An intelligent monitoring method and system for train gap of a platform door based on AI video analysis

Abstract

This invention relates to an AI-based video analysis-based intelligent monitoring method and system for platform screen door train gaps, belonging to the field of platform safety monitoring. The method includes acquiring real-time video; when train door features are detected, monitoring the gap area between the train door and the platform screen door; when foreign object features are found in the gap area, retrieving a laser obstruction signal; if the laser obstruction signal exists, retrieving the platform screen door status and the train door status; if the platform screen door status and the train door status are inconsistent, intercepting the foreign object video and outputting the foreign object video and door linkage fault signal to the train driver; if the platform screen door status and the train door status are consistent, intercepting the foreign object video and outputting the foreign object video and foreign object alarm signal to the train driver; when there are no foreign object features in the gap area, no laser obstruction signal, and the platform screen door status and the train door status are consistent, outputting a departure prompt signal to prompt the train driver to depart. This invention has the effect of reducing the occurrence of gap safety accidents.

Description

Technical Field

[0001] This invention relates to the field of platform safety monitoring, and in particular to an intelligent monitoring method and system for platform screen door train gaps based on AI video analysis. Background Technology

[0002] Platform screen doors are protective devices installed at the edge of the platform. After the train arrives at the station and comes to a complete stop, the train doors and platform screen doors open simultaneously, with a fixed gap between them as designed. With the rapid development of urban rail transit, the safety of the gap between platform screen doors and trains has become a key aspect of operational safety management.

[0003] Currently, there are three main monitoring methods: First, staff visually inspect the train when the doors are closing; second, install through-beam sensors on platform doors or along the tracks to detect any obstructions; and third, install cameras on the platform to review the recordings later.

[0004] Regarding the aforementioned technologies, the monitoring methods mostly rely on manual visual inspection, simple sensors, and traditional cameras. These methods are prone to missed detections and are difficult to detect in real time the potential hazards of people, objects, or foreign objects getting stuck in gaps. They cannot accurately and proactively prevent safety accidents in gaps and there is still room for improvement. Summary of the Invention

[0005] To reduce the occurrence of safety accidents caused by gaps, this invention provides a method and system for intelligent monitoring of train gaps in platform screen doors based on AI video analysis.

[0006] In a first aspect, the present invention provides a method for intelligent monitoring of train gaps at platform screen doors based on AI video analysis, employing the following technical solution: A method for intelligent monitoring of train gaps at platform screen doors based on AI video analysis includes: Step S1: Acquire real-time video of each platform gate; Step S2: When the preset train door features are detected in the real-time video, monitor the gap area between the train door and the platform door in the real-time video based on the train door features and the preset platform door features; Step S3: When a preset foreign object feature appears in the gap area, retrieve the laser blocking signal corresponding to the gap area; Step S4: If the laser blocking signal exists, retrieve the platform door status and train door status; Step S40: If the platform door status and the train door status are inconsistent, the real-time video is captured according to the foreign object characteristics to obtain the foreign object video, and the foreign object video and the preset door linkage fault signal are output to the train driver. Step S41: If the platform door status and the train door status are consistent, the real-time video is captured according to the foreign object characteristics to obtain the foreign object video, and the foreign object video and the preset foreign object alarm signal are output to the train driver. Step S5: When there are no foreign objects in the gap area, no laser blocking signal, and the platform door status and train door status are consistent, output a preset departure prompt signal to prompt the train driver to depart.

[0007] By adopting the above technical solution, real-time video of the platform screen doors is first collected, and the characteristics of the train doors and platform screen doors are identified to determine the gap area. After detecting foreign objects in the gap, the system is verified by combining the detection with laser obstruction signals. Then, the opening and closing status of the two doors is retrieved, and linkage fault alarms or foreign object alarms are issued according to the differences in status. When there are no foreign objects, no obstructions, and the status of the two doors is synchronized, a departure prompt signal is issued. This solves the problem of not being able to monitor the gap area in real time for people, objects, or foreign objects stuck in it, and achieves the effect of reducing the occurrence of gap safety accidents.

[0008] Optionally, it also includes a method for handling situations where the platform door status and the train door status are consistent, the method including: Step S410: If both the platform door status and the train door status are in the preset fully closed state, acquire the three-dimensional point cloud, compare the three-dimensional point cloud with the preset three-dimensional background, and identify abnormal point clouds. Step S411: If the abnormal point cloud matches the preset human point cloud, extract the image of the foreign object from the video of the foreign object. Step S412: Platform doors include the first platform door and the second platform door; Step S413: Map the anomalous point cloud onto the foreign object image to obtain the anomalous image; Step S414: Analyze the proportion of abnormal point clouds located at the first platform gate and the proportion of abnormal point clouds located at the second platform gate in the abnormal image; Step S415: If the first proportion is 0, the second platform door is opened at a preset slow speed to obtain the resistance of the second platform door. If the resistance exceeds the preset safety threshold, the opening of the second platform door is stopped and a preset personnel guidance signal is output to the platform staff for guidance. If the resistance does not exceed the safety threshold, the opening continues until it is fully open and a personnel guidance signal is output. Step S416: If the second proportion is 0, the first platform door is opened at a preset slow speed to obtain the resistance of the first platform door. If the resistance exceeds the preset safety threshold, the opening of the first platform door is stopped and a preset personnel guidance signal is output to the platform staff for guidance. If the resistance does not exceed the safety threshold, the opening continues until it is fully open and a personnel guidance signal is output. Step S417: If both the first and second proportions are not 0, then execute the preset cross-door rescue process.

[0009] By adopting the above technical solution, when both doors close simultaneously, a three-dimensional point cloud is collected, and abnormal point clouds are identified by comparing them with the three-dimensional background. It is then determined whether the abnormal point cloud belongs to a human body. If it is a person, an image of the foreign object is captured, and the abnormal point cloud is mapped to obtain an abnormal image. The proportion of abnormal point clouds in the left and right platform doors is counted. Based on the difference in proportion, the doors are opened at low speed on one side and the resistance is monitored. If both are occupied, the corresponding cross-door rescue process is initiated, and a signal is pushed simultaneously for staff to carry out the rescue. This solves the problem that it is impossible to rescue people based on their location when both doors are closed, and achieves the effect of timely and proactive rescue of trapped people.

[0010] Optionally, if both the first and second proportions are not 0, the methods for executing the preset cross-door rescue process include: Step S4170: If the first proportion is greater than the second proportion, the second platform door is opened slowly to obtain the resistance of the second platform door. If the resistance exceeds the preset safety threshold, the opening of the second platform door is stopped and a personnel guidance signal is output to the platform staff. Step S4171: If the resistance does not exceed the safety threshold, open the second platform door to full opening and output a preset self-rescue guidance signal to prompt the trapped personnel to use the space created by the opening of the second platform door to adjust their posture and escape. Step S4172: If the first proportion is less than the second proportion, the first platform door is opened slowly to obtain the resistance of the first platform door. If the resistance exceeds the preset safety threshold, the opening of the first platform door is stopped and a personnel guidance signal is output to the platform staff. Step S4173: If the resistance does not exceed the safety threshold, continue to open the first platform door until it is fully open, and output a preset self-rescue guidance signal to prompt the trapped personnel to use the space created by the opening of the first platform door to adjust their posture and escape. Step S4174: If the first proportion is equal to the second proportion, then output the preset personnel assistance signal to the platform staff.

[0011] By adopting the above technical solution, if one side has a higher proportion, the opposite door will open slowly, the resistance will be monitored, and if the resistance exceeds the limit, the door will stop opening and personnel will be notified for rescue; if the resistance is normal, the door will be fully opened to guide the trapped person to escape on their own; if the proportions of both sides are equal, no automatic door opening command will be issued, and a personnel assistance signal will be directly output to hand over the rescue to the staff on site. This solves the problem that trapped people cannot be directly rescued when they are between two doors, and achieves the effect of differentiated rescue based on the proportions of the two doors.

[0012] Optional, also includes: Step S4110: Human point cloud includes adult point cloud and child point cloud; Step S4111: If the abnormal point cloud matches the adult point cloud, then execute steps S411 to S417 and steps S4170 to S4174. Step S4112: If the abnormal point cloud matches the child point cloud, then extract the platform door number from the abnormal image; Step S4113: Output the platform door number and the preset child trapped signal to the platform staff, instructing the platform staff to carry out the rescue.

[0013] By adopting the above technical solution, the point cloud categories of adults and children are distinguished. Adults follow the existing rescue process, while children report the platform door number and send a child trapped signal to prompt personnel to come to the scene for rescue. This solves the problem that using the same rescue plan for adults and children may cause injury to children, and achieves the effect of real-time targeted rescue for children and adults respectively.

[0014] Optionally, methods for outputting the platform door number and a preset child trapped signal to platform staff to instruct them to carry out rescue operations include: Step S41130: If the first proportion is 0, the first platform door is opened slowly to obtain the resistance of the first platform door. If the resistance exceeds the preset safety threshold, the opening of the first platform door is stopped and the platform door number and the preset child trapped signal are output to the platform staff, instructing the platform staff to enter the gap from the side of the opened first platform door to rescue the trapped child. Step S41131: If the second proportion is 0, the second platform door is opened slowly to obtain the resistance of the second platform door. If the resistance exceeds the preset safety threshold, the opening of the second platform door is stopped and the platform door number and the preset child trapped signal are output to the platform staff, instructing the platform staff to enter the gap from the side of the opened second platform door to rescue the trapped child. Step S41132: If both the first and second proportions are not 0, output the platform door number and the preset child trapped signal to the platform staff, instructing the platform staff to carry out the rescue.

[0015] By adopting the above technical solution, when a child is stuck on one side, the door on the opposite side is opened at low speed and the resistance is monitored. If there is an abnormality, the door is stopped and an alarm is triggered. When a child occupies the area of ​​the first platform door and the second platform door at the same time, a child trapped signal is sent directly to notify personnel to come to the scene for rescue, which solves the problem that the door cannot be opened directly when a child is trapped in a gap area.

[0016] Optionally, a method for inspecting train doors may also be included, which includes: Step S4100: Retrieve the in-car video corresponding to the platform door and the corresponding train door; Step S4101: Monitor the train door closure line based on the characteristics of the train door and the in-car video; Step S4102: Extract the initial closed line from the train door features; Step S4103: Compare the train door closing line with the initial closing line. If the train door closing line and the initial closing line are inconsistent, control the train door to open to the preset release gap and output a preset departure prompt signal to the trapped personnel.

[0017] By adopting the above technical solution, the in-vehicle video monitoring of the train door closure line is retrieved and compared with the initial closure line. If there is a deviation, the door is opened to release the gap and a departure prompt signal is output to prompt the personnel to leave, which solves the problem of passengers being unable to move due to clothing, bag straps and other small objects being caught when the train door closes.

[0018] Optionally, it also includes a method for handling inconsistencies between the platform door status and the train door status, the method including: Step S400: Acquire a 3D point cloud and compare it with a preset 3D background to identify abnormal point clouds; Step S401: When the abnormal point cloud is a preset human point cloud, if the platform door is in a preset unclosed state and the train door is in a preset fully closed state, then control the platform door to fully open and output a preset departure prompt signal. Step S402: If the platform door is in a preset fully closed state and the train door is in a preset unclosed state, then control the train door to fully open and output a preset departure prompt signal.

[0019] By adopting the above technical solution, when the door status is inconsistent, after identifying that a person is trapped, the door on the side that is not closed will be fully opened and an exit prompt signal will be output to prompt the trapped person to leave, thus solving the problem of people being trapped due to inconsistent platform door status and train door status.

[0020] Optionally, it also includes a method for checking whether the train door is in a preset fully closed state or the platform door is in a preset fully closed state, the method comprising: Step S403: Define the train door or platform door that is in a preset fully closed state as a closed door; Step S404: Obtain the pressure at the edge of the closed door; Step S405: If the pressure at the edge of the door exceeds the preset normal closing pressure, it is determined that there is a foreign object trapped in the closing door; Step S406: Control the closing door to open to the preset release gap, and output the departure prompt signal again.

[0021] By adopting the above technical solution, the pressure on the edge of the closed door is detected, and if the pressure is abnormal, it is determined that something is stuck. The closed door is then controlled to open to release the gap, and a departure prompt signal is output again. This solves the problem of people being unable to leave when the door is stuck due to inconsistent door conditions.

[0022] Optionally, it also includes a method for handling abnormal point clouds that are not preset human point clouds, the method including: Step S4010: If the platform door is in the preset unclosed state and the train door is in the preset fully closed state, then control the platform door to open fully and locate the owner of the items at the platform door based on the gap area and real-time video. Step S4011: If the platform door is in the preset fully closed state and the train door is in the preset unclosed state, then control the train door to open fully, retrieve the in-car video corresponding to the platform door and the corresponding train door, and find the owner of the item at the train door in the in-car video. Step S4012: If the owner of the item exists, output a preset passenger pick-up signal; Step S4013: If the owner of the item does not exist, output a preset item removal signal to the platform staff.

[0023] By adopting the above technical solution, after identifying the objects stuck in the gap as items, the corresponding doors are opened according to the opening and closing status of the two doors. The owner of the items is traced by combining real-time video or in-vehicle video. If the owner is found, a voice prompt is given to retrieve the items. If no one claims them, a signal is sent to hand them over to the staff for cleaning, which solves the problem of items being stuck in the gap area and not being claimed.

[0024] Secondly, this invention provides an intelligent monitoring system for train gaps at platform screen doors based on AI video analysis, employing the following technical solution: An AI-based video analysis-based intelligent monitoring system for train gaps at platform screen doors includes: The acquisition module is used to acquire real-time video, 3D point cloud, resistance, and pressure at the edge of the door. The memory is used to store the program of the AI ​​video analysis-based intelligent monitoring method for train gaps at platform screen doors, as described above. The processor loads and executes programs from memory.

[0025] In summary, the present invention has at least one of the following beneficial technical effects: It solves the problem of not being able to monitor people, objects, or foreign objects trapped in gaps in real time, thus reducing the occurrence of safety accidents in gaps; This solved the problem of being unable to conduct rescues based on the location of people when both doors are closed, achieving the effect of timely and proactive rescue of trapped personnel. Attached Figure Description

[0026] Figure 1 This is a flowchart of an AI-based video analysis-based intelligent monitoring method for train gaps at platform screen doors, as described in this application.

[0027] Figure 2This is a schematic diagram of a scenario in which the platform door and train door are not in the same state as in this application embodiment.

[0028] The locations indicated by the numbers in the attached diagrams are as follows: 1. Platform door; 2. Train door. Detailed Implementation

[0029] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.

[0030] This invention discloses an intelligent monitoring method for train gaps at platform screen doors based on AI video analysis. (Refer to...) Figure 1 A method for intelligent monitoring of train gaps at platform screen doors based on AI video analysis includes: Step S1: Obtain real-time video of each platform gate 1.

[0031] Platform screen door 1 refers to a sliding door installed on the edge of a subway or light rail station platform, primarily used to separate the track area from the waiting area. Platform screen door 1 is obtained by directly reading the number and physical location information of each platform screen door from the station's existing platform screen door control system. Based on this data, a mapping relationship between platform screen door 1 and the corresponding high-definition camera is established in the existing monitoring system.

[0032] Real-time video refers to the real-time video of the area where platform door 1 is located. The video range covers the entire platform door 1 and the gap area after the train doors are closed. It is collected by a high-definition camera installed above platform door 1 or on the platform top.

[0033] Step S2: When the preset train door features are detected in the real-time video, monitor the gap area between train door 2 and platform door 1 in the real-time video based on the train door features and the preset platform door features.

[0034] Train door features refer to the identifiable characteristics of train doors in video images, including the rectangular outline of the door glass area, the straight-line features of the door edge, the specific shape of the emergency unlocking device, and the parallel line pair formed by the door and platform door 1 when the door is closed. Train door features are generated by collecting historical video samples from multiple train stops by those skilled in the art. The train door area in each frame of the sample is manually labeled, including the rectangular outline of the door glass, the straight-line features of the door edge, the specific shape of the emergency unlocking device, and the parallel line pair formed by the door and platform door frame when the door is fully closed. The labeled door area images are collected to form a train door feature dataset. During monitoring, real-time video frames are compared with this feature dataset. When the similarity between an image area in the real-time video and a certain type of sample in the dataset exceeds a preset first similarity threshold, a train door feature is detected, and the position coordinates of the door in the image are output. The first similarity threshold is predetermined by those skilled in the art through sampling tests based on the actual scenario. The requirement is that the value can stably identify the train door area with an acceptable false detection rate.

[0035] Platform screen door features refer to the identifiable features of platform screen door 1 in video images, including the reflective texture of the sliding door's glass panel, the straight lines of the sealing strip along the door's edge, and safety warning signs or numbered stickers on the door. Platform screen door features are determined by a person skilled in the art through the collection of historical video samples of platform screen door 1 under different lighting conditions. Areas such as the reflective texture of the glass, the straight lines of the sealing strip along the door's edge, and the safety warning signs or numbered stickers on the door are manually marked. The marked images are then collected to form a platform screen door feature dataset. During monitoring, real-time video frames are compared with the platform screen door feature dataset. When the similarity exceeds a preset second similarity threshold, the platform screen door feature is detected, and the position coordinates of platform screen door 1 in the image are output. The second similarity threshold is obtained in the same way as the first similarity threshold.

[0036] Train door 2 refers to the sliding door installed on the side of the subway or urban rail train car, through which passengers board and alight. Train door 2 corresponds to platform door 1, and is the door on the side that forms a gap area with platform door 1 when the train is stopped; it is an inherent door device of the train.

[0037] The gap area refers to the narrow space between the edges of platform screen door 1 and train door 2 when the train has come to a complete stop and both doors are closed. Its physical width is typically 100 mm to 250 mm, and its height is the same as the door height. The intermediate area between the detected train door and platform screen door boundaries is calculated. Specifically, a horizontal strip-shaped area is taken between the right vertical edge of train door 2 and the left vertical edge of platform screen door 1. If platform screen door 1 is open, the area between adjacent edges of the two doors is taken. Using pre-calibrated camera intrinsic and extrinsic parameters, this area is mapped from the image coordinate system to actual physical space coordinates.

[0038] When the preset train door features are detected in the real-time video, it means that the train has entered the station and has come to a complete stop. Therefore, it is necessary to determine the gap area between platform door 1 and train door 2 for subsequent monitoring.

[0039] If the preset train door features are not detected in the real-time video, it means that the train has not yet entered the station and no monitoring is required.

[0040] Step S3: When a preset foreign object feature appears in the gap area, retrieve the laser blocking signal corresponding to the gap area.

[0041] Foreign object features refer to characteristics in a video image of a gap area that significantly differ from the normal background. Their shape and texture typically conform to the features of a part of the human body (such as an arm or head), the overall outline of a child, or the shape of a large piece of luggage. The normal background refers to a gap without passengers or luggage. The foreign object feature is obtained as follows: the fixed position of the gap area in the image is determined by the combined features of the train door and platform door; during runtime, a sub-image of the gap area is extracted from the real-time video, the sub-image is converted to grayscale, and then binarized using adaptive thresholding to separate the foreground object (i.e., the foreign object) from the background gap; connected component analysis is performed on the binary image, and the area of ​​each connected component is calculated; if the area of ​​any connected component is greater than a preset area threshold, it is determined that a foreign object feature exists in the gap area, and the presence of the foreign object and its outline coordinates are output. The area threshold is determined experimentally by those skilled in the art based on the camera resolution and the actual size of the gap area.

[0042] The laser obstruction signal is a signal indicating whether there is an object obstructing the gap area. It is emitted by the lidar installed at both ends of the gap area below or to the side of each platform door 1 (e.g., between the bottom of platform door 1 and the train floor).

[0043] When a preset foreign object feature appears in the gap area, it indicates that a person or object has entered the gap area. It is necessary to check whether there is a laser obstruction signal in the gap area to verify whether a foreign object feature has actually appeared in the gap area.

[0044] If no pre-defined foreign object features are found in the gap area, it means that no person or object has entered the gap area, and no additional operation is required.

[0045] Step S4: If the laser blocking signal exists, retrieve the platform door status and train door status.

[0046] The platform door status refers to the current operating state of platform door 1. The train door status refers to the current operating state of train door 2. The status includes at least a fully closed state (fully closed and locked) and an unclosed state (closing, opening, or malfunction). The platform door status is obtained by reading the "door closed and locked" dry contact signal and the "door open" limit switch signal from each door unit through the platform door control system. The train door status is obtained by acquiring the status frame reported by the train door controller through the train arrival information system or wireless communication. The status frame contains the "closed and locked" flag for each door in each car.

[0047] If the laser blocking signal is present, it indicates that there is indeed a person or object in the gap area. Therefore, at this time, the status of the platform door and the train door should be retrieved, and the subsequent operation should be determined based on whether the two statuses are synchronized.

[0048] If the laser blocking signal is not present, it indicates an error in the video analysis. In this case, the video should be reviewed, and multiple frames should be retrieved for tracking and confirmation. If foreign object features are present in multiple consecutive frames, the video should still be treated as if a foreign object is present.

[0049] Step S40: If the platform door status and the train door status are inconsistent, the real-time video is captured according to the foreign object characteristics to obtain the foreign object video, and the foreign object video and the preset door linkage fault signal are output to the train driver.

[0050] Foreign object video refers to a video segment containing foreign object characteristics, which is extracted from real-time video. Specifically, the start time is the moment when the foreign object characteristics first appear, and the end time is the moment when the door linkage fault signal or foreign object alarm signal is output.

[0051] The door linkage fault signal refers to an alarm signal that prompts the driver that there is a door asynchrony fault and that the train needs to be stopped immediately to take measures. This signal is preset by those skilled in the art and is triggered when there are foreign objects in the gap area, laser obstruction signals are present, and the status of the platform door and the train door are inconsistent. It sends an alarm signal containing fault information to the train driver's cab display screen, specifically the fault type code of door asynchrony, the time of the fault occurrence, and the corresponding foreign object video path.

[0052] Reference Figure 2If the status of the platform door and the train door are inconsistent, it indicates that the linkage between platform door 1 and train door 2 is faulty. The video of the foreign object and the linkage fault signal are sent to the driver, indicating that there is a foreign object in the gap area and that the linkage between platform door 1 and train door 2 is faulty. It is necessary to stop the train and release the trapped personnel or items, and to check and deal with the doors.

[0053] Step S41: If the platform door status and the train door status are consistent, the real-time video is captured according to the foreign object characteristics to obtain the foreign object video, and the foreign object video and the preset foreign object alarm signal are output to the train driver.

[0054] The foreign object alarm signal is an alarm signal that prompts the driver to immediately stop the train due to the presence of a foreign object in the gap area. It is pre-set by those skilled in the art. It is triggered when there are foreign object characteristics in the gap area, a laser blocking signal is present, and the platform door status and train door status are consistent. The alarm signal, which contains fault information, is sent to the train driver's cab display screen. Specifically, it includes the fault type code of the foreign object in the gap, the time of the fault occurrence, and the corresponding video path of the foreign object.

[0055] If the status of the platform door and the train door are the same, it means that the door linkage between platform door 1 and train door 2 is normal. At this time, a foreign object video and a foreign object alarm signal are output to remind the driver that there is a person or object in the gap area, and the train needs to be stopped and the foreign object in the gap area needs to be removed.

[0056] Step S5: When there are no foreign objects in the gap area, no laser blocking signal, and the platform door status and train door status are consistent, output a preset departure prompt signal to prompt the train driver to depart.

[0057] The departure warning signal is a safety signal indicating that the vehicle can depart after the gap area is free of foreign objects, the laser beam is unobstructed, and both doors are fully closed and locked. The departure warning signal is pre-set by those skilled in the art, and when all the above conditions are met, it is displayed on the driver's cab screen with both sound and light to indicate that departure is possible.

[0058] When there are no foreign objects in the gap area, no laser blocking signal, and the status of the platform door and the train door are consistent, it means that there are no foreign objects in the gap area and the laser blocking signal has been verified. The platform door 1 and the train door 2 are linked normally and are in a fully closed and locked state. At this time, the conditions for departure are met, so a departure prompt signal is issued to the driver to indicate that the train can depart.

[0059] This also includes a method for handling situations where the platform door status and the train door status are the same, and this method includes: Step S410: If both the platform door status and the train door status are in the preset fully closed state, acquire the three-dimensional point cloud, compare the three-dimensional point cloud with the preset three-dimensional background, and identify abnormal point clouds.

[0060] The fully closed state refers to the state in which platform door 1 and train door 2 are completely closed and locked. This state is preset by those skilled in the art, and the specific determination method refers to the method of obtaining platform door 1 and train door 2 in step S4.

[0061] A 3D point cloud refers to the three-dimensional coordinates of the gap area and its surrounding surface. Each point contains X (along the track direction), Y (horizontal direction perpendicular to the platform edge), and Z (vertical height). The 3D point cloud is generated by installing a solid-state LiDAR (such as Ouster OS1) above or to the side of each platform door 1, covering the entire gap area between platform door 1 and train door 2. When the train stops and the doors are fully closed, the LiDAR scans and outputs raw point cloud data, which is then transmitted to a server. The server divides the raw point cloud data into a 1 cm side-length cube grid, retaining only one center point within each grid to reduce data volume while maintaining shape. A statistical filtering method is then used to calculate the average distance within a certain range around each center point (e.g., a neighborhood within a 5 cm radius). If the average distance of a point exceeds twice the overall mean standard deviation, it is considered an isolated noise point and needs to be deleted. Finally, the pre-processed 3D point cloud is obtained.

[0062] The three-dimensional background refers to the baseline three-dimensional point cloud of the gap area and its surrounding environment in the absence of any foreign objects. The three-dimensional background is obtained by a person skilled in the art during a period when multiple trains are normally stopped, no passengers are staying, and the gap area is completely empty (if it is after the trains stop running at night). The background point cloud data is obtained by scanning with the installed lidar, and the background point cloud data is processed according to the three-dimensional point cloud preprocessing method described above to obtain the three-dimensional background.

[0063] Anomaly point clouds refer to a set of point clouds where there is a significant difference between the 3D point cloud and the 3D background. These are non-background objects (such as human figures or luggage) appearing in the gap regions. Spatial registration is performed between the 3D point cloud and the 3D background. For each point in the 3D point cloud, the nearest 1cm cube grid in the 3D background is found, and the Euclidean distance between them is calculated. If this distance exceeds a preset anomaly threshold (e.g., 5cm), the center point is marked as an anomaly. All anomaly points constitute the initial anomaly point cloud. Then, referring to statistical filtering methods in 3D point clouds, the average distance from each point to all neighboring points within a 5cm radius is calculated. If this average distance exceeds twice the standard deviation of the overall average distance, the point is considered an isolated noise point and is deleted. The final retained point cloud is the anomaly point cloud.

[0064] If both the platform door and train door are in the preset fully closed state, it means that platform door 1 and train door 2 are completely closed. People or objects in the gap area are trapped between the two doors. At this time, the 3D point cloud is obtained and compared with the 3D background to obtain the abnormal point cloud. The abnormal point cloud is used to determine whether there is a person or an object in the gap area.

[0065] Step S411: If the abnormal point cloud matches the preset human point cloud, extract the image of the foreign object from the video of the foreign object.

[0066] Human point cloud refers to a three-dimensional point cloud that conforms to the geometric and motion characteristics of the human body and can represent the human torso, limbs, or head. The human point cloud is obtained by those skilled in the art in advance by collecting three-dimensional point cloud data of multiple volunteers of different heights and body types in different postures (such as standing, turning to the side, and squatting) in the gap area of ​​the platform. After preprocessing these data, the mean point cloud of all samples is taken as the human point cloud.

[0067] The method for determining whether an abnormal point cloud matches a preset human point cloud is as follows: The abnormal point cloud is registered with the human point cloud, for example, using the Iterative Closest Point (ICP) algorithm to calculate the shape similarity between the two point clouds (such as the average Euclidean distance or overlap rate between the point clouds). If the similarity exceeds a preset similarity threshold (such as an average distance of less than 2 cm or an overlap rate greater than 80%), the abnormal point cloud is determined to match the human point cloud. The similarity threshold is preset by those skilled in the art based on actual scenarios through sampling tests.

[0068] A foreign object image refers to an image with foreign object characteristics, typically the frame with the clearest outline of the foreign object in a foreign object video. In each frame of the foreign object video, background subtraction is performed on the gap area, specifically referring to the method for obtaining foreign object characteristics in step S3, to calculate the pixel area of ​​the foreground object, and the frame with the largest pixel area is selected as the foreign object image.

[0069] If the abnormal point cloud matches the preset human point cloud, it means that a person is trapped. First, extract the image of the foreign object from the video of the foreign object, and use it for subsequent analysis to obtain a method to rescue the trapped person.

[0070] If the abnormal point cloud does not match the preset human body point cloud, it means that the object is trapped. In this case, in addition to outputting foreign object video and foreign object alarm signals to the train driver, the foreign object image and the platform door number corresponding to the foreign object video camera are sent to the platform staff. The platform door 1 corresponding to the platform door number is then opened, and the platform staff remove the trapped object.

[0071] Step S412: Platform door 1 includes a first platform door and a second platform door.

[0072] The first platform screen door refers to the left sliding door in each platform screen door unit consisting of two sliding doors, one on the left and one on the right. The second platform screen door refers to the right sliding door in each platform screen door unit consisting of two sliding doors, one on the left and one on the right. Those skilled in the art, based on the numbering mapping relationship of the platform screen door control system, pre-assign left and right doors to each door unit, corresponding the left door to the first platform screen door and the right door to the second platform screen door.

[0073] Step S413: Map the abnormal point cloud onto the foreign object image to obtain the abnormal image.

[0074] An anomaly image is an image generated by projecting anomaly point clouds onto an image of a foreign object, thus accurately showing the location of the foreign object. First, the LiDAR and camera are jointly calibrated to obtain the rotation matrix and translation vector between them, as well as the camera's intrinsic parameter matrix. For each point in the anomaly point cloud, the 3D coordinates of the point cloud are projected onto the pixel coordinate system of the foreign object image using the aforementioned parameters, yielding the pixel coordinates. Color markers (such as red dots) are then drawn at the corresponding pixel positions in the foreign object image. The convex hull contour of all points is drawn by connecting the outer contours formed by these color markers, resulting in the final anomaly image.

[0075] Step S414: Analyze the proportion of abnormal point clouds located at the first platform gate and the proportion of abnormal point clouds located at the second platform gate in the abnormal image.

[0076] The first percentage refers to the proportion of pixels in the anomalous point cloud located within the first platform screen door area of ​​the anomalous image, relative to the total number of pixels in the first platform screen door area. The second percentage refers to the proportion of pixels in the anomalous point cloud located within the second platform screen door area of ​​the anomalous image, relative to the total number of pixels in the second platform screen door area. First, determine the pixel area occupied by each platform screen door 1 (first platform screen door and second platform screen door) in the image; this is all the pixels inside the polygon enclosed by the four vertices of the door frame. Then, based on the projection points of all anomalous point clouds in the anomalous image, calculate the number of projection points falling into the pixel area of ​​the first platform screen door and the number falling into the pixel area of ​​the second platform screen door. Finally, divide the number of projection points falling into the first platform screen door area by the total number of pixels in the first platform screen door area to obtain the first percentage, and divide the number of projection points falling into the second platform screen door area by the total number of pixels in the second platform screen door area to obtain the second percentage. For example, if the first platform screen door area has 10,000 pixels, and 200 anomalous point cloud projection points fall within this area, then the first percentage is 2%.

[0077] Step S415: If the first proportion is 0, the second platform door is opened at a preset slow speed to obtain the resistance of the second platform door. If the resistance exceeds the preset safety threshold, the opening of the second platform door is stopped and a preset personnel guidance signal is output to the platform staff for guidance. If the resistance does not exceed the safety threshold, the door is opened until it is fully open and a personnel guidance signal is output.

[0078] Slow speed refers to the speed at which the motor drives the door to move at a low speed, usually 20% of the normal opening and closing speed, and is preset by those skilled in the art based on the door type and on-site safety requirements.

[0079] Resistance refers to the torque value output by the motor during the movement of the door. It can reflect the mechanical resistance encountered by the door. The current value of the drive motor is collected in real time by the door controller, and then the current is converted into torque through the motor torque constant.

[0080] The safety threshold refers to the upper limit of permissible resistance. According to industry standards (such as GB / T 33667), the anti-pinch force of platform screen door 1 should not exceed 150N. Corresponding to the motor torque, during commissioning, a person skilled in the art should apply a force of 150N to the edge of the door using a standard force gauge, and record the torque value reported by the door controller. This torque value is then multiplied by a safety factor of 0.8 to obtain the safety threshold.

[0081] A personnel guidance signal is a signal that prompts platform staff to guide and rescue trapped personnel. This signal is pre-set by those skilled in the art and is sent via a wireless network to the staff terminal of the corresponding platform. It contains the platform door number where the trapped personnel are located, the recommended platform door 1 (first or second), the path to the abnormal image file, and text prompts for rescue suggestions (such as "Please enter through the gap from the right-hand door to guide the passenger to safety").

[0082] If the percentage of the first platform screen door is 0, it means that the trapped person is completely inside the gap area of ​​the second platform screen door. Therefore, the second platform screen door should be opened slowly. During this period, it is necessary to continuously obtain resistance from the second platform screen door to prevent it from being forcibly opened and injuring the trapped person.

[0083] If the resistance exceeds the preset safety threshold, it indicates that the door may be squeezing a person. At this time, stop driving the second platform door to open and send a personnel guidance signal to the platform staff, who will then guide the trapped person to safety.

[0084] If the resistance does not exceed the safety threshold, it means that the door has not yet squeezed the person. It can continue to open until it is fully open, and then send a personnel guidance signal to the platform staff, who will then guide the trapped person and carry out follow-up work.

[0085] Step S416: If the second percentage is 0, the first platform door is opened at a preset slow speed to obtain the resistance of the first platform door. If the resistance exceeds the preset safety threshold, the opening of the first platform door is stopped and a preset personnel guidance signal is output to the platform staff for guidance. If the resistance does not exceed the safety threshold, the opening continues until it is fully open and a personnel guidance signal is output.

[0086] If the second percentage is 0, it means that the trapped person is completely in the gap area inside the first platform door. Therefore, the first platform door should be opened slowly. During this period, it is necessary to continuously obtain resistance from the first platform door to prevent it from being forcibly opened and injuring the trapped person.

[0087] If the resistance exceeds the preset safety threshold, it indicates that the door may be squeezing a person. At this time, stop driving the first platform door to open and send a personnel guidance signal to the platform staff, who will then guide the trapped person to safety.

[0088] Step S417: If both the first and second proportions are not 0, then execute the preset cross-door rescue process.

[0089] The cross-door rescue procedure refers to a set of rescue operations performed when an abnormal point cloud is simultaneously located in the first platform screen door area and the second platform screen door area. For details, please refer to the detailed descriptions of steps S4170 to S4174.

[0090] If neither the first nor the second percentage is 0, it means that the trapped person is between the first and second platform doors. In this case, the cross-door rescue procedure should be executed, and a special rescue operation should be carried out.

[0091] If both the first and second proportions are not 0, the methods for executing the preset cross-door rescue process include: Step S4170: If the first proportion is greater than the second proportion, the second platform screen door is opened slowly to obtain the resistance of the second platform screen door. If the resistance exceeds the preset safety threshold, the opening of the second platform screen door is stopped and a personnel guidance signal is output to the platform staff.

[0092] If the first proportion is greater than the second proportion, it means that the gap area corresponding to the first platform door is larger than the gap area corresponding to the second platform door. Therefore, you can try to open the second platform door slowly and continuously monitor the resistance of the second platform door.

[0093] If the resistance exceeds the preset safety threshold, it means that the second platform door can no longer be opened, otherwise it will crush the trapped personnel. Therefore, the opening of the second platform door is stopped, and a personnel guidance signal is sent to the platform staff so that the platform staff can rescue the trapped personnel.

[0094] Step S4171: If the resistance does not exceed the safety threshold, open the second platform door to full opening and output a preset self-rescue guidance signal to prompt the trapped personnel to use the space created by the opening of the second platform door to adjust their posture and escape.

[0095] Self-rescue guidance signals refer to the voice and visual prompts that guide trapped personnel to independently adjust their posture and escape using the space of the opened platform door 1. These signals are pre-set by those skilled in the art. The signals are played through speakers above the platform door 1, such as "The door is open, please slowly move out from the side of the open platform door," while a green arrow and text prompts are displayed on the passenger information display screen.

[0096] The space created by the opening of the second platform screen door refers to the gap area formed between the second platform screen door and the train door after the second platform screen door is opened slowly and fully, allowing people to pass through or adjust their posture. The size of this space increases with the degree of opening of the second platform screen door. Trapped people can use this space and the freed parts of their body located in the second platform screen door area to exert force to move their body or limbs out from the side that was not compressed, or use this space to adjust their posture to eventually escape.

[0097] If the resistance does not exceed the safety threshold, it means that the second platform door can continue to be opened until it is fully open. At this time, a self-rescue guidance signal is output to guide the trapped personnel to escape from the space formed by the opening of the second platform door.

[0098] Step S4172: If the first proportion is less than the second proportion, the first platform door is opened slowly to obtain the resistance of the first platform door. If the resistance exceeds the preset safety threshold, the opening of the first platform door is stopped and a personnel guidance signal is output to the platform staff.

[0099] If the first proportion is less than the second proportion, it means that the gap area corresponding to the second platform door is larger than the gap area corresponding to the first platform door. Therefore, we can try to open the first platform door slowly and continuously monitor the resistance of the first platform door.

[0100] If the resistance exceeds the preset safety threshold, it means that the first platform door can no longer be opened, otherwise it will crush the trapped personnel. Therefore, the opening of the first platform door is stopped, and a personnel guidance signal is sent to the platform staff so that the platform staff can rescue the trapped personnel.

[0101] Step S4173: If the resistance does not exceed the safety threshold, continue to open the first platform door until it is fully open, and output a preset self-rescue guidance signal to prompt the trapped personnel to use the space created by the opening of the first platform door to adjust their posture and escape.

[0102] The space created by the opening of the first platform screen door refers to the gap area formed between the first platform screen door and the train door after the first platform screen door is opened slowly and fully, allowing people to pass through or adjust their posture. The size of this space increases with the degree of opening of the first platform screen door. Trapped people can use this space and the freed parts of their body located in the first platform screen door area to exert force to move their body or limbs out from the side that was not compressed, or use this space to adjust their posture to eventually escape.

[0103] If the resistance does not exceed the safety threshold, it means that the first platform door can continue to be opened until it is fully open. At this time, a self-rescue guidance signal is output to guide the trapped personnel to escape from the space formed by the opening of the first platform door.

[0104] Step S4174: If the first proportion is equal to the second proportion, then output the preset personnel assistance signal to the platform staff.

[0105] A personnel assistance signal is a signal that requests platform staff to come to the scene for manual rescue. It is triggered when the first proportion equals the second proportion and is preset by those skilled in the art. The personnel assistance signal sends information to the platform staff's terminal, such as "The trapped person is located between the two doors. Please use the special key to open the emergency door or sliding door from the platform side to carry out the rescue."

[0106] If the first proportion equals the second proportion, it means that the trapped person is trapped between the first platform door and the second platform door. At this time, there is a high probability that opening platform door 1 will crush the trapped person. Therefore, a personnel assistance signal is sent to the platform staff, prompting them to carry a special key to open the emergency door or sliding door from the platform side to carry out the rescue and free the trapped person.

[0107] This also includes: Step S4110: Human point cloud includes adult point cloud and child point cloud.

[0108] Adult point clouds refer to three-dimensional point clouds that conform to the geometric characteristics of an adult human body. Those skilled in the art first collect three-dimensional point cloud samples from multiple adult volunteers of different heights. After clustering, they statistically determine the characteristic range of height between 150cm and 200cm, projected area greater than 0.3m², and aspect ratio between 2:1 and 4:1. This range is used as the threshold for determining adult point clouds. During runtime, abnormal point clouds are clustered and their geometric characteristics are calculated. If they simultaneously meet the above ranges, they are determined to be adult point clouds.

[0109] Children's point clouds refer to three-dimensional point clouds that conform to the geometric characteristics of children's bodies. Referring to the calibration method of adult point clouds, point cloud samples were collected from multiple children volunteers of different heights (80cm to 140cm). In the same way, feature ranges with heights between 80cm and 140cm, projected areas greater than 0.1m² and less than 0.3m², and aspect ratios between 1.2:1 and 3:1 were obtained. These ranges were used as the judgment thresholds for children's point clouds. During operation, if the geometric characteristics of abnormal point clouds simultaneously meet the above ranges, they are judged as children's point clouds.

[0110] Step S4111: If the abnormal point cloud matches the adult point cloud, then execute steps S411 to S417 and steps S4170 to S4174.

[0111] If the abnormal point cloud matches the adult point cloud, it indicates that the trapped person is an adult. Then, execute steps S411 to S417 and steps S4170 to S4174 to determine the proportion of the abnormal point cloud occupying the first platform door area and the proportion of the abnormal point cloud occupying the second platform door area. Compare these two and execute the corresponding rescue operation.

[0112] Step S4112: If the abnormal point cloud matches the child point cloud, then extract the platform door number from the abnormal image.

[0113] The platform door number refers to the number of each platform door 1. The method of obtaining the platform door 1 is the same as that in step S1.

[0114] If the abnormal point cloud matches the point cloud of a child, it means that the trapped person is a child. At this time, the rescue procedure for adults cannot be followed. Children need to be cared for by adults, and children's bones are more fragile than adults. Opening platform door 1 may frighten or crush the child. Therefore, the platform door number should be determined first based on the camera corresponding to the abnormal image.

[0115] Step S4113: Output the platform door number and the preset child trapped signal to the platform staff, instructing the platform staff to carry out the rescue.

[0116] A child entrapment signal is a signal indicating to platform staff that a child is currently trapped and that priority should be given to responding and sending appropriate rescue tools (such as a special key and comforting items) to the scene. It is triggered when the anomalous point cloud matches a child point cloud and is sent to the platform staff's terminal. This signal includes a text message indicating the child is trapped (e.g., "Child trapped, please proceed immediately"), a link to an anomalous image, and a high priority level, all pre-set by those skilled in the art.

[0117] The methods for instructing platform staff to carry out rescue operations include: (The methods described include:) Outputting the platform door number and a pre-set signal indicating a child is trapped. Step S41130: If the first proportion is 0, the first platform door is opened slowly to obtain the resistance of the first platform door. If the resistance exceeds the preset safety threshold, the opening of the first platform door is stopped and the platform door number and the preset child trapped signal are output to the platform staff, instructing the platform staff to enter the gap from the side of the opened first platform door to rescue the trapped child.

[0118] If the first percentage is 0, it means that the child is currently at the second platform door. The first platform door can be opened first. It is necessary to continuously monitor the resistance of the first platform door to prevent the door from squeezing the child when the child moves in the gap area.

[0119] If the resistance exceeds the preset safety threshold, it indicates that the child may be moving within the gap area. Therefore, the first platform door is stopped from opening, and the platform door number and a child trapped signal are sent to the platform staff, instructing the platform staff to enter the gap from the side of the opened first platform door to rescue the trapped child.

[0120] If the resistance does not exceed the preset safety threshold, it means that the child is not moving within the gap area. The first platform door can be opened until it is fully open. The platform door number and the child being trapped signal should be sent to the platform staff, who will then guide the child out of the gap area.

[0121] Step S41131: If the second proportion is 0, the second platform door is opened slowly to obtain the resistance of the second platform door. If the resistance exceeds the preset safety threshold, the opening of the second platform door is stopped and the platform door number and the preset child trapped signal are output to the platform staff, instructing the platform staff to enter the gap from the side of the opened second platform door to rescue the trapped child.

[0122] If the second percentage is 0, it means that the child is currently in the gap area corresponding to the first platform door. The second platform door can be opened first. It is necessary to continuously monitor the resistance of the second platform door to prevent the door from squeezing the child when the child moves in the gap area.

[0123] If the resistance exceeds the preset safety threshold, it indicates that the child may be moving within the gap area. Therefore, the second platform door is stopped from opening, and the platform door number and the preset child trapped signal are output to the platform staff, instructing the platform staff to enter the gap from the side of the opened second platform door to rescue the trapped child.

[0124] If the resistance does not exceed the preset safety threshold, it means that the child is not moving within the gap area. The second platform door can be opened until it is fully open. Similarly, the platform door number and the child being trapped signal should be sent to the platform staff, who will then guide the child out of the gap area.

[0125] Step S41132: If both the first and second proportions are not 0, output the platform door number and the preset child trapped signal to the platform staff, instructing the platform staff to carry out the rescue.

[0126] If both the first and second proportions are not 0, it means that the child is located in the gap area between the two doors. At this time, neither the first nor the second platform door can be opened. Therefore, the platform door number and the child being trapped signal are sent to the platform staff to instruct them to carry out the rescue.

[0127] This also includes a method for inspecting train door 2, which includes: Step S4100: Retrieve the in-vehicle video corresponding to platform door 1 and the corresponding train door 2.

[0128] The in-car video refers to the video footage inside the carriage corresponding to train door 2. A camera is fixedly installed above each door, which can capture the entire door and surrounding area; the in-car video is captured by this camera.

[0129] Step S4101: Monitor the closing line of the train door based on the characteristics of the train door and the video inside the train.

[0130] The train door closure line refers to the line appearing in the video image as the contact gap between the left and right door leaves when the train door 2 is fully closed. It is usually represented as a vertical dark or light stripe with a certain width. First, a person skilled in the art captures a video frame inside the train with the door 2 fully closed and free of obstructions. The position of the closure line between the door leaves is manually marked on this frame (for example, by clicking and drawing a vertical line segment with a mouse). The image column coordinates corresponding to this position, along with the grayscale or color values ​​of a series of pixels at that coordinate, are used as the initial closure line and stored in a file. During runtime, the vertical line segment at the fixed column coordinate is directly extracted from the video frame inside the train, and the brightness or color values ​​of a series of pixels on that line segment are collected, which is the train door closure line at the current moment.

[0131] Step S4102: Extract the initial closed line from the train door features.

[0132] The initial closed line refers to the baseline closed line extracted from the in-vehicle video image where the train door 2 is completely closed and there are no foreign objects. The method of obtaining it is the same as the description of the initial closed line in step S4101.

[0133] Step S4103: Compare the train door closing line with the initial closing line. If the train door closing line and the initial closing line are inconsistent, control the train door 2 to open to the preset release gap and output a preset departure prompt signal to the trapped personnel.

[0134] The release gap refers to the width of the opening when the door is opened to release small foreign objects such as clothing and bags that are caught in it, which is preset by those skilled in the art.

[0135] The exit prompt signal is a signal indicating that the trapped personnel have been freed and the trapped foreign object can be removed. It is preset by those skilled in the art and output when the train door 2 is opened to the release gap. It is usually presented in the form of a voice command, such as "The door is open. Please carefully remove the trapped object."

[0136] If the closing line of the train door is inconsistent with the initial closing line, it means that a foreign object is stuck in the fully closed train door 2. The train door 2 needs to be opened to the release gap and an exit warning signal needs to be output to the trapped personnel so that the trapped personnel can remove the stuck foreign object and leave the trapped area.

[0137] If the closing line of the train door is consistent with the initial closing line, it means that no foreign object is trapped in the fully closed train door 2, and no additional operation is required.

[0138] This includes a method for handling situations where the status of the platform doors and the train doors are inconsistent. This method includes: Step S400: Acquire a 3D point cloud and compare it with a preset 3D background to identify abnormal point clouds.

[0139] Step S401: When the abnormal point cloud is a preset human point cloud, if the platform door is in a preset unclosed state and the train door is in a preset fully closed state, then control the platform door 1 to fully open and output a preset departure prompt signal.

[0140] The unclosed state refers to the state in which platform door 1 or train door 2 has not reached the fully closed and locked position, which is preset by those skilled in the art.

[0141] The fully closed state refers to the platform door 1 or train door 2 being completely closed and in a mechanically locked state, which is preset by those skilled in the art.

[0142] When the abnormal point cloud is the preset human point cloud, it means that a person is trapped in the gap area. At this time, the specific opening and closing status of train door 2 and platform door 1 should be determined first.

[0143] Reference Figure 2 In the upper part, if the platform door is in the preset unclosed state and the train door is in the preset fully closed state, it means that platform door 1 is not closed and train door 2 is fully closed. Therefore, platform door 1 will be fully opened and an exit prompt signal will be output to prompt the trapped personnel to leave.

[0144] Step S402: If the platform door is in the preset fully closed state and the train door is in the preset unclosed state, then control the train door 2 to fully open and output a preset departure prompt signal.

[0145] Reference Figure 2 In the second half, if the platform door is in the preset fully closed state and the train door is in the preset unclosed state, it means that platform door 1 is fully closed and train door 2 is not closed. Therefore, train door 2 will be fully opened and an exit prompt signal will be output to prompt the trapped personnel to leave.

[0146] This includes a method for checking whether the train door or platform door is in a preset fully closed state, the method comprising: Step S403: Define the train door 2 or platform door 1, which is in a preset fully closed state, as a closed door.

[0147] Step S404: Obtain the pressure at the edge of the closed door.

[0148] Door edge pressure refers to the pressure detected at the edge of platform screen door 1 or train door 2. Multiple thin-film pressure sensors are installed at the edge of each platform screen door 1 and train door 2, and the door edge pressure is obtained through detection by the thin-film pressure sensors.

[0149] Step S405: If the pressure at the edge of the door exceeds the preset normal closing pressure, it is determined that there is a foreign object trapped in the closing door.

[0150] Normal closing pressure refers to the reference pressure value detected by the pressure sensor at the edge of the platform door 1 or train door 2 when it is fully closed and free of any foreign objects. The normal closing pressure is pre-measured by those skilled in the art when the door is fully closed and free of any obstructions; this pressure is the normal closing pressure.

[0151] If the pressure at the edge of the door exceeds the preset normal closing pressure, it indicates that a foreign object is trapped between the door panels, so it is determined that a foreign object is trapped when closing the door.

[0152] Step S406: Control the closing door to open to the preset release gap, and output the departure prompt signal again.

[0153] Open the closed door to the preset release gap and output the exit prompt signal again to prompt the trapped personnel to remove the trapped foreign object and leave.

[0154] This also includes a method for handling abnormal point clouds that are not preset human point clouds, the method comprising: Step S4010: If the platform door is in the preset unclosed state and the train door is in the preset fully closed state, then control the platform door 1 to fully open, and find the owner of the item at the platform door 1 based on the gap area and real-time video.

[0155] The owner of the item refers to the person carrying or owning the item that is caught in the closing process of platform door 1 or train door 2, and is therefore unable to leave. The method for determining the owner of the item at platform door 1 is as follows: using real-time video, within the gap area and near platform door 1, a target tracking algorithm is used to detect passengers whose spatial location is closest to the foreign object and whose movement trajectory overlaps with or moves in the same direction as the foreign object within a preset time window (e.g., the first 10 seconds) before the foreign object enters the gap area. These passengers are identified as the owner of the item. If no passenger meeting the above criteria is detected, the owner of the item is determined to be non-existent.

[0156] When the abnormal point cloud is not the preset human body point cloud, it means that the object is trapped in the gap area.

[0157] If the platform door is in the preset unclosed state and the train door is in the preset fully closed state, it means that platform door 1 is not fully closed and train door 2 is fully closed. The trapped item is stuck between platform door 1. Therefore, platform door 1 should be fully opened, and the owner of the item at platform door 1 should be found based on the gap area and real-time video.

[0158] Step S4011: If the platform door is in the preset fully closed state and the train door is in the preset unclosed state, then control the train door 2 to open fully. Based on the platform door 1 and the corresponding train door 2, retrieve the in-car video corresponding to the train door 2, and find the owner of the item at the train door 2 in the in-car video.

[0159] The method for determining the owner of the item in this step is as follows: First, refer to the method for acquiring the in-vehicle video in step S4100 to find the in-vehicle video corresponding to the current train door 2. Using the in-vehicle video, a target tracking algorithm is used to detect passengers who are closest to the spatial position of the foreign object and whose movement trajectory overlaps with or moves in the same direction as the foreign object within a preset time window (e.g., the first 10 seconds) before the foreign object enters the gap area. These passengers are identified as the owner of the item. If no passenger meeting the above conditions is detected, it is determined that the owner of the item does not exist.

[0160] If the platform door is in the preset fully closed state and the train door is in the preset unclosed state, it means that platform door 1 is fully closed and train door 2 is not fully closed. The trapped item is stuck between train door 2. Therefore, fully open train door 2, retrieve the in-vehicle video corresponding to train door 2 based on platform door 1 and the corresponding train door 2, and find the owner of the item at train door 2 in the in-vehicle video.

[0161] Step S4012: If the owner of the item exists, output the preset passenger pick-up signal.

[0162] The passenger retrieval signal is a signal that prompts a passenger whose item is stuck to retrieve it. It is pre-set by a person skilled in the art and is presented in the form of a voice command, such as "Please retrieve your item".

[0163] If the owner of the item is present, it means that the passenger whose item is trapped is right next to it. A signal can be sent to the passenger to retrieve the item.

[0164] Step S4013: If the owner of the item does not exist, output a preset item removal signal to the platform staff.

[0165] The object removal signal is a signal that prompts staff to remove the stuck object. It is pre-set by those skilled in the art and sent to the platform staff's terminal in text form, which includes a video of the foreign object and the platform door number.

[0166] If the owner of the item does not exist, it means that the owner of the stuck item has left and the stuck item needs to be manually removed by the platform staff. Therefore, an item removal signal is output to the platform staff.

[0167] Based on the same inventive concept, embodiments of the present invention provide an intelligent monitoring system for train gaps at platform screen doors based on AI video analysis.

[0168] An AI-based video analysis-based intelligent monitoring system for train gaps at platform screen doors includes: The acquisition module is used to acquire real-time video, 3D point cloud, resistance, and pressure at the edge of the door. The memory stores a computer program that can be loaded and executed by the processor, which is a method for intelligent monitoring of train gaps at platform screen doors based on AI video analysis. The processor loads and executes programs from memory.

[0169] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional modules is used as an example. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. The specific working process of the system, device, and unit described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0170] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A method for intelligent monitoring of train gaps at platform screen doors based on AI video analysis, characterized in that, include: Step S1: Obtain real-time video of each platform door (1); Step S2: When the preset train door features are detected in the real-time video, monitor the gap area between the train door (2) and the platform door (1) in the real-time video according to the train door features and the preset platform door (1) features; Step S3: When a preset foreign object feature appears in the gap area, retrieve the laser blocking signal corresponding to the gap area; Step S4: If the laser blocking signal exists, retrieve the platform door status and train door status; Step S40: If the platform door status and the train door status are inconsistent, the real-time video is captured according to the foreign object characteristics to obtain the foreign object video, and the foreign object video and the preset door linkage fault signal are output to the train driver. Step S41: If the platform door status and the train door status are consistent, the real-time video is captured according to the foreign object characteristics to obtain the foreign object video, and the foreign object video and the preset foreign object alarm signal are output to the train driver. Step S5: When there are no foreign objects in the gap area, no laser blocking signal, and the platform door status and train door status are consistent, output a preset departure prompt signal to prompt the train driver to depart.

2. The method for intelligent monitoring of train gaps at platform screen doors based on AI video analysis according to claim 1, characterized in that, It also includes a handling method if the platform door status and the train door status are the same, the method including: Step S410: If both the platform door status and the train door status are in the preset fully closed state, acquire the three-dimensional point cloud, compare the three-dimensional point cloud with the preset three-dimensional background, and identify abnormal point clouds. Step S411: If the abnormal point cloud matches the preset human point cloud, extract the image of the foreign object from the video of the foreign object. Step S412: Platform doors (1) include the first platform door and the second platform door; Step S413: Map the anomalous point cloud onto the foreign object image to obtain the anomalous image; Step S414: Analyze the proportion of abnormal point clouds located at the first platform gate and the proportion of abnormal point clouds located at the second platform gate in the abnormal image; Step S415: If the first proportion is 0, the second platform door is opened at a preset slow speed to obtain the resistance of the second platform door. If the resistance exceeds the preset safety threshold, the opening of the second platform door is stopped and a preset personnel guidance signal is output to the platform staff for guidance. If the resistance does not exceed the safety threshold, the opening continues until it is fully open and a personnel guidance signal is output. Step S416: If the second proportion is 0, the first platform door is opened at a preset slow speed to obtain the resistance of the first platform door. If the resistance exceeds the preset safety threshold, the opening of the first platform door is stopped and a preset personnel guidance signal is output to the platform staff for guidance. If the resistance does not exceed the safety threshold, the opening continues until it is fully open and a personnel guidance signal is output. Step S417: If both the first and second proportions are not 0, then execute the preset cross-door rescue process.

3. The method for intelligent monitoring of train gaps at platform screen doors based on AI video analysis according to claim 2, characterized in that, If neither the first nor the second proportion is 0, the methods for executing the preset cross-door rescue process include: Step S4170: If the first proportion is greater than the second proportion, the second platform door is opened slowly to obtain the resistance of the second platform door. If the resistance exceeds the preset safety threshold, the opening of the second platform door is stopped and a personnel guidance signal is output to the platform staff. Step S4171: If the resistance does not exceed the safety threshold, open the second platform door to full opening and output a preset self-rescue guidance signal to prompt the trapped personnel to use the space created by the opening of the second platform door to adjust their posture and escape. Step S4172: If the first proportion is less than the second proportion, the first platform door is opened slowly to obtain the resistance of the first platform door. If the resistance exceeds the preset safety threshold, the opening of the first platform door is stopped and a personnel guidance signal is output to the platform staff. Step S4173: If the resistance does not exceed the safety threshold, continue to open the first platform door until it is fully open, and output a preset self-rescue guidance signal to prompt the trapped personnel to use the space created by the opening of the first platform door to adjust their posture and escape. Step S4174: If the first proportion is equal to the second proportion, then output the preset personnel assistance signal to the platform staff.

4. The method for intelligent monitoring of train gaps at platform screen doors based on AI video analysis according to claim 2, characterized in that, Also includes: Step S4110: Human point cloud includes adult point cloud and child point cloud; Step S4111: If the abnormal point cloud matches the adult point cloud, then execute steps S411 to S417 and steps S4170 to S4174. Step S4112: If the abnormal point cloud matches the child point cloud, then extract the platform door number from the abnormal image; Step S4113: Output the platform door number and the preset child trapped signal to the platform staff, instructing the platform staff to carry out the rescue.

5. The method for intelligent monitoring of train gaps at platform screen doors based on AI video analysis according to claim 4, characterized in that, Methods for instructing platform staff to implement rescue operations by providing them with the platform door number and a pre-set child entrapment signal include: Step S41130: If the first proportion is 0, the first platform door is opened slowly to obtain the resistance of the first platform door. If the resistance exceeds the preset safety threshold, the opening of the first platform door is stopped and the platform door number and the preset child trapped signal are output to the platform staff, instructing the platform staff to enter the gap from the side of the opened first platform door to rescue the trapped child. Step S41131: If the second proportion is 0, the second platform door is opened slowly to obtain the resistance of the second platform door. If the resistance exceeds the preset safety threshold, the opening of the second platform door is stopped and the platform door number and the preset child trapped signal are output to the platform staff, instructing the platform staff to enter the gap from the side of the opened second platform door to rescue the trapped child. Step S41132: If both the first and second proportions are not 0, output the platform door number and the preset child trapped signal to the platform staff, instructing the platform staff to carry out the rescue.

6. The method for intelligent monitoring of train gaps at platform screen doors based on AI video analysis according to claim 2, characterized in that, It also includes a method for inspecting train doors (2), which includes: Step S4100: Retrieve the in-vehicle video corresponding to the platform door (1) and the corresponding train door (2); Step S4101: Monitor the train door closure line based on the characteristics of the train door and the in-car video; Step S4102: Extract the initial closed line from the train door features; Step S4103: Compare the train door closing line with the initial closing line. If the train door closing line and the initial closing line are inconsistent, control the train door (2) to open to the preset release gap and output the preset departure prompt signal to the trapped personnel.

7. The method for intelligent monitoring of train gaps at platform screen doors based on AI video analysis according to claim 1, characterized in that, It also includes a method for handling situations where the status of the platform doors and the train doors are inconsistent, the method including: Step S400: Acquire a 3D point cloud and compare it with a preset 3D background to identify abnormal point clouds; Step S401: When the abnormal point cloud is the preset human point cloud, if the platform door is in the preset unclosed state and the train door is in the preset fully closed state, then control the platform door (1) to fully open and output the preset departure prompt signal. Step S402: If the platform door is in the preset fully closed state and the train door is in the preset unclosed state, then control the train door (2) to fully open and output the preset departure prompt signal.

8. The method for intelligent monitoring of train gaps at platform screen doors based on AI video analysis according to claim 7, characterized in that, It also includes a method for checking whether the train door or platform door is in a preset fully closed state, the method comprising: Step S403: Define the train door (2) or platform door (1) that is in a preset fully closed state as a closed door; Step S404: Obtain the pressure at the edge of the closed door; Step S405: If the pressure at the edge of the door exceeds the preset normal closing pressure, it is determined that there is a foreign object trapped in the closing door; Step S406: Control the closing door to open to the preset release gap, and output the departure prompt signal again.

9. The method for intelligent monitoring of train gaps at platform screen doors based on AI video analysis according to claim 7, characterized in that, It also includes a method for handling abnormal point clouds that are not preset human point clouds, the method including: Step S4010: If the platform door is in the preset unclosed state and the train door is in the preset fully closed state, then control the platform door (1) to open fully and find the owner of the item at the platform door (1) based on the gap area and real-time video. Step S4011: If the platform door is in the preset fully closed state and the train door is in the preset unclosed state, then control the train door (2) to open fully. Retrieve the in-car video corresponding to the platform door (1) and the corresponding train door (2), and find the owner of the item at the train door (2) in the in-car video. Step S4012: If the owner of the item exists, output a preset passenger pick-up signal; Step S4013: If the owner of the item does not exist, output a preset item removal signal to the platform staff.

10. An intelligent monitoring system for train gaps at platform screen doors based on AI video analysis, characterized in that, include: The acquisition module is used to acquire real-time video, 3D point cloud, resistance, and pressure at the edge of the door. A memory for storing a program for an AI video analysis-based intelligent monitoring method for train gaps at platform screen doors, as described in any one of claims 1 to 9; The processor loads and executes programs from memory.

Images