively determined. When the door closing state is not in place, a low-speed compensation jamming control instruction is output, and the termination stage determination is repeated after compensation. If the door is still not in place after exceeding the preset number of times, an interlocking signal for prohibiting departure is output. The application improves the reliability of in-place sealing, reduces the risk of false departure and false reverse door opening, and realizes anti-pinch linkage and leaning / danger detection.

Description

Technical Field

[0001] This invention belongs to the field of safety interlocking control technology, specifically, it relates to a method and device for safety control of subway car doors integrating a TOF camera. Background Technology

[0002] With the high-density operation of urban rail transit and the widespread adoption of ATO (Automatic Train Operation) / unmanned scenarios, the interlocking of door closing and departure constitutes a critical link in the train operation safety chain. Any "departure before door is closed," "entrapment of objects or people," or "misjudgment due to insufficient sealing after door is closed" can lead to safety risks and operational delays. Existing door safety controls typically rely on door controller position / locking signals, door edge pressure / contact strips, infrared light curtains, or reaction force judgments based solely on motor current thresholds. However, under conditions of train vibration, door wear, door frame deformation, uneven gaps, and complex lighting / dust environments, these single sensors or criteria are prone to false alarms, missed alarms, or the need for frequent manual calibration.

[0003] Therefore, the invention patent CN114516344A discloses a solution that uses a time-of-flight scanning / detection device to scan the plane of the gap between the platform screen door and the train door, and determines whether there is an object obstructing the area within the warning zone, so as to output a safety contact signal to the platform screen door control system and issue an alarm. This type of solution is biased towards the existence detection and alarm linkage of "whether the gap area is obstructed": it usually uses the obstruction of a fixed plane / area as the criterion, and it is difficult to benchmark and quantify "morphological deviations" such as uneven gaps along the height of the door gap, door rebound, and insufficient sealing when locked in place during the door closing stage; and it only outputs alarms or safety contacts, which is difficult to form a reproducible state machine linkage with the door controller's compensation insertion, repeated judgment, and reverse opening due to jamming, and may still have engineering pain points such as "difficulty in distinguishing between not in place and jamming leading to mishandling", "misjudgment due to only current or only vision", and "inability to complete self-recovery within a limited number of times and give a clear prohibition of departure interlocking conditions", and lacks anti-pinch linkage and leaning / danger detection functions.

[0004] Therefore, it is necessary to design a subway car door safety control method and device integrating a TOF camera to solve the problems existing in the current technology. Summary of the Invention

[0005] In view of this, the present invention proposes a subway door safety control method and device integrating a TOF camera, aiming to solve the problem in the current technology that the traditional position or locking signal-based determination of the door's position at the closing stage of the subway door is difficult to reliably distinguish between in-position sealing, non-in-position and foreign object obstruction under working conditions such as vibration, resulting in misjudgment and mismatch between interlocking measures.

[0006] This invention proposes a subway car door safety control method integrating a TOF camera, comprising:

[0007] When the door controller confirms that the subway door is in the locked / closed state, it controls the TOF camera to acquire multiple frames of depth images, aligns and filters them based on the door frame reference area, sets multiple sampling lines along the height direction of the door gap in the door gap detection area, extracts the door gap width, and forms a closed reference door gap width distribution.

[0008] When the door closes and enters the termination phase, a depth image is acquired according to the steady-state time window and aligned with the door frame reference area to obtain the real-time door gap width distribution;

[0009] Based on the door gap determination threshold at the termination stage, the real-time door gap width distribution is compared with the closed reference door gap width distribution to obtain door gap deviation characteristics, and the end fluctuation characteristics of the motor drive current are collected to comprehensively determine the closing state. The closing state includes the in-place sealing state, the in-place state, and the foreign object blocking state.

[0010] When the door is in the closed and sealed state, a closing and departure interlock signal is output.

[0011] When the door is not in the closed state, a low-speed compensation insertion control command is output and the termination stage judgment is repeated after compensation. If the door is still not in the closed state after exceeding the preset number of times, a prohibition on departure interlock signal is output.

[0012] When the door is in a state of obstruction by a foreign object, the system outputs a stop door closing and reverse door opening control command and a prohibition departure interlock signal.

[0013] Furthermore, when obtaining the door gap deviation characteristics, the following are included:

[0014] Within each steady-state time window, the real-time door gap width is extracted along the multiple sampling lines and compared with the reference door gap width of the corresponding sampling line in the closed reference door gap width distribution to obtain the door gap deviation of each sampling line.

[0015] Based on the door gap deviation, the maximum door gap deviation and the door gap dispersion deviation are determined. When the door gap deviation corresponding to a preset number of consecutive sampling lines along the door gap height direction all exceed the abrupt change threshold, the corresponding height segment is determined as the door gap abrupt change segment; and the maximum door gap deviation and the maximum door gap dispersion deviation are used as the door gap deviation features.

[0016] Furthermore, when collecting the end-of-phase fluctuation characteristics of the motor drive current, the following are included:

[0017] The motor drive current during the termination phase is collected and a closed-loop reference current characteristic is formed, which includes the peak range at the end and the fluctuation range at the end.

[0018] When the door closes and enters the termination phase, the motor drive current is collected according to the steady-state time window to form a real-time current characteristic, which includes the terminal peak value and the terminal fluctuation amplitude. The real-time current characteristic is compared with the closed reference current characteristic to obtain the terminal fluctuation characteristic of the motor drive current.

[0019] Furthermore, when comprehensively determining the door's closed status, the following factors are considered:

[0020] When the maximum deviation of the door gap is not greater than the preset door gap threshold and the door gap dispersion deviation is not greater than the preset dispersion threshold, and the real-time current characteristic falls within the end peak range and end fluctuation range corresponding to the closed reference current characteristic, the door closing state is determined to be the in-place sealing state.

[0021] When the door gap abruptly changes and the peak value at the end exceeds the peak value range at the end or the fluctuation amplitude at the end exceeds the fluctuation range at the end, the closed state is determined to be a foreign object jamming state.

[0022] Otherwise, the door closing state is determined to be in an incomplete state.

[0023] Furthermore, when the door is in an incomplete closed state.

[0024] The low-speed compensation insertion control command includes compensation insertion speed, compensation insertion stroke, and compensation holding duration; wherein, the door control controller determines the compensation insertion stroke based on the door gap deviation characteristics, sets the compensation insertion speed to a preset low speed that is less than the closing speed of the termination stage, controls the subway door to perform insertion according to the compensation insertion stroke, and maintains the drive according to the compensation holding duration after insertion is completed.

[0025] Furthermore, when performing the compensation insertion, it also includes:

[0026] The gate controller collects the motor drive current during the compensation holding time and compares the collected motor drive current with the preset compensation current upper limit; when the motor drive current exceeds the preset compensation current upper limit, the gate controller stops the compensation insertion and outputs a stop closing and reverse opening control command.

[0027] Furthermore, the door controller counts the number of compensation insertions, and after each compensation insertion, it re-acquires the real-time door gap width distribution according to the steady-state time window and repeats the termination stage determination; when the number of compensation insertions reaches the preset number and the door closing state is still not in the correct state, it outputs a prohibition on departure interlock signal.

[0028] Furthermore, forming the closed reference door gap width distribution includes:

[0029] Geometric features of the door frame are extracted within the door frame reference area, and inter-frame registration is performed on multiple depth images to obtain an aligned depth image sequence. Multiple sampling lines are set along the height direction of the door gap in the door gap detection area, and the door gap width is extracted along each sampling line in each depth image. The door gap width of the same sampling line in the multiple depth images is taken as a preset statistical value and summarized to form the closed reference door gap width distribution.

[0030] Furthermore, the subway door safety control method integrating a TOF camera also includes anti-pinch linkage and leaning / danger detection: before the door enters the termination stage, a depth image of the danger detection area at the doorway is acquired based on the TOF camera and a person or object target is identified. The determination is made based on the minimum distance between the target and the door movement boundary and / or the intrusion amount. When the minimum distance is less than a preset anti-pinch threshold or the intrusion amount exceeds a preset intrusion threshold, a deceleration, stop closing, and / or reverse opening control command is output. When the target is within a preset leaning detection area and the distance between the target and the door is less than a preset leaning threshold and the duration exceeds a preset duration threshold, the door opening command and / or closing command are suppressed, and the warning module is triggered to output a buzzer and voice prompt.

[0031] Compared with existing technologies, the advantages of this invention are as follows: It elevates the determination of the arrival status of subway doors at the termination stage from traditional single-point position or locking signal criteria to a distributed quantitative authentication system based on the door gap width distribution acquired by a TOF camera and using the closed reference door gap width distribution as a comparable reference. Furthermore, by aligning the door frame reference area and acquiring data within a steady-state time window, it establishes repeatable measurement conditions resistant to vibration and light disturbances, enabling door gap deviation characteristics to reflect non-uniform anomalies such as local non-insertion and local jamming. Simultaneously, it collaboratively determines the door gap deviation characteristics with the fluctuation characteristics at the end of the motor drive current, forming mutually exclusive current splitting of the in-place sealing state, the non-in-place state, and the foreign object jamming state. This allows for differentiated actions at the control level that match the state, reducing the risk of missed detections and false releases when the locking signal is in place but the seal is not complete. It also avoids misjudging non-in-place as jamming, leading to frequent reverse opening and reduced operational efficiency. Finally, by consistently outputting interlocking signals, it directly incorporates the authenticity of the door gap seal into the departure safety link, improving the reliability and maintainability of the termination stage determination.

[0032] On the other hand, this application also provides a subway door safety control device integrating a TOF camera, used to apply the above-mentioned subway door safety control method integrating a TOF camera, including:

[0033] The door control controller is used to control the TOF camera to acquire multiple frames of depth images when the subway car door is confirmed to be in the locked / closed state, and to set multiple sampling lines in the door gap detection area; it is used to control the TOF camera to obtain the real-time door gap width distribution; it is used to obtain door gap deviation characteristics and to collect the end fluctuation characteristics of motor drive current to comprehensively determine the door closing state; and it outputs a control signal according to the door closing state.

[0034] A TOF camera, connected to the gating controller, is used to output depth images to the gating controller;

[0035] A motor drive current acquisition interface is connected to the gate controller and is used to provide the motor drive current to the gate controller.

[0036] The door operator control interface is connected to the door controller and is used to receive the stop closing, reverse opening, and low-speed compensation insertion control commands output by the door controller.

[0037] An interlocking signal output interface is connected to the gate controller and is used to output the departure permission interlocking signal and the departure prohibition interlocking signal.

[0038] It is understandable that the aforementioned subway door safety control method and device integrating a TOF camera have the same beneficial effects, and will not be elaborated further here. Attached Figure Description

[0039] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0040] Figure 1 A flowchart of a subway door safety control method integrating a TOF camera provided in an embodiment of the present invention;

[0041] Figure 2 This is a structural block diagram of a subway door safety control device integrating a TOF camera, provided in an embodiment of the present invention. Detailed Implementation

[0042] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the disclosure to those skilled in the art. It should be noted that, unless otherwise specified, embodiments and features in the embodiments of the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0043] In some embodiments of this application, see Figure 1 As shown, this application proposes a subway car door safety control method integrating a TOF camera, including:

[0044] S100: When the door controller confirms that the subway door is in the locked / closed state, it controls the TOF camera to acquire multiple frames of depth images, aligns and filters them based on the door frame reference area, sets multiple sampling lines along the height direction of the door gap in the door gap detection area, extracts the door gap width, and forms a closed reference door gap width distribution.

[0045] S200: When the door closes and enters the termination phase, a depth image is acquired according to the steady-state time window and aligned with the door frame reference area to obtain the real-time door gap width distribution.

[0046] S300: Based on the comparison of the real-time door gap width distribution and the closing reference door gap width distribution with the door gap judgment threshold at the termination stage, the door gap deviation characteristics are obtained, and the fluctuation characteristics of the end of the motor drive current are collected to comprehensively determine the closing state. The closing state includes the in-place sealing state, the in-place state, and the foreign object blocking state.

[0047] When the door is in the closed and sealed state, the output signal is closed and interlocked with the departure permission signal.

[0048] When the door is not in the closed state, a low-speed compensation insertion control command is output and the termination stage judgment is repeated after compensation. If the door is still not in the closed state after exceeding the preset number of times, a prohibition departure interlock signal is output.

[0049] When the door is blocked by a foreign object while closed, the system outputs a stop closing and reverse opening control command and a prohibition departure interlock signal.

[0050] Specifically, in this embodiment, the door control controller is electrically and / or communicatively connected to the TOF camera and connected to the door operator drive unit. The integrated TOF camera refers to the TOF camera being integrated into the subway car door control system as a door gap detection sensor. The TOF camera can be separately installed from the door control controller or integrated into the same control box. The door control controller executes the door closing state machine and outputs door operator control commands and departure interlocking signals. When the door control controller confirms that the subway car door is in a locked / closed state, it triggers a reference establishment process. This confirmation is obtained by satisfying both the locking detection signal and the door closing completion state, ensuring that the reference corresponds to the closed condition where the door leaf has completed the final stage of insertion and the door lock mechanism is active. The door control controller controls the TOF camera to acquire multiple frames of depth images under this condition, filters each frame of depth image, and removes invalid depth points. Simultaneously, a door frame reference area with a stable door frame structure that does not move with the door leaf is selected from the depth images, and the multiple frames of depth images are aligned based on the door frame reference area to suppress depth jitter caused by vehicle vibration and unify the door gap measurement coordinate reference. Subsequently, the gate controller sets multiple sampling lines along the height of the door gap in the door gap detection area. The door gap detection area covers the bonding boundary between the door leaf and the door frame. The sampling lines are arranged along the height of the door gap to characterize the bonding differences of the door gap at different heights. The gate controller extracts the door gap width along each sampling line in the aligned depth image of each frame, and takes a preset statistical value for the door gap width of the same sampling line in multiple frames to reduce the influence of noise. The results are summarized to form a closed reference door gap width distribution and stored for comparison benchmarks in the subsequent termination stage of in-place authentication.

[0051] When the door controller executes the closing process and enters the termination phase, it acquires depth images according to a steady-state time window and aligns them based on the door frame reference area to obtain the real-time door gap width distribution during the termination phase. Specifically, the steady-state time window corresponds to the last time period of the termination phase, during which the relative pose change of the door leaf is relatively small, facilitating stable extraction of the door gap width distribution. After acquiring and aligning multiple frames of depth images within each steady-state time window, the door controller extracts the real-time door gap width along the aforementioned multiple sampling lines and takes a preset statistical value for the door gap width of the same sampling line within the steady-state time window to form a real-time door gap width distribution that can be directly compared with the closed reference door gap width distribution. The door controller compares the real-time door gap width distribution with the closed reference door gap width distribution based on the door gap judgment threshold during the termination phase to obtain door gap deviation characteristics. Simultaneously, the door controller acquires the motor drive current during the termination phase and extracts the end fluctuation characteristics of the motor drive current to characterize the force and resistance conditions during the termination phase. The door controller comprehensively determines the door gap deviation characteristics and the fluctuation characteristics of the motor drive current, and outputs the door closing state. The closing state includes a sealed state, a not sealed state, and a foreign object jamming state. The sealed state corresponds to the situation where the door gap width distribution meets the termination stage threshold and the fluctuation of the motor drive current falls within the closed reference current characteristic range. The foreign object jamming state corresponds to the situation where the door gap deviation shows a local abrupt change and the fluctuation of the motor drive current exceeds the closed reference current characteristic range. The not sealed state corresponds to the remaining situations that do not meet the aforementioned sealed state and foreign object jamming state. In one implementation, the door controller estimates the door speed v and acceleration a based on the door encoder position or door displacement. When |v|≤v_th and |a|≤a_th and the duration T≥T_th is satisfied, this duration is determined as the steady-state time window. v_th, a_th, and T_th are preset parameters. v_th can be set to 0.2 to 0.6 times the rated closing speed of the termination stage, and T_th can be set to 100ms to 500ms to ensure the repeatability of door gap width extraction within the window.

[0052] After the door is closed, the door controller executes corresponding control and interlocking outputs based on the closing status. When the door is in a sealed state, the door controller outputs a closing pass signal and a departure permission interlocking signal, enabling the train control system to confirm the departure conditions based on this interlocking signal. When the door is not in a sealed state, the door controller outputs a low-speed compensation insertion control command, controlling the door operator drive unit to perform compensation insertion at a speed lower than the closing speed of the termination stage, and repeats the termination stage judgment after compensation is completed. If the door is still not in a sealed state after the preset number of compensation attempts, the door controller outputs a departure prohibition interlocking signal to block the departure conditions and prevent accidental release when the door gap does not meet the sealing requirements. When the door is blocked by a foreign object, the door controller outputs a stop closing and reverse opening control command, causing the door operator drive unit to stop closing and open in reverse to release the obstruction, while simultaneously outputting a departure prohibition interlocking signal to maintain safety constraints. The closing pass signal is used to characterize the gate controller's determination of the sealing status. This signal can be used as an internal status flag for the gate controller's state machine to call, or it can be output synchronously with the departure permission interlock signal. When the departure permission interlock signal is valid, the closing pass signal can also be configured as an external output signal identical to the departure permission interlock signal.

[0053] Understandably, by establishing a closed reference door gap width distribution in the locked state, and obtaining a comparable real-time door gap width distribution by aligning it with the door frame reference area through a steady-state time window during the termination phase, the arrival authentication of the subway door during the termination phase is upgraded from a single-point signal judgment to a distributed quantitative judgment along the door gap height direction. Furthermore, by combining the end-of-line fluctuation characteristics of the motor drive current to achieve mutual exclusion and current diversion for the arrival sealing state, the non-arrival state, and the foreign object jamming state, differentiated control strategies and departure interlocking signals are output. This allows for closed-loop correction of the non-arrival situation through low-speed compensation insertion and re-judgment, timely stopping the door closing and reversing the door opening in the jamming situation, and mandatory safety constraints through the prohibition of departure interlocking when the arrival sealing condition is not met. Overall, this improves the reliability of the termination phase judgment, the matching of the handling strategy, and the consistency of the interlocking output.

[0054] In this embodiment, the logic for determining the foreign object jamming state is as follows: under the premise of the door gap abrupt change segment, as long as either the end peak value exceeds the end peak value range or the end fluctuation amplitude exceeds the end fluctuation range, it is determined to be a foreign object jamming state.

[0055] Specifically, in this embodiment, after the gate controller enters the termination phase of closing, it extracts the gap width from the depth image output by the TOF camera using a steady-state time window as the processing unit, and forms a real-time gap width distribution accordingly. The steady-state time window corresponds to the end time period of the termination phase. During this time period, the gate controller acquires multiple frames of depth images and maintains the gap detection area and sampling line positions consistent with the closed reference gap width distribution, thereby ensuring comparability of each sampling line. Within each steady-state time window, the gate controller extracts the real-time gap width along multiple sampling lines and compares it one by one with the reference gap width of the corresponding sampling line in the closed reference gap width distribution to obtain the gap deviation of each sampling line. Based on the gap deviation of each sampling line, the maximum gap deviation is determined, and the gap dispersion deviation is determined according to the dispersion of the gap deviation in the sampling line set to reflect the uniformity of the gap deviation in the height direction. To identify non-uniform door gap anomalies caused by local jamming or partial lack of insertion, the door controller also detects a preset number of consecutive sampling lines along the height direction of the door gap. When the door gap deviation corresponding to the consecutive sampling lines all exceeds the abrupt change threshold, the corresponding height segment is determined as a door gap abrupt change segment. The maximum door gap deviation and the maximum door gap dispersion deviation are used as door gap deviation features for subsequent door closing state determination. In one implementation, the door gap dispersion deviation D is defined as the range of the door gap deviation Δw_i of each sampling line, i.e., D=max(Δw_i)−min(Δw_i), used to characterize the uniformity of the door gap deviation along the height direction. When D is calculated for multiple steady-state time windows, the maximum value can be recorded as the maximum door gap dispersion deviation. Optionally, the door gap dispersion deviation can also be the standard deviation or interquartile range of Δw_i to improve robustness to outliers.

[0056] Furthermore, in this embodiment, the motor drive current is synchronously acquired during the termination phase to extract the end-of-phase fluctuation characteristics of the motor drive current, which are used to characterize the force changes and resistance signs during the termination phase. Under the locking and closing condition corresponding to the established closed reference door gap width distribution, the gate controller synchronously acquires the motor drive current during the termination phase to form a closed reference current characteristic. The closed reference current characteristic includes at least the end-of-phase peak range and the end-of-phase fluctuation range, used to describe the typical fluctuation range of the motor drive current at the end of the termination phase during normal sealing. When the door closes and enters the termination phase, the gate controller acquires the motor drive current according to a steady-state time window and forms a real-time current characteristic. The real-time current characteristic includes at least the end-of-phase peak value and the end-of-phase fluctuation amplitude. The gate controller compares the real-time current characteristic with the closed reference current characteristic to obtain the end-of-phase fluctuation characteristics of the motor drive current, which, together with the door gap deviation characteristics, supports the current distribution determination of whether the door is in place, not in place, or stuck.

[0057] In this embodiment, the door controller comprehensively determines the closing state based on a combination of door gap deviation characteristics and motor drive current end-fluctuation characteristics, thereby avoiding misjudgments caused by a single information source. When the maximum door gap deviation is not greater than a preset door gap threshold and the door gap dispersion deviation is not greater than a preset dispersion threshold, and the real-time current characteristic falls within the end-peak range and end-fluctuation range corresponding to the closed reference current characteristic, the door controller determines the closing state to be a sealed state. When there is a sudden change in the door gap and the end-peak exceeds the end-peak range or the end-fluctuation amplitude exceeds the end-fluctuation range, the door controller determines the closing state to be a foreign object jamming state. Other situations are determined to be an incomplete state, thus providing clear state input for subsequent low-speed compensation insertion re-judgment or reverse opening.

[0058] Understandably, by comparing the closed-loop reference gap width distribution with the real-time gap width distribution along a sampling line, a gap deviation characteristic is formed that simultaneously includes overall deviation and local abrupt change indications. This provides a distributed quantitative basis for determining the sealing position during the termination stage. Furthermore, by constraining the fluctuations at the end of the motor drive current using the closed-loop reference current characteristics, the geometric deviation of the gap and abnormal drive force are consistently integrated. This allows for a more reliable differentiation between sealing position, non-sealing position, and foreign object obstruction during the termination stage, reducing the probability of misjudgments caused by vibration, noise, or uneven local force, and improving the matching and maintainability of the closing handling strategy and interlock output.

[0059] In this embodiment, when the door controller determines that the door is not fully closed during the termination phase, it does not directly output a reverse opening. Instead, it enters a low-speed compensation insertion process to achieve closed-loop correction. The door controller generates a low-speed compensation insertion control command, which includes at least a compensation insertion speed, a compensation insertion stroke, and a compensation holding duration. The compensation insertion speed is a preset low speed, less than the closing speed during the termination phase, used to reduce the risk of passenger pinching and mechanical impact during the final insertion. The compensation holding duration is used to maintain short-term drive after the compensation insertion is completed to counteract elastic rebound and promote door gap sealing. The door controller determines the compensation insertion stroke based on door gap deviation characteristics. Specifically, it can use a method where the maximum door gap deviation and the door gap dispersion deviation are jointly mapped to the compensation insertion stroke, so that situations with large overall deviations or high unevenness in the height direction result in more sufficient compensation insertion. To avoid overloading the mechanism due to excessive compensation insertion, the gate controller can set an upper limit constraint on the compensation insertion stroke, and under this constraint, control the subway door to perform insertion according to the compensation insertion stroke and maintain the drive for the compensation holding duration after insertion is completed.

[0060] Furthermore, this embodiment introduces current constraints during the compensation insertion process to identify the risk of compensation insertion turning into jamming. The gate controller collects the motor drive current during the compensation holding time and compares the collected motor drive current with a preset upper limit of the compensation current. The preset upper limit of the compensation current characterizes the maximum allowable force for compensation insertion, and can be derived from the peak range and fluctuation range at the end of the closed reference current characteristic, with a safety margin. When the motor drive current exceeds the preset upper limit of the compensation current, the gate controller determines that there is an overload risk in the compensation insertion and stops the compensation insertion, outputting stop closing and reverse opening control commands to avoid clamping damage or damage to the gate mechanism caused by continuous insertion, thereby distinguishing between "incomplete compensation" and "suspected jamming release" at the control action level.

[0061] In this embodiment, to ensure the deterministic confirmation of the state after compensation insertion, the gate controller counts the number of compensation insertion attempts. After each compensation insertion, it re-acquires the real-time door gap width distribution within a steady-state time window and repeats the termination stage determination to determine whether the compensation insertion has changed the closed state to a properly sealed state or a foreign object blocking state. If the number of compensation insertion attempts reaches a preset number and the closed state is still not in the properly sealed state, the gate controller outputs a prohibition on departure interlock signal to include the abnormal situation of "multiple compensation attempts still not in place" in the interlock constraint, preventing accidental departure when the door gap does not meet the sealing requirements. The preset number of attempts limits the upper limit of compensation attempts, preventing the compensation process from looping infinitely and reducing processing efficiency.

[0062] Understandably, the introduction of a closed-loop control chain—"low-speed compensation insertion—maintaining compression—steady-state reassessment—number of attempts interlocking"—to address the issue of insufficient door positioning prioritizes gentle compensation rather than direct reverse door opening, thereby reducing the impact of malfunctions on operational efficiency. Simultaneously, using the upper limit of the motor drive current as a force constraint during the compensation and maintenance phase allows for timely identification of overload trends and switching between stop-closing and reverse-opening measures, preventing compensation insertion from evolving into jamming damage. Furthermore, by outputting a prohibition-departure interlocking signal when the maximum number of compensation attempts has been reached, the risk of insufficient door positioning at the end of the line is constrained by a deterministic interlocking method, improving the consistency and maintainability of the termination phase handling strategy and safety interlocking outputs.

[0063] In some embodiments of this application, forming a closed reference door gap width distribution includes: extracting door frame geometric features within a door frame reference area and performing inter-frame registration on multiple depth images to obtain an aligned depth image sequence. Multiple sampling lines are set along the door gap height direction in the door gap detection area, and the door gap width is extracted along each sampling line in each depth image frame. A preset statistical value is taken for the door gap width of the same sampling line in multiple depth images, and these values ​​are summarized to form a closed reference door gap width distribution.

[0064] In some embodiments of this application, obtaining the real-time door gap width distribution includes: acquiring multiple frames of depth images within each steady-state time window, and performing inter-frame registration of the multiple frames of depth images based on the door frame reference area; statistically analyzing the proportion of invalid depth points within the door gap detection area, and discarding the current steady-state time window and re-acquiring data when the proportion of invalid depth points exceeds a preset threshold; and extracting the door gap width along multiple sampling lines within the steady-state time window when the proportion of invalid depth points does not exceed the preset threshold, and taking a preset statistical value for the door gap width of the same sampling line within the steady-state time window, summarizing them to form the real-time door gap width distribution.

[0065] Specifically, in this embodiment, to ensure that the closed reference door gap width distribution and the real-time door gap width distribution have consistent coordinate references and comparability, the door controller extracts the door frame geometric features within the door frame reference area and uses these features as the basis for inter-frame registration to register multiple frames of depth images, resulting in an aligned depth image sequence. The door frame reference area is selected as the region corresponding to the fixed structure of the door frame to ensure that this region remains relatively stationary during the closing and locking phases, thereby suppressing overall pose jitter caused by vehicle vibration and reducing door gap measurement errors caused by slight changes in viewing angle. When forming the closed reference door gap width distribution, the door controller sets multiple sampling lines along the door gap height direction in the door gap detection area and extracts the door gap width along each sampling line in each aligned depth image. The door gap detection area covers the depth variation region of the door leaf and door frame contact boundary, and the sampling lines are used to characterize the width difference of the door gap at different heights. To reduce the impact of single-frame noise and local flying spots on the door gap width, the gate controller takes a preset statistical value for the door gap width of the same sampling line in multiple frames of depth images, and summarizes the preset statistical values ​​corresponding to each sampling line to form a closed reference door gap width distribution. The preset statistical values ​​are used to robustly converge the door gap widths of the same sampling line across multiple frames to obtain a reusable reference distribution. In one implementation, the gate controller projects the pixel set corresponding to the sampling line in the depth image into a point cloud, extracts the point set of the outer surface of the door leaf and the point set of the door frame sealing surface in the neighborhood of a preset width on both sides of the sampling line, respectively, performs local planar fitting or surface fitting on the two point sets, and calculates the shortest distance between the two fitted surfaces in the door gap normal direction as the door gap width at that sampling line. When the fitting residual exceeds a preset residual threshold, the measurement value of that sampling line in that frame is discarded, and robust convergence is performed from the statistical values ​​of multiple frames.

[0066] In this embodiment, when acquiring the real-time door gap width distribution, the gate controller acquires multiple frames of depth images in steady-state time windows and performs inter-frame registration of the multiple frames of depth images based on the door frame reference area to ensure that the real-time door gap measurement and the closed reference door gap measurement use the same coordinate reference. Considering that strong reflections, occlusions, or sensor noise may cause invalid depth points in the door gap detection area, this embodiment introduces an invalid depth point ratio gating mechanism: the gate controller counts the proportion of invalid depth points in the door gap detection area. When the proportion of invalid depth points exceeds a preset ratio threshold, the current steady-state time window is discarded and re-acquisition is performed, thereby avoiding the output of the real-time door gap width distribution and subsequent misjudgments when the measurement reliability is insufficient. When the proportion of invalid depth points does not exceed the preset ratio threshold, the gate controller extracts the door gap width along multiple sampling lines within the steady-state time window, and takes a preset statistical value for the door gap width of the same sampling line within the steady-state time window. Then, the preset statistical values ​​corresponding to each sampling line are summarized to form the real-time door gap width distribution. The above processing ensures high consistency in the real-time door gap width distribution within the same steady-state time window, providing reliable input for subsequent door gap deviation feature calculation. Invalid depth points include pixels with a depth value of 0, exceeding the upper / lower limit of the measurement range, having a depth confidence level below a preset confidence threshold, or being marked as abnormal by the filter. The invalid depth point ratio is the proportion of invalid depth points within the door gap detection area to the total number of pixels in that area.

[0067] Understandably, by extracting the geometric features of the door frame from the reference area and performing inter-frame registration, the closure reference and real-time measurements during the termination stage are unified to the same coordinate reference, reducing the impact of vehicle vibration on the extraction of the door gap width distribution from the source. By setting multiple sampling lines along the height direction of the door gap and using preset statistical values ​​to robustly converge multi-frame measurements, the door gap width distribution can reflect local differences in the height direction rather than single-point results. Furthermore, measurement reliability gating is achieved through the proportion of invalid depth points and a preset proportion threshold, preventing the generation of real-time door gap width distribution and its transmission to subsequent judgment logic when the depth data quality is insufficient. This improves the comparability and reliability of the door gap width distribution, providing a stable data foundation for the termination stage's sealing certification and state diversion judgment.

[0068] In this embodiment, the subway door safety control method integrating a TOF camera further includes: in S150, anti-pinch linkage and leaning / danger detection: before the door enters the termination stage, a depth image of the danger detection area at the door is acquired based on the TOF camera and a person or object target is identified. The determination is made based on the minimum distance between the target and the door movement boundary and / or the intrusion amount; when the minimum distance is less than the preset anti-pinch threshold or the intrusion amount exceeds the preset intrusion threshold, a deceleration, stop closing the door and / or reverse opening control command is output; when the distance between the target and the door body in the preset leaning detection area is less than the preset leaning threshold and the duration exceeds the preset duration threshold, the door opening command and / or closing command are suppressed, and the warning module is triggered to output a buzzer and voice prompt.

[0069] Specifically, in this embodiment, before the door closes and the door enters the termination phase, the door controller periodically triggers a TOF camera to acquire depth images (e.g., sampling at 20–60 fps) and performs inter-frame alignment and depth filtering based on the door frame reference area to obtain a stable 3D point cloud / depth matrix for the doorway. The door controller uses the door frame coordinate system as a unified reference and pre-defines a hazard detection area and a leaning detection area. The hazard detection area covers the spatial range that the door movement may sweep across and its outward safety margin, while the leaning detection area covers the outer surface neighborhood of the door and the side neighborhood of the door frame, reflecting risk scenarios such as people leaning against each other, limbs probing in, or luggage approaching. The geometric boundaries of these areas can be jointly defined using polygonal ROIs and depth threshold surfaces, and mapped to depth image coordinates by installing calibration parameters (camera extrinsic parameters, door frame dimensions, door edge line / surface model), thereby ensuring the consistency and reusability of area definitions for different vehicles and door types.

[0070] Specifically, the door controller extracts targets from the depth image within the danger detection area at the doorway. Targets can be obtained through "background depth model difference + connected component clustering" or "point cloud density clustering," and suspected targets are filtered out using minimum voxel filtering to suppress noise such as dust and reflection. Subsequently, the door controller constructs the door motion boundary, which is preferably the spatiotemporal sweep envelope formed by the door edge line and the door plane during the process from the current opening degree to the target opening degree / closing opening degree (which can be generated offline and solidified as an envelope surface, or updated online based on door displacement / encoder information), and calculates the minimum distance d_min between the target and the door motion boundary; at the same time, it calculates the intrusion amount I, which can be defined as "the percentage of points / area where the target points are concentrated within the motion boundary (or within the buffer zone extended outward by Δ)" or "the integral value of the intrusion depth," to simultaneously cover two types of situations: slender limbs protruding and large luggage approaching each other. The gate controller compares d_min with the preset anti-pinch threshold D_pin and I with the preset intrusion threshold I_in. The thresholds can be set in combination with TOF ranging accuracy, gate speed and braking response time (e.g., D_pin is 0.10–0.25m, Δ is 0.02–0.05m, I_in is 5%–15%), and can use N consecutive frames to meet the conditions (e.g., N=3–8 frames) as the trigger judgment to avoid false triggering due to instantaneous noise. When d_min < D_pin or I > I_in, the gate controller outputs control commands according to the gradient strategy of "deceleration-stop-reverse opening". The deceleration trigger can use a more lenient threshold D_pre (D_pre > D_pin) to achieve early warning braking, and the stop / reverse opening trigger uses a more stringent threshold or is combined with a sudden increase in motor current as the confirmation condition, thereby ensuring the timeliness and stability of the safety action.

[0071] Furthermore, to achieve verifiable determination of leaning / danger detection, this embodiment adopts a composite condition of "spatial proximity + time duration" for leaning behavior: when the distance d_lean between the target and the outer surface of the door or the side of the door frame within the preset leaning detection area is less than the preset leaning threshold D_lean and the duration T_lean exceeds the preset duration threshold T_th, it is determined that there is leaning / dangerous proximity; where D_lean can be 0.05–0.15m, T_th can be 0.5–2.0s, and continuous frame counting can also be introduced to achieve anti-jitter. When the judgment is valid, the door controller suppresses the door opening and / or closing commands and triggers the warning module to output a buzzer and voice prompt. To avoid conflict with the anti-pinch linkage action, this embodiment specifies the control priority as follows: the anti-pinch linkage (deceleration / stop / reverse door opening) has the highest priority, and its control output is not limited by the "suppression of opening / closing commands"; the leaning suppression only acts on the regular opening and closing command channel and is automatically released after the target leaves the leaning detection area or d_lean≥D_lean and the release condition T_rel is continuously met (e.g., 0.3–1.0s), thereby ensuring the consistency and reproducibility of the control state machine closed loop. For example, during the door closing stage and before entering the termination stage, the TOF camera identifies a passenger arm target in the danger detection area at the doorway, calculates the minimum distance between the target and the door movement boundary as 0.12m, which is less than the preset anti-pinch threshold of 0.15m, then the door controller first outputs a deceleration command; if the minimum distance subsequently decreases further to 0.08m or the intrusion exceeds the preset intrusion threshold, then the door controller outputs a stop closing command and executes reverse door opening. For example, before the train is about to depart, if it is detected that a passenger's body is continuously pressed against the outer surface of the door, the distance between the target and the door is 0.06m and the duration exceeds 1.0s, the door control controller will suppress the door closing command and trigger the warning module to output a buzzer and voice prompt; the suppression will be lifted when the target distance returns to not less than 0.10m and lasts for 0.5s.

[0072] Understandably, using TOF depth perception to achieve non-contact advance detection of the door opening and closing path shifts the traditional "post-contact response" relying on edge strips / light curtains to "early warning and braking upon approach," significantly reducing the risk of pinching in densely populated areas and minimizing repeated opening and closing due to accidental triggering. Simultaneously, the highly perceptible warnings of leaning / dangerous proximity and the buzzer-like voice alerts allow for proactive passenger intervention, reducing high-risk stops in the door area. The progression from the overall solution lies in this: the overall solution focuses on state recognition and interlocking output based on door gap deviation and current end-point fluctuations during the termination phase. This embodiment preemptively eliminates external risk sources such as "personnel / objects entering the movement boundary, leaning / approaching" before the termination phase, reducing the frequency of abnormal conditions and triggering of incomplete or jammed states during the termination phase. This enhances safety while further stabilizing interlocking decisions and reducing unnecessary actions like compensation insertion and reverse door opening, thus contributing to the overall beneficial effects.

[0073] In one specific embodiment, the gate controller is electrically connected to the TOF camera, the gate drive unit, the motor drive current acquisition interface, and the interlocking signal output interface. The TOF camera is fixedly installed above the gate frame, and its field of view covers the gate gap detection area. The gate frame reference area is selected from the depth-stable region corresponding to the fixed structure inside the gate frame for inter-frame registration. The TOF camera outputs depth images at a frame rate of 30 frames per second. The gate controller uses a steady-state time window for processing during the termination phase. The steady-state time window length is set to 200 milliseconds, and approximately 6 frames of depth images can be obtained in each steady-state time window.

[0074] During the baseline establishment phase, the gate controller triggers data acquisition after confirming that the subway doors are in a locked / closed state, continuously acquiring 24 frames of depth images. The gate controller extracts the geometric features of the door frame within the door frame reference area, performs inter-frame registration on the 24 depth images to obtain an aligned depth image sequence, and then filters and removes invalid depth points from the sequence. Subsequently, 12 sampling lines are arranged along the height direction of the door gap in the door gap detection area, and the door gap width of each of the 12 sampling lines is extracted for each depth image frame. Taking the first sampling line as an example, the door gap widths extracted in the 24 frames are 0.62, 0.60, 0.61, 0.63, 0.60, and 0.62 mm, respectively. The gate controller takes a preset statistical value for the door gap width of the same sampling line; in this embodiment, the median value is used as the preset statistical value. Therefore, the baseline door gap width of the first sampling line is 0.61 mm. The reference gap widths of 12 sampling lines were obtained in the same manner, and summarized to form a closed reference gap width distribution, for example, [0.61, 0.60, 0.62, 0.61, 0.60, 0.61, 0.62, 0.61, 0.60, 0.61, 0.62, 0.61] mm. Synchronously with the closed reference gap width distribution, the gate controller collected the motor drive current during the termination phase and statistically obtained the closed reference current characteristics: the peak value range at the end is 3.8 to 4.6 A, and the fluctuation range at the end is 0.5 to 0.9 A. The fluctuation range at the end is defined as the difference between the maximum and minimum values ​​of the motor drive current during the final time period of the termination phase. The peak range and fluctuation range at the end can be obtained by statistically analyzing the terminal current collected by the gate controller during the M consecutive normal locking and closing conditions. For example, the peak range at the end can be formed by taking [median value ± k × standard deviation] for the M peak values ​​at the end, and the fluctuation range at the end can be formed by taking [median value ± k × standard deviation] for the M fluctuation amplitudes at the end, where M is an integer not less than 3 and k is a preset coefficient.

[0075] During an actual door closing process, the door controller calculates the door speed and acceleration based on the position of the door operator's encoder. When |v|≤v_th and |a|≤a_th, and the duration T≥T_th, the closing is considered to have entered the termination stage, and the first steady-state time window is entered for data acquisition and judgment. Within this steady-state time window, the door controller acquires 6 frames of depth images and performs inter-frame registration. A total of 2000 depth sampling points are counted within the door gap detection area, with 140 invalid depth points, resulting in an invalid depth point ratio of 7%. The preset ratio threshold is set to 10%, therefore the current steady-state time window is valid, and the real-time door gap width continues to be extracted. The gate controller extracts the real-time door gap width along 12 sampling lines, and takes the median value of the door gap width over 6 frames for each sampling line as a preset statistical value, resulting in a real-time door gap width distribution of [0.84, 0.82, 0.83, 0.85, 0.82, 0.84, 0.83, 0.84, 0.82, 0.83, 0.84, 0.83] mm. Then, it compares each sample with the corresponding sampling line of the closed reference door gap width distribution, obtaining door gap deviations of [0.23, 0.22, 0.21, 0.24, 0.22, 0.23, 0.21, 0.23, 0.22, 0.22, 0.22, 0.22] mm. The gate controller determines the maximum door gap deviation to be 0.24 mm. The door gap dispersion deviation is used to characterize the degree of dispersion of the deviation in the height direction. In this embodiment, the difference between the maximum and minimum door gap deviation is used as the door gap dispersion deviation, which is 0.24 minus 0.21, resulting in 0.03 mm. To identify abrupt door gap changes, three sampling lines are taken consecutively with a preset threshold of 0.35 mm. Since the door gap deviation of all sampling lines is less than 0.35 mm, there are no abrupt door gap changes. Simultaneously, the door controller collects the motor drive current within this steady-state time window to form real-time current characteristics: the peak value at the end is 4.2 A, and the fluctuation amplitude at the end is 0.7 A. This real-time current characteristic falls within the peak value and fluctuation range of the closed reference current characteristic. The door controller makes a comprehensive judgment: the preset door gap threshold is 0.20 mm, and the preset dispersion threshold is 0.10 mm. The current maximum door gap deviation of 0.24 mm is greater than 0.20 mm, failing to meet the sealing condition criterion. Furthermore, since there is no abrupt change in the door gap and the real-time current characteristic does not exceed the closed reference current characteristic range, the foreign object jamming condition criterion is not met. Therefore, the door is determined to be in an incomplete closed state.

[0076] In the pre-positioned state, the door controller outputs a low-speed compensation insertion control command and starts a compensation insertion count counter. The compensation insertion speed is set to 0.5 times the closing speed of the final stage; for example, if the closing speed of the final stage is 0.08 m / s, then the compensation insertion speed is 0.04 m / s. The compensation hold duration is set to 300 milliseconds. The compensation insertion stroke is determined by the door gap deviation characteristics. In this embodiment, a segmented mapping rule is adopted: when the maximum door gap deviation is between 0.20 and 0.30 mm and the door gap dispersion deviation is not greater than 0.10 mm, the compensation insertion stroke is set to 0.8 mm. Therefore, the compensation insertion stroke in this case is 0.8 mm. The door controller controls the subway door to perform insertion at 0.8 mm and holds the drive for 300 milliseconds after insertion. During the compensation hold duration, the motor drive current is collected and compared with the preset compensation current upper limit. The preset compensation current upper limit is set to 5.0A. The actual collected maximum motor drive current is 4.7A, which does not exceed the upper limit, so the compensation insertion does not stop. After compensation, the gate controller reacquires the real-time door gap width distribution according to the steady-state time window and repeats the termination phase determination: within the new steady-state time window, the real-time door gap width distribution becomes [0.70, 0.68, 0.69, 0.70, 0.68, 0.69, 0.69, 0.69, 0.69, 0.69, 0.69, 0.69] mm, corresponding to door gap deviations of [0.09, 0.08, 0.07, 0.09, 0.08, 0.08, 0.07, 0.08, 0.08, 0.08, 0.07, 0.08] mm, with a maximum door gap deviation of 0.09 mm and a door gap dispersion deviation of 0.02 mm (0.09 minus 0.07). Simultaneously, the real-time current characteristics are a peak value of 4.3 A and a fluctuation amplitude of 0.8 A at the end, still falling within the closed reference current characteristic range. Since the maximum deviation of the door gap is not greater than the preset threshold for the door gap and the deviation of the door gap dispersion is not greater than the preset threshold for dispersion, and the real-time current characteristics meet the constraints of the closed reference current characteristics, the door controller determines that the closed state is in a sealed state, outputs a closed pass and allow departure interlock signal, and records the number of compensation insertions as 1. If the compensation insertion count reaches the preset number during the compensation re-judgment process but the door is still not in place, in this embodiment the preset number is set to 2, and the door controller outputs a prohibition departure interlock signal. If the motor drive current exceeds the preset compensation current upper limit during the compensation holding phase, the door controller stops the compensation insertion and outputs a stop closing and reverse opening control command, and simultaneously outputs a prohibition departure interlock signal to achieve control diversion for incomplete compensation and foreign object obstruction handling.

[0077] In summary, this application elevates the determination of the arrival status of subway doors during the termination phase from the traditional single-point position or locking signal criteria to a distributed quantitative authentication based on the door gap width distribution acquired by a TOF camera and using the closed reference door gap width distribution as a comparable reference. Furthermore, by aligning the door frame reference area and acquiring data within a steady-state time window, it establishes repeatable measurement conditions resistant to vibration and light disturbances, enabling door gap deviation characteristics to reflect non-uniform anomalies such as local non-insertion and local jamming. Simultaneously, it coordinates the determination of door gap deviation characteristics with the fluctuation characteristics of the motor drive current, forming mutually exclusive current splitting of the in-place sealing state, the non-in-place state, and the foreign object jamming state. This allows for differentiated actions at the control level that match the state, reducing the risk of missed detections and false releases when the locking signal is in place but the seal is not complete. It also avoids misjudging non-in-place as jamming, leading to frequent door reversals and reduced operational efficiency. Finally, by ensuring consistent output of interlocking signals, the authenticity of the door gap seal is directly incorporated into the departure safety link, improving the reliability and maintainability of the termination phase determination.

[0078] Based on another preferred embodiment described above, see [link to preferred embodiment]. Figure 2 As shown, this embodiment provides a subway door safety control device integrating a TOF camera, used in applying the above-described subway door safety control method integrating a TOF camera, including:

[0079] The door control controller is used to control the TOF camera to acquire multiple frames of depth images when the subway car doors are confirmed to be locked or closed, and to set multiple sampling lines in the door gap detection area. It is used to control the TOF camera to obtain the real-time door gap width distribution. It is used to obtain door gap deviation characteristics and to collect the fluctuation characteristics of the motor drive current to comprehensively determine the door closing status. It then outputs a control signal based on the door closing status.

[0080] A TOF camera, connected to a gating controller, is used to output depth images to the gating controller.

[0081] The motor drive current acquisition interface is connected to the gate controller and is used to provide the motor drive current to the gate controller.

[0082] The door operator control interface connects to the door controller and is used to receive stop closing, reverse opening, and low-speed compensation insertion control commands output by the door controller.

[0083] The interlocking signal output interface is connected to the gate controller and is used to output the departure permission interlocking signal and the departure prohibition interlocking signal.

[0084] Specifically, the TOF camera and motor drive current acquisition interface are connected to the gate controller as input terminals for depth images and motor drive current, respectively. The gate controller performs door gap detection and door closing status determination, and generates control outputs. The gate controller outputs control commands such as stop closing, reverse opening, and low-speed compensation insertion to the door operator control interface, which is connected to the door operator drive unit to drive the door movement. Simultaneously, the gate controller outputs a departure interlocking signal or a departure prohibition interlocking signal to the interlocking signal output interface, which is connected to the train control system to implement departure interlocking constraints, thus forming a closed-loop control link of sensing input, status determination, door operator control, and interlocking output.

[0085] Understandably, the determination of the subway door's arrival at the termination stage is upgraded from the traditional single-point position or locking signal criterion to a distributed quantitative authentication based on the door gap width distribution acquired by a TOF camera, with the closed reference door gap width distribution serving as a comparable reference. Furthermore, by aligning the door frame reference area and acquiring data within a steady-state time window, repeatable measurement conditions resistant to vibration and light disturbances are established, enabling door gap deviation characteristics to reflect non-uniform anomalies such as local non-insertion and local jamming. Simultaneously, the door gap deviation characteristics are collaboratively determined with the fluctuation characteristics of the motor drive current, forming mutually exclusive current separation between the in-place sealing state, the non-in-place state, and the foreign object jamming state. This allows for differentiated actions at the control level that match the state, reducing the risk of missed detections and false releases when the locking signal is in place but the seal is not complete. It also avoids misjudging non-in-place as jamming, leading to frequent door reversals and reduced operational efficiency. Finally, by ensuring consistent output of interlocking signals, the authenticity of the door gap seal is directly incorporated into the departure safety link, improving the reliability and maintainability of the termination stage determination.

[0086] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A subway car door safety control method integrating a TOF camera, characterized in that, include: When the door controller confirms that the subway door is in the locked / closed state, it controls the TOF camera to acquire multiple frames of depth images, aligns and filters them based on the door frame reference area, sets multiple sampling lines along the height direction of the door gap in the door gap detection area, extracts the door gap width, and forms a closed reference door gap width distribution. When the door closes and enters the termination phase, a depth image is acquired according to the steady-state time window and aligned with the door frame reference area to obtain the real-time door gap width distribution; Based on the door gap determination threshold at the termination stage, the real-time door gap width distribution is compared with the closed reference door gap width distribution to obtain door gap deviation characteristics, and the end fluctuation characteristics of the motor drive current are collected to comprehensively determine the closing state. The closing state includes the in-place sealing state, the in-place state, and the foreign object blocking state. When the door is in the closed and sealed state, a closing pass and departure permission interlock signal is output. When the door is not in the closed state, a low-speed compensation insertion control command is output and the termination stage judgment is repeated after compensation. If the door is still not in the closed state after exceeding the preset number of times, a prohibition on departure interlock signal is output. When the door is in a foreign object blocking state, output a stop closing and reverse opening control command and output a prohibition departure interlock signal. When obtaining door gap deviation characteristics, the following are included: Within each steady-state time window, the real-time door gap width is extracted along the multiple sampling lines and compared with the reference door gap width of the corresponding sampling line in the closed reference door gap width distribution to obtain the door gap deviation of each sampling line. Based on the door gap deviation, the maximum door gap deviation and the door gap dispersion deviation are determined. When the door gap deviation corresponding to a preset number of consecutive sampling lines along the door gap height direction exceeds the abrupt change threshold, the corresponding height segment is determined as the door gap abrupt change segment. The maximum door gap deviation and the maximum door gap dispersion deviation are used as the door gap deviation features. Forming the closed reference door gap width distribution includes: Geometric features of the door frame are extracted within the door frame reference area, and inter-frame registration is performed on multiple depth images to obtain an aligned depth image sequence. Multiple sampling lines are set along the height direction of the door gap in the door gap detection area, and the door gap width is extracted along each sampling line in each depth image. The door gap width of the same sampling line in the multiple depth images is taken as a preset statistical value and summarized to form the closed reference door gap width distribution.

2. The subway door safety control method integrating a TOF camera according to claim 1, characterized in that, When collecting the fluctuation characteristics of the motor drive current at the end, the following are included: The motor drive current during the termination phase is collected and a closed-loop reference current characteristic is formed, which includes the peak range at the end and the fluctuation range at the end. When the door closes and enters the termination phase, the motor drive current is collected according to the steady-state time window to form a real-time current characteristic, which includes the terminal peak value and the terminal fluctuation amplitude. The real-time current characteristic is compared with the closed reference current characteristic to obtain the terminal fluctuation characteristic of the motor drive current.

3. The subway door safety control method integrating a TOF camera according to claim 2, characterized in that, When comprehensively determining the door's closed status, the following are included: When the maximum deviation of the door gap is not greater than the preset door gap threshold and the door gap dispersion deviation is not greater than the preset dispersion threshold, and the real-time current characteristic falls within the end peak range and end fluctuation range corresponding to the closed reference current characteristic, the door closing state is determined to be the in-place sealing state. When the door gap abruptly changes and the peak value at the end exceeds the peak value range at the end or the fluctuation amplitude at the end exceeds the fluctuation range at the end, the closed state is determined to be a foreign object jamming state. Otherwise, the door closing state is determined to be in an incomplete state.

4. The subway door safety control method integrating a TOF camera according to claim 1, characterized in that, When the door is in the not-closed state The low-speed compensation insertion control command includes compensation insertion speed, compensation insertion stroke, and compensation holding duration; wherein, the door control controller determines the compensation insertion stroke based on the door gap deviation characteristics, sets the compensation insertion speed to a preset low speed that is less than the closing speed of the termination stage, controls the subway door to perform insertion according to the compensation insertion stroke, and maintains the drive according to the compensation holding duration after insertion is completed.

5. The subway door safety control method integrating a TOF camera according to claim 4, characterized in that, When performing the compensation insertion, it also includes: The gate controller collects the motor drive current during the compensation holding time and compares the collected motor drive current with the preset compensation current upper limit; when the motor drive current exceeds the preset compensation current upper limit, the gate controller stops the compensation insertion and outputs a stop closing and reverse opening control command.

6. The subway door safety control method integrating a TOF camera according to claim 5, characterized in that, The gate controller counts the number of compensation insertions and, after each compensation insertion, re-acquires the real-time door gap width distribution according to the steady-state time window and repeats the termination stage determination; when the number of compensation insertions reaches the preset number and the door closing state is still not in place, it outputs a prohibition on departure interlock signal.

7. The subway door safety control method integrating a TOF camera according to claim 1, characterized in that, It also includes anti-pinch linkage and leaning / hazard detection: Before the door enters the termination stage, a depth image of the hazard detection area at the doorway is acquired based on a TOF camera and a person or object target is identified. The determination is based on the minimum distance between the target and the door movement boundary and / or the intrusion amount. When the minimum distance is less than a preset anti-pinch threshold or the intrusion amount exceeds a preset intrusion threshold, a deceleration, stop closing the door and / or reverse opening control command is output. When the target is in a preset leaning detection area and the distance between the target and the door is less than a preset leaning threshold and the duration exceeds a preset duration threshold, the door opening command and / or closing command is suppressed, and the warning module is triggered to output a buzzer and voice prompt.

8. A subway door safety control device integrating a TOF camera, used to apply the subway door safety control method integrating a TOF camera as described in any one of claims 1-7, characterized in that, include: The door control controller is used to control the TOF camera to acquire multiple frames of depth images when it is confirmed that the subway car door is in the locked / closed state, and to set multiple sampling lines in the door gap detection area; Used to control the TOF camera to obtain the real-time door gap width distribution; This is used to obtain door gap deviation characteristics and collect the end fluctuation characteristics of motor drive current to comprehensively determine the door closing state; and output a control signal according to the door closing state. A TOF camera, connected to the gating controller, is used to output depth images to the gating controller; A motor drive current acquisition interface is connected to the gate controller and is used to provide the motor drive current to the gate controller. The door operator control interface is connected to the door controller and is used to receive the stop closing, reverse opening, and low-speed compensation insertion control commands output by the door controller. An interlocking signal output interface is connected to the gate controller and is used to output the departure permission interlocking signal and the departure prohibition interlocking signal.

Images