An elevator electric vehicle recognition method and system based on anomaly detection

Abstract

The present invention relates to the technical field of elevator electric vehicle recognition, and specifically to an elevator electric vehicle recognition method and system based on anomaly detection, including the following steps: Based on the picture data of the elevator camera and the light curtain infrared signal data, frame by frame extract the object edge contour in the picture during the door opening period, use edge detection to separate the object and the background area, record the width and height ratio of the object edge, calculate the change characteristics of the edge shape on the time axis, and generate an object edge shape data set. By dynamically extracting the object edge contour and combining the light curtain signal data, analyze the object width-to-height ratio and movement trajectory, and through the comprehensive analysis of the occlusion time, light curtain signal frequency and reflection intensity, improve the accuracy of occlusion behavior recording. Use the correlation verification between the load change and the trajectory data to achieve accurate recording of the abnormal behavior of the electric vehicle, and trigger the elevator locking, voice warning and linkage alarm mechanism in real time, providing comprehensive guarantee for the safe operation of the elevator.

Description

Technical Field

[0001] The present invention relates to the technical field of elevator electric vehicle recognition, and particularly to an elevator electric vehicle recognition method and system based on anomaly detection. Background Art

[0002] The technical field of elevator electric vehicle recognition includes the monitoring of elevator operation status and the detection and analysis of the behavior of electric vehicles entering and leaving the elevator. The core content of this technical field is to identify and process abnormal behaviors in the elevator based on elevator operation data, video surveillance data, and information collected by other sensors. The overall elevator electric vehicle recognition technology involves multiple aspects, including elevator status information collection and transmission, elevator usage behavior analysis, item recognition and classification, elevator safety analysis and detection, etc. Through information collection, analysis, and feedback mechanisms, it provides technical support for elevator safety management.

[0003] Among them, the elevator electric vehicle recognition method based on anomaly detection refers to identifying and analyzing the behavior of electric vehicles entering the elevator by constructing a model and method based on anomaly detection. In view of the possible abnormal situations during the process of electric vehicles entering and leaving the elevator, using the data collected by elevator sensors and the image information obtained from video surveillance, combined with specific signal processing methods and classification algorithms, to achieve the recognition of electric vehicle characteristics. Specifically, it includes steps such as behavior detection based on elevator operation status, extracting the appearance characteristics of electric vehicles through image analysis, and classification and judgment based on multi-modal data fusion, forming a technical process for electric vehicle recognition.

[0004] In the recognition of electric vehicles entering the elevator in the prior art, it usually relies on a single data source. For example, simple feature extraction and judgment are carried out through elevator operation status or video surveillance data, lacking in-depth fusion and dynamic analysis of multi-dimensional data, resulting in low recognition accuracy. The analysis of object morphological characteristics is only limited to static or low-dimensional parameters, such as a single aspect ratio or image contour, without combining dynamic behavior data and light curtain signal changes for in-depth judgment, and it is easy to make misjudgments in complex scenarios. For the recording and evaluation of occlusion behaviors, it mostly relies on simple field-of-view occlusion information, lacking comprehensive correlation analysis of occlusion duration, frequency, and its relationship with light curtain signals, and it is unable to effectively establish a complete chain of abnormal behaviors, resulting in some abnormal situations being difficult to be accurately captured. In addition, the prior art lacks the correlation verification between load changes and object behaviors. Simply relying on a certain feature for abnormal behavior recognition may lead to misjudgment of events in multi-target scenarios. At the same time, during the disposal process after an abnormal event occurs, the triggering and response of the linkage mechanism in the prior art are lagging, and alarm and control fail to form an efficient coordination, restricting the overall safety of elevator operation management, directly affecting the comprehensiveness and timeliness of the recognition and disposal of abnormal behaviors of electric vehicles in the elevator, and there are significant safety hazards. Summary of the Invention

[0005] The object of the present invention is to solve the drawbacks existing in the prior art, and to propose a method and system for identifying electric vehicles in elevators based on anomaly detection.

[0006] To achieve the above object, the present invention adopts the following technical solutions:

[0007] A method for identifying electric vehicles in elevators based on anomaly detection, comprising the following steps:

[0008] S1: Based on the picture data of the elevator camera and the light curtain infrared signal data, extract the edge contours of the objects frame by frame during the door opening period, analyze the width-to-height ratio of the objects, calculate the contour closure parameter, associate the width and reflection intensity data of the light curtain crossing path, perform the recognition processing of the object shape characteristics and the signal path, and generate the determination result of the object geometric characteristics;

[0009] S2: Based on the determination result of the object geometric characteristics, analyze the width-to-height ratio of the target object and the change range of the continuous movement trajectory, combine the object shape characteristics and the position offset data, statistically analyze the light curtain signal crossing speed and the reflection intensity fluctuation, compare the geometric shape data with the speed fluctuation, and generate the matching result of the target object characteristics;

[0010] S3: Based on the matching result of the target object characteristics, extract the time period and the change area of the elevator camera field of view occlusion, statistically analyze the light curtain trigger frequency and the reflection change during the occlusion time, analyze the occlusion duration and frequency, calculate the overlap rate of the occlusion duration and frequency, and perform a comparative analysis in combination with the occlusion data to generate the record result of the abnormal occlusion behavior chain;

[0011] S4: Based on the record result of the abnormal occlusion behavior chain, analyze the load change before and after the door opening, verify the load change value and the object trajectory data extracted by the light curtain, screen the event chains in which the trajectory behavior is consistent with the load change, analyze the behavior patterns and occlusion characteristics in the event chains, and generate the record of the abnormal behavior of the electric vehicle;

[0012] S5: Based on the record of the abnormal behavior of the electric vehicle, transmit the record data to the elevator control mechanism, trigger the elevator door lock and issue a voice warning, upload the data to the remote monitoring mechanism, link the building alarm mechanism, and generate the response result of the elevator abnormal state event.

[0013] As a further solution of the present invention, the determination result of the geometric characteristics of the object includes the object edge contour parameters, the width-to-height ratio analysis result, the contour closure parameter, the through-path width feature, and the reflection intensity feature. The matching result of the target object characteristics includes the object width-to-height ratio feature, the motion trajectory range feature, the position offset data, the through-speed characteristic, and the reflection intensity fluctuation characteristic. The record result of the abnormal occlusion behavior chain includes the occlusion time period data, the occlusion area change feature, the light curtain signal trigger frequency statistics, the cumulative reflection change feature, and the occlusion duration frequency overlap rate. The record of the abnormal behavior of the electric vehicle includes the load change feature, the trajectory behavior chain, the event chain mode characteristic, and the occlusion characteristic analysis result. The response result of the abnormal state event of the elevator includes the door lock trigger signal, the voice warning instruction, the remote monitoring upload data, and the building alarm linkage mechanism.

[0014] As a further solution of the present invention, based on the video data of the elevator camera and the light curtain infrared signal data, the object edge contour during the door opening period is extracted frame by frame, the width-to-height ratio of the object is analyzed, the contour closure parameter is calculated, the through-path width of the light curtain is associated with the reflection intensity data, and the recognition process of the object shape characteristics and the signal path is carried out. The specific steps for generating the determination result of the geometric characteristics of the object are as follows:

[0015] S101: Based on the video data of the elevator camera and the light curtain infrared signal data, the object edge contour in the video during the door opening period is extracted frame by frame. The edge detection is used to separate the object and the background area, the width and height ratio of the object edge is recorded, and the change characteristics of the edge shape on the time axis are calculated to generate the object edge shape data set.

[0016] S102: Based on the object edge shape data set, the edge closure deviation value is calculated, the geometric positions of the edge start point and the end point are compared, the contour closure is judged, the spatial continuity of the inter-frame closure change is analyzed, the edge closure degree is curve-fitted, and the key feature points are recorded to generate the edge contour closure characteristic data.

[0017] S103: Based on the edge contour closure characteristic data, combined with the through-path width and reflection intensity information of the light curtain infrared signal data, the spatial correlation between the signal path and the contour closure parameter is analyzed, and the object feature mapping relationship is constructed by a matrix through multi-dimensional characteristic data to generate the determination result of the geometric characteristics of the object.

[0018] As a further solution of the present invention, the specific formula for the edge closure deviation value is:

[0019]

[0020] Among them, C b represents the closure deviation value, represents the abscissa of the starting point of the edge of the i-th frame, represents the ordinate of the starting point of the edge of the i-th frame, represents the abscissa of the ending point of the edge of the i-th frame, represents the ordinate of the ending point of the edge of the i-th frame, A i represents the area of the closed region of the i-th frame, n represents the total number of frames, and π is the constant of pi.

[0021] As a further solution of the present invention, based on the determination result of the geometric characteristics of the object, analyzing the width-to-height ratio of the target object and the change range of the continuous movement trajectory, combining the morphological characteristics of the object and the position offset data, statistically analyzing the penetration speed of the light curtain signal and the fluctuation of the reflection intensity, comparing the geometric shape data with the speed fluctuation, and generating the specific steps of the matching result of the target object characteristics are as follows:

[0022] S201: Based on the determination result of the geometric characteristics of the object, extract the change range of the width-to-height ratio of the target object, group according to the width and height data between frames, calculate the ratio change of each group, statistically analyze the frequency distribution of the ratio change interval, extract the key points of the continuous change trend on the time axis, and generate the width-to-height ratio change data of the target object;

[0023] S202: Based on the width-to-height ratio change data of the target object, combining the movement trajectory and the position offset data, analyze the change amplitude of each frame position in the trajectory, calculate the cumulative displacement of the change direction, screen the key points with abnormal offset characteristics for the trajectory curve, and generate the trajectory and position characteristic data of the target object;

[0024] S203: Based on the trajectory and position characteristic data of the target object, combining the penetration speed of the light curtain signal and the fluctuation range of the reflection intensity, compare the time nodes of the signal parameters with the trajectory change characteristics, statistically analyze the parameter correlation between the corresponding signal data and the geometric characteristics, and generate the matching result of the target object characteristics.

[0025] As a further solution of the present invention, based on the matching result of the target object characteristics, extract the occlusion time period and the change area in the field of view of the elevator camera, statistically analyze the light curtain trigger frequency and the reflection change during the occlusion time, analyze the occlusion duration and frequency, calculate the overlap rate of the occlusion duration and frequency, and perform a comparative analysis in combination with the occlusion data to generate the specific steps of the abnormal occlusion behavior chain record result are as follows:

[0026] S301: Based on the matching result of the target object characteristics, extract the occlusion time period and the change area in the field of view of the elevator camera, detect the boundary point positions of the occlusion area frame by frame, record the change value of the occlusion area on the time axis, statistically analyze the trend of the occlusion area change and the boundary stability, and generate the occlusion time and area change data;

[0027] S302: Based on the occlusion time and area change data, extract the frequency and reflection intensity of the light curtain trigger signal during the occlusion period, synchronously analyze the change amplitude of the light curtain signal in the time interval and the occlusion time, calculate the overlapping ratio of the two, screen the characteristic intervals where occlusion and signal match, and generate the relationship data between the occlusion time and the light curtain trigger frequency;

[0028] S303: Based on the relationship data between the occlusion time and the light curtain trigger frequency, extract the change of the occlusion area and the light curtain signal trigger parameters, perform segmented comparison on the signal and occlusion data, record the correlation between signal anomalies and regional fluctuations at time nodes, extract and classify the abnormal correlation behaviors, and generate the record result of the abnormal occlusion behavior chain.

[0029] As a further solution of the present invention, the specific calculation formula for the overlapping intensity value of the occlusion and the light curtain signal is:

[0030]

[0031] wherein, R o represents the overlapping intensity value of the occlusion and the light curtain signal, ΔF k represents the change amount of the trigger frequency of the light curtain signal in the k-th time period, T k represents the duration of the k-th time period, B k represents the area value of the k-th occlusion area, π is the constant of pi, d represents the total number of occlusion time periods, m represents the total number of signal trigger time periods, and T j represents the time value of the j-th signal trigger.

[0032] As a further solution of the present invention, based on the record result of the abnormal occlusion behavior chain, analyze the load change before and after the door opening, verify the load change value and the object trajectory data extracted by the light curtain, screen the event chains where the trajectory behavior and the load change are consistent, analyze the behavior patterns and occlusion characteristics in the event chains, and the specific steps for generating the record of abnormal behaviors of the electric vehicle are:

[0033] S401: Based on the record result of the abnormal occlusion behavior chain, extract the elevator load change data before and after the door opening, statistically analyze the load change data of each frame on the time axis, record the change value of the load of each frame, calculate the load change trend, and mark the time points with significant changes to generate the load change and time series data;

[0034] S402: Based on the load change and time series data, verify the load change value and the object trajectory data extracted by the light curtain, extract the synchronization between the load change events and the trajectory behaviors on the time axis, screen the matching event chains according to the change range, and statistically analyze the event correlation characteristics to generate the load and trajectory event chain data;

[0035] S403: Based on the load and trajectory event chain data, analyze the correspondence between the continuity of the behavior pattern and the occlusion characteristics in the event chain, calculate the correlation parameters between the trajectory change and the occlusion behavior, screen and sort out the abnormal behavior characteristics, and generate the abnormal behavior records of the electric vehicle.

[0036] As a further solution of the present invention, based on the abnormal behavior records of the electric vehicle, the specific steps of transmitting the record data to the elevator control mechanism, triggering the elevator door lock and issuing a voice warning, uploading the data to the remote monitoring mechanism, and linking the building alarm mechanism to generate the elevator abnormal state event response result are as follows:

[0037] S501: Based on the abnormal behavior records of the electric vehicle, transmit the record data to the elevator control mechanism, analyze the abnormal behavior characteristic data and corresponding control parameters, activate the locking function of the elevator door, trigger the voice warning device and record the response status, and generate the elevator locking and voice warning response data;

[0038] S502: Based on the elevator locking and voice warning response data, upload the abnormal behavior records and elevator operation state data to the remote monitoring mechanism, establish a data transmission network interface and monitor the integrity of the data transmission, record the feedback status information of the remote monitoring mechanism, and generate the remote monitoring abnormal behavior upload result;

[0039] S503: Based on the remote monitoring abnormal behavior upload result, link the building alarm mechanism, analyze the abnormal behavior data to trigger the alarm condition and synchronize the control parameters, activate the alarm device in the building and record the alarm response time and execution status, and generate the elevator abnormal state event response result.

[0040] Based on the same inventive concept, an elevator electric vehicle identification system based on anomaly detection is also proposed, including:

[0041] The object edge extraction module extracts the object edge contours frame by frame based on the image data of the elevator camera and the light curtain infrared signal data, calculates the contour closure parameter, correlates the light curtain crossing path width and reflection intensity data, and generates the object geometric characteristic determination result;

[0042] The target characteristic analysis module analyzes the width-height ratio and continuous motion trajectory change range of the target object based on the object geometric characteristic determination result, counts the light curtain signal crossing speed and reflection intensity fluctuation, compares the geometric shape data with the speed fluctuation, and generates the target object characteristic matching result;

[0043] The abnormal occlusion analysis module extracts the elevator camera field of view occlusion time period and change area based on the target object characteristic matching result, counts the light curtain trigger frequency and reflection change during the occlusion time, calculates the occlusion duration and frequency overlap rate, and conducts a comparative analysis in combination with the occlusion data to generate the abnormal occlusion behavior chain record result;

[0044] Based on the result of the abnormal occlusion behavior chain record, the load behavior verification module analyzes the load change before and after the door opening, verifies the load change value and the object trajectory data extracted by the light curtain, screens the event chain where the trajectory behavior is consistent with the load change, and generates an abnormal behavior record of the electric vehicle.

[0045] Based on the abnormal behavior record of the electric vehicle, the abnormal event response module transmits the record data to the elevator control mechanism, triggers the elevator door lock and issues a voice warning, links the building alarm mechanism, and generates an abnormal state event response result of the elevator.

[0046] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0047] In the present invention, by dynamically extracting the object edge contour and combining the light curtain signal data, analyzing the object width-to-height ratio and motion trajectory, and comprehensively analyzing the occlusion time, light curtain signal frequency and reflection intensity, the accuracy of the occlusion behavior record is improved. Using the correlation verification between the load change and the trajectory data, the accurate record of the abnormal behavior of the electric vehicle is realized, and the elevator lock, voice warning and linked alarm mechanism are triggered in real-time feedback, providing comprehensive protection for the safe operation of the elevator. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0049] Figure 1 It is a schematic diagram of the step flow of the present invention.

[0050] Figure 2 It is a flowchart of step S1 of the present invention.

[0051] Figure 3 It is a flowchart of step S2 of the present invention.

[0052] Figure 4 It is a flowchart of step S3 of the present invention.

[0053] Figure 5 It is a flowchart of step S4 of the present invention.

[0054] Figure 6 It is a flowchart of step S5 of the present invention.

[0055] Figure 7 It is a system module diagram of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0056] The technical solutions in the present invention will be described below with reference to the accompanying drawings.

[0057] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to represent examples, illustrations or explanations. Any embodiment or design solution described as an "example" in the present invention should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Rather, the use of the word "example" is intended to present concepts in a specific manner. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one of the two.

[0058] In the embodiments of the present invention, "image" and "picture" can sometimes be used interchangeably. It should be noted that when the difference is not emphasized, the meanings they express are the same. "Of", "corresponding", and "corresponding" can sometimes be used interchangeably. It should be noted that when the difference is not emphasized, the meanings they express are the same.

[0059] In the embodiments of the present invention, sometimes subscripts such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meanings they express are the same.

[0060] To make the technical problems to be solved, technical solutions and advantages of the present invention clearer, the following will be described in detail with reference to the accompanying drawings and specific embodiments.

[0061] Please refer to Figure 1 , an elevator electric vehicle recognition method based on anomaly detection, comprising the following steps:

[0062] S1: Based on the frame data of the elevator camera and the light curtain infrared signal data, extract the object edge contour frame by frame during the door opening period, analyze the aspect ratio of the object, calculate the contour closure parameter, correlate the light curtain crossing path width with the reflection intensity data, perform the recognition processing of the object morphological characteristics and the signal path, and generate the object geometric characteristic determination result;

[0063] S2: Based on the object geometric characteristic determination result, analyze the aspect ratio of the target object and the change range of the continuous motion trajectory, combine the object morphological characteristics and the position offset data, statistically analyze the light curtain signal crossing speed and the reflection intensity fluctuation, compare the geometric shape data with the speed fluctuation, and generate the target object characteristic matching result;

[0064] S3: Based on the target object characteristic matching result, extract the elevator camera field of view occlusion time period and the change area, statistically analyze the light curtain trigger frequency and the reflection change during the occlusion time, analyze the occlusion duration and frequency, calculate the overlap rate of the occlusion duration and frequency, and perform a comparative analysis in combination with the occlusion data to generate the abnormal occlusion behavior chain record result;

[0065] S4: Based on the recorded results of the abnormal occlusion behavior chain, analyze the load change before and after the door opening, verify the load change value and the object trajectory data extracted by the light curtain, screen the event chains where the trajectory behavior is consistent with the load change, analyze the behavior patterns and occlusion characteristics in the event chains, and generate the abnormal behavior record of the electric vehicle.

[0066] S5: Based on the abnormal behavior record of the electric vehicle, transmit the recorded data to the elevator control mechanism, trigger the elevator door lock and issue a voice warning, upload the data to the remote monitoring mechanism, link the building alarm mechanism, and generate the event response result of the elevator abnormal state.

[0067] The determination results of the object geometric characteristics include the object edge contour parameters, width-to-height ratio analysis results, contour closure parameters, penetration path width characteristics, and reflection intensity characteristics. The target object characteristic matching results include the object width-to-height ratio characteristics, motion trajectory range characteristics, position offset data, penetration speed characteristics, and reflection intensity fluctuation characteristics. The recorded results of the abnormal occlusion behavior chain include the occlusion time period data, occlusion area change characteristics, light curtain signal trigger frequency statistics, cumulative reflection change characteristics, and occlusion duration frequency overlap rate. The abnormal behavior record of the electric vehicle includes the load change characteristics, trajectory behavior chain, event chain mode characteristics, and occlusion characteristic analysis results. The event response result of the elevator abnormal state includes the door lock trigger signal, voice warning instruction, remote monitoring upload data, and building alarm linkage mechanism.

[0068] Please refer to Figure 2 , and the specific steps of S1 are as follows:

[0069] S101: Based on the video data of the elevator camera and the light curtain infrared signal data, frame by frame extract the object edge contour in the video during door opening, use edge detection to separate the object and the background area, record the width and height ratio of the object edge, calculate the change characteristics of the edge shape on the time axis, and generate the object edge shape data set.

[0070] Extract the edge contour of the object in the video through intra-frame analysis. First, use the gradient edge detection technology to calculate the gradient intensity and direction of each pixel in the image, complete the edge detection using the Sobel operator or the Canny algorithm, remove the noise background and separate the foreground object. Then, construct the preliminary contour of the object edge by connecting the edge pixel points, perform pixel measurement on the width and height of the edge contour, calculate the width-to-height ratio and store it. Combine time series analysis, and obtain the change characteristics through curve fitting of the edge shape change, and generate the object edge shape data set.

[0071] S102: Based on the object edge morphology dataset, calculate the edge closure deviation value, compare the geometric positions of the edge start point and the end point, judge the contour closure, analyze the spatial continuity of the inter-frame closure change, perform curve fitting on the edge closure degree, record the key feature points, and generate the edge contour closure feature data;

[0072] The specific formula for calculating the edge closure deviation value is as follows:

[0073]

[0074] Among them, C b represents the closure deviation value, represents the abscissa of the edge start point of the i-th frame, represents the ordinate of the edge start point of the i-th frame, represents the abscissa of the edge end point of the i-th frame, represents the ordinate of the edge end point of the i-th frame, A i represents the area of the closed region of the i-th frame, and n represents the total number of frames.

[0075] The specific parameters involved are as follows:

[0076] C b : The closure deviation value, used to evaluate the overall deviation between the object edge closure and the ideal circular closed region.

[0077] The abscissa and ordinate of the edge start point of the i-th frame. The starting point position of the object edge in the video frame is extracted through the edge detection algorithm and can be obtained by using a pixel coordinate measurement tool. The measurement accuracy should match the frame resolution.

[0078] The abscissa and ordinate of the edge end point of the i-th frame. The end point position of the object edge in the video frame is extracted through the edge detection algorithm.

[0079] A i : The area of the closed region of the edge of the i-th frame, obtained through the closed calculation method of the in-frame edge points. The area calculation uses the pixel measurement method, multiplying the number of pixels in the closed contour region by the single-pixel area.

[0080] n: The total number of frames, used to average the deviation values of all frames. The number of frames is directly calculated from the video duration and the frame rate.

[0081] π: The constant pi, used to calculate the radius value of the ideal circle through the area.

[0082] The specific calculation and derivation process:

[0083] Given a video with 5 frames, a total resolution of 1920×1080, and a frame rate of 30 frames per second, the extracted edge start and end coordinates and the closed area data are as follows: For the first frame, the start coordinate is (100, 200), the end coordinate is (400, 600), and the closed area is 50265 pixels; for the second frame, the start coordinate is (150, 250), the end coordinate is (420, 590), and the closed area is 48750 pixels; for the third frame, the start coordinate is (120, 220), the end coordinate is (410, 620), and the closed area is 51500 pixels; for the fourth frame, the start coordinate is (130, 210), the end coordinate is (430, 610), and the closed area is 50500 pixels; for the fifth frame, the start coordinate is (140, 230), the end coordinate is (440, 620), and the closed area is 49800 pixels; The Euclidean distance between the calculated start and end points of the first frame:

[0084]

[0085] Radius of the ideal circle:

[0086]

[0087] Absolute value of the deviation:

[0088]

[0089] The Euclidean distance between the calculated start and end points of the second frame:

[0090]

[0091] Radius of the ideal circle:

[0092]

[0093] Absolute value of the deviation:

[0094]

[0095] The Euclidean distance between the calculated start and end points of the third frame:

[0096]

[0097] Radius of the ideal circle:

[0098]

[0099] Absolute value of the deviation:

[0100]

[0101] The Euclidean distance between the calculated start and end points of the fourth frame:

[0102]

[0103] Radius of the ideal circle:

[0104]

[0105] Absolute value of deviation:

[0106]

[0107] Euclidean distance between the starting point and the ending point of the 5th frame calculation:

[0108]

[0109] Radius of the ideal circle:

[0110]

[0111] Absolute value of deviation:

[0112]

[0113] Calculation of the closure deviation value:

[0114]

[0115]

[0116] Result description:

[0117] The closure deviation value is 357.69, indicating that there is a certain geometric deviation between the object edge and the ideal closed contour. This value is used for subsequent feature fitting and key point extraction to further describe the variation characteristics of the edge closed shape.

[0118] S103: Based on the edge contour closure feature data, combined with the penetration path width and reflection intensity information of the light curtain infrared signal data, analyze the spatial correlation between the signal path and the contour closure parameters, and construct the object feature mapping relationship through the multi-dimensional characteristic data to generate the object geometric characteristic determination result;

[0119] First, perform pixel-level measurement on the infrared penetration path of the light curtain signal, use the sensor to collect the time series data of the infrared signal intensity, calculate the width value of the path through the reflection intensity analysis of the optical path, then correlate the edge contour closure parameters with the light curtain signal path width in the spatial dimension, map the infrared signal intensity of each spatial point to the specific position on the geometric feature through the construction of the feature mapping matrix, and then analyze the correlation relationship according to the dynamic changes of the path width and the closure characteristics in the time series in the multi-dimensional matrix, and finally generate the object geometric characteristic analysis result.

[0120] Please refer toFigure 3 , the specific steps of S2 are as follows:

[0121] S201: Based on the determination result of the object's geometric characteristics, extract the range of changes in the width-to-height ratio of the target object, group according to the width and height data between frames, calculate the ratio change of each group, statistically analyze the frequency distribution of the ratio change interval, extract the key points of the continuous change trend on the time axis, and generate the width-to-height ratio change data of the target object;

[0122] First, calibrate each frame of the image for regions and detect the width and height boundaries. Numerically quantify the specific values of the width and height using the method of pixel measurement. Calculate the width-to-height ratio of each group of frames, divide and statistically analyze the ratio values according to intervals, and statistically analyze the frequency distribution of the ratio changes within each interval. Subsequently, conduct a continuous trend analysis on the ratio change data on the time axis, extract the key points with significant continuous change trends in the frame sequence, and mark their specific positions on the time axis. Finally, generate the width-to-height ratio change data of the target object.

[0123] S202: Based on the width-to-height ratio change data of the target object, combined with the motion trajectory and position offset data, analyze the change amplitude of each frame position in the trajectory, calculate the cumulative displacement of the change direction, screen for key points with abnormal offset characteristics in the trajectory curve, and generate the trajectory and position feature data of the target object;

[0124] For screening key points with abnormal offset characteristics in the trajectory curve, according to the formula:

[0125]

[0126] Calculate the change amplitude of each frame position and accumulate the displacement. In the formula, x i,t and y i,t respectively represent the horizontal and vertical coordinate positions of the object in the i-th frame, D t represents the cumulative displacement of the trajectory, and n represents the total number of frames.

[0127] The trajectory change amplitude is accumulated by calculating the Euclidean distance between the position coordinates of adjacent frames to characterize the trajectory change trend. In the calculation, by locating the centroid position of the object in each frame, extracting its horizontal and vertical coordinates, calculating the distance value between adjacent frames and accumulating. In a certain example, assuming that the total number of frames is 5, and the frame sequence coordinates are (10, 20), (13, 24), (18, 30), (22, 36), (25, 40) respectively, then the distances between adjacent frames are as follows:

[0128] From the 1st frame to the 2nd frame:

[0129]

[0130] From the 2nd frame to the 3rd frame:

[0131]

[0132] After repeated calculations, the cumulative displacement is:

[0133] D t = 5 + 7.81 + 7.21 + 5.83 ≈ 25.85;

[0134] This result indicates that the trajectory of the object has changed significantly. By analyzing the cumulative displacement curve, key points with abnormal offset characteristics can be marked, and finally, the trajectory and position feature data of the target object are generated.

[0135] S203: Based on the trajectory and position feature data of the target object, combined with the crossing speed and the fluctuation range of the reflection intensity of the light curtain signal, compare the time nodes of the signal parameters with the trajectory change characteristics, statistically analyze the parameter correlation between the corresponding signal data and the geometric characteristics, and generate the characteristic matching result of the target object;

[0136] By calculating the average crossing time of the crossing path and the peak value of the fluctuation of the reflection intensity, analyze the degree of correlation with the spatial position characteristics of the trajectory, and use a two-dimensional correlation matrix to map the dynamic position changes of the trajectory data in the geometric characteristics, and finally generate the characteristic matching result of the target object.

[0137] Please refer to Figure 4 , and the specific steps of S3 are as follows:

[0138] S301: Based on the characteristic matching result of the target object, extract the occlusion time period and the changing area within the field of view of the elevator camera, detect the position of the boundary points of the occlusion area frame by frame, record the change value of the occlusion area on the time axis, statistically analyze the trend of the occlusion area change and the boundary stability, and generate the occlusion time and area change data;

[0139] Use the edge detection algorithm to extract the outer boundary of the occlusion area as a set of pixel points. Calculate the area of the minimum circumscribed rectangle of these boundary points frame by frame as the occlusion area value. Associate the area change of each frame with the time axis to form the time series data of the occlusion area. Conduct trend analysis on these data, statistically analyze the interval distribution of the change range of the occlusion area, calculate the stability of the occlusion boundary points, and characterize its change range through the mean and variance, and finally generate the occlusion time and area change data.

[0140] S302: Based on the occlusion time and area change data, extract the frequency and reflection intensity of the light curtain trigger signal within the occlusion time period, synchronously analyze the change range of the light curtain signal and the occlusion time in the time interval, calculate the overlapping ratio of the two, screen the characteristic intervals where the occlusion and the signal match, and generate the relationship data between the occlusion time and the light curtain trigger frequency;

[0141] The specific calculation formula for the overlapping intensity value between the occlusion and the light curtain signal is:

[0142]

[0143] Among them, R o represents the overlap intensity value between the occlusion and the light curtain signal, and ΔF k represents the change amount of the trigger frequency of the light curtain signal within the k-th time period, T k represents the duration of the k-th time period, B k represents the area value of the k-th occlusion area, π is the constant of pi, d represents the total number of occlusion time periods, m represents the total number of time periods when the signal is triggered, and T j represents the time value when the j-th signal is triggered.

[0144] The parameters involved are specifically as follows:

[0145] R o : The overlap intensity value between the occlusion and the light curtain signal, indicating the dynamic overlap degree between the two.

[0146] ΔF k : The change amount of the trigger frequency of the light curtain signal within the k-th time period. By monitoring the number of signal triggers of the light curtain sensor, the increment or decrement of the trigger frequency is calculated, and the unit is times per second.

[0147] T k : The duration of the k-th time period, obtained by recording the time interval of the occlusion, and the unit is seconds.

[0148] B k : The area value of the k-th occlusion area. By calculating the number of pixels in the occlusion area through image edge detection technology and converting it in combination with the pixel area (assuming the single pixel area is 0.01 square centimeters), the unit is square centimeters.

[0149] π: The constant pi, used to calculate the equivalent circular radius of the occlusion area, and the value is 3.14159.

[0150] d: The total number of occlusion time periods, indicating the total number of divided occlusion event segments.

[0151] m: The total number of time periods when the signal is triggered, indicating the total number of time periods of the light curtain signal trigger event.

[0152] T j : The time value when the j-th signal is triggered, obtained by recording the time of the light curtain signal, and the unit is seconds.

[0153] Parameter assignment and calculation:

[0154] The following parameters were monitored in a certain scenario: the change in the triggering frequency of the light curtain signal within the first period was 10 times per second, the duration was 5 seconds, and the occluded area was 31,416 square pixels; within the second period, the change in the triggering frequency of the light curtain signal was 15 times per second, the duration was 6 seconds, and the occluded area was 45,239 square pixels; within the third period, the change in the triggering frequency of the light curtain signal was 8 times per second, the duration was 4 seconds, and the occluded area was 20,106 square pixels. The time periods during which the light curtain signal was triggered were 5 seconds, 6 seconds, and 4 seconds respectively.

[0155] The calculation of the equivalent circular radius for the first occlusion is as follows:

[0156]

[0157] The area adjustment value is:

[0158] ΔF1·T1·r1 = 10·5·100 = 5000;

[0159] The calculation of the equivalent circular radius for the second occlusion is as follows:

[0160]

[0161] The area adjustment value is:

[0162] ΔF2·T2·r2 = 15·6·120 = 10800;

[0163] The calculation of the equivalent circular radius for the third occlusion is as follows:

[0164]

[0165] The area adjustment value is:

[0166] ΔF3·T3·r3 = 8·4·80 = 2560;

[0167] Calculation of the total occlusion overlap value:

[0168]

[0169] Calculation of the total signal time:

[0170]

[0171] Calculation of the overlap intensity value between the occlusion and the light curtain signal:

[0172]

[0173] Result description:

[0174] The overlapping intensity value of the occlusion and the light curtain signal is 1224, which represents the comprehensive correlation degree of the area, time of the occlusion event and the triggering frequency of the light curtain signal in terms of time and space. This value is further used to screen the characteristic intervals where the occlusion and the signal match, providing a quantitative basis for generating the relationship data between the occlusion time and the light curtain triggering frequency.

[0175] S303: Based on the relationship data between the occlusion time and the light curtain triggering frequency, extract the change of the occlusion area and the triggering parameters of the light curtain signal, conduct segmented comparison on the signal and the occlusion data, record the correlation between the signal anomaly and the area fluctuation at the time node, extract and classify the abnormal correlation behaviors, and generate the record result of the abnormal occlusion behavior chain;

[0176] Use the segmented analysis method to conduct pairwise comparison on the time series data of the occlusion area and the time node data of the light curtain signal, perform correlation calculation on the dynamic change value of the occlusion area and the fluctuation range of the triggering amplitude of the light curtain signal, record the fluctuation value of the associated occlusion area through the time node of the abnormal triggering of the light curtain signal, classify the associated behaviors, mark different fluctuation characteristics as independent behavior chains, and sort and classify the behavior chains in chronological order based on the statistical results of the fluctuations, and finally generate the record result of the abnormal occlusion behavior chain.

[0177] Please refer to Figure 5 , the specific steps of S4 are as follows:

[0178] S401: Based on the record result of the abnormal occlusion behavior chain, extract the elevator load change data before and after the door opening, statistically analyze the load change data for each frame on the time axis, record the change value of each frame of the load, calculate the load change trend, and mark the time points with significant changes, generating the load change and time series data;

[0179] First, calibrate the door opening time node on the time axis, collect the load data recorded frame by frame by the elevator load sensor, extract the change value of each frame of the load data, calculate the load change trend by accumulating the load difference between adjacent frames, mark the frames with a load change amplitude greater than the threshold as significant change time points, and generate the complete load change data in combination with the time series, and finally form the load change and time series data.

[0180] S402: Based on the load change and time series data, verify the load change value and the object trajectory data extracted by the light curtain, extract the synchronization of the load change event and the trajectory behavior in the time axis, screen the matching event chains according to the change range, and statistically analyze the event correlation characteristics, generating the load and trajectory event chain data;

[0181] Extract the synchronization of the load change event and the trajectory behavior, according to the formula:

[0182]

[0183] Calculate the correlation ratio between the load change event and the trajectory behavior. In the formula, ΔW i represents the change value of the load in the i-th segment, and ΔT i represents the time period of the load change, and ΔX j,t represents the displacement change value of the j-th segment of the trajectory behavior, and R v represents the correlation ratio between the load change and the trajectory behavior.

[0184] The correlation between the load change and the trajectory behavior is obtained by calculating the ratio of the load change value to the trajectory displacement change value. For each time interval, the change amount of the load and the cumulative displacement change amount of the trajectory behavior are counted, compared, and the synchronization ratio is calculated. In an example, assume that the time is divided into 3 segments, the load change values are 50, 40, and 60 (unit: kg) respectively, the corresponding time periods are 2, 3, and 1 (unit: s), and the trajectory behavior displacement change amounts are 10, 15, and 5 (unit: m) respectively. Then:

[0185] Cumulative value of load change:

[0186]

[0187] Cumulative value of trajectory behavior displacement:

[0188]

[0189] Correlation ratio between load change and trajectory behavior:

[0190]

[0191] This result shows that there is a strong correlation between the load change and the trajectory behavior. By analyzing this correlation ratio, matching event chains can be screened out, and load and trajectory event chain data can be generated.

[0192] S403: Based on the load and trajectory event chain data, analyze the corresponding relationship between the continuity of the behavior pattern and the occlusion characteristics in the event chain, calculate the correlation parameters between the trajectory change and the occlusion behavior, screen and sort out the abnormal behavior characteristics, and generate the electric vehicle abnormal behavior record;

[0193] Using the segmented comparison method of the trajectory position and the occlusion time series data, calculate the fluctuation range of the trajectory change amount and the occlusion behavior segment by segment, screen out the event segments related to the trajectory characteristics in the occlusion behavior, record and sort out the associated data of the abnormal behavior, and by classifying the characteristics of each abnormal behavior chain, mark the behaviors that do not conform to the correlation rules of the trajectory behavior continuity and the occlusion characteristics as abnormal, and finally generate the electric vehicle abnormal behavior record.

[0194] Please refer to Figure 6 , the specific steps of S5 are as follows:

[0195] S501: Based on the abnormal behavior records of the electric vehicle, transmit the record data to the elevator control mechanism, analyze the abnormal behavior characteristic data and corresponding control parameters, activate the locking function of the elevator door, trigger the voice warning device and record the response status, and generate elevator locking and voice warning response data;

[0196] By analyzing the behavior characteristic parameters included in the abnormal behavior record data, extract the key control parameters, use the elevator door lock control module to activate the locking function, call the voice warning device to trigger the warning sound effect, and at the same time record the response status of the locking function and the warning device. Corresponding storage of the response status data and control parameters forms elevator locking and voice warning response data, and finally generates a complete elevator control response data set.

[0197] S502: Based on the elevator locking and voice warning response data, upload the abnormal behavior records and elevator operation status data to the remote monitoring mechanism, establish a data transmission network interface and monitor the integrity of data transmission, record the feedback status information of the remote monitoring mechanism, and generate the remote monitoring abnormal behavior upload result;

[0198] Establish a data transmission network interface and monitor the transmission integrity. According to the formula:

[0199]

[0200] Calculate the ratio of data transmission integrity. In the formula, D k,t represents the transmission data volume of the k-th record, R k,t represents the received data volume of the corresponding feedback, and I t represents the ratio of data transmission integrity.

[0201] The data transmission integrity is calculated by monitoring the cumulative ratio of the difference between the transmitted data volume and the received data volume. For each transmission record, count its transmission volume and received volume, calculate the cumulative difference, and divide it by the cumulative value of the transmission volume to obtain the data transmission integrity ratio. In an example, assume that the number of records is 3, the transmitted data volumes are 1000, 800, 600 (unit: bytes) respectively, and the received data volumes are 980, 760, 580 (unit: bytes) respectively. Then:

[0202] Cumulative transmission difference:

[0203]

[0204] Cumulative transmission volume:

[0205]

[0206] Data transmission integrity ratio:

[0207]

[0208] The result shows that the data transmission integrity reaches 96.7%. By analyzing this ratio, the data transmission status of the remote monitoring mechanism can be monitored, and the upload result of remote monitoring abnormal behavior can be generated.

[0209] S503: Based on the upload result of remote monitoring abnormal behavior, link the building alarm mechanism, analyze the abnormal behavior data to trigger the alarm condition and synchronize the control parameters, start the alarm device in the building and record the alarm response time and execution status, and generate the elevator abnormal state event response result;

[0210] By analyzing the alarm trigger condition and the associated control parameters in the abnormal behavior data, extract the abnormal feature information associated with the alarm, call the alarm device module to start the alarm sound effect, calibrate the specific time node of alarm trigger, and record the operation response status of the alarm module, form the record data of alarm trigger time and execution status, and combine the trigger condition parameters with the linkage response rule of the building alarm mechanism to generate the elevator abnormal state event response result.

[0211] Please refer to Figure 7 , an elevator electric vehicle identification system based on anomaly detection, including:

[0212] The object edge extraction module extracts the object edge contour frame by frame based on the picture data of the elevator camera and the light curtain infrared signal data, calculates the contour closure parameter, correlates the light curtain crossing path width and reflection intensity data, and generates the object geometric characteristic determination result;

[0213] The target characteristic analysis module analyzes the width-height ratio of the target object and the change range of the continuous motion trajectory based on the object geometric characteristic determination result, statistically analyzes the light curtain signal crossing speed and reflection intensity fluctuation, compares the geometric shape data with the speed fluctuation, and generates the target object characteristic matching result;

[0214] The abnormal occlusion analysis module extracts the elevator camera field of view occlusion time period and change area based on the target object characteristic matching result, statistically analyzes the light curtain trigger frequency and reflection change during the occlusion time, calculates the occlusion duration and frequency overlap rate, and conducts comparative analysis in combination with the occlusion data to generate the abnormal occlusion behavior chain record result;

[0215] The load behavior verification module analyzes the load change before and after the door opening based on the abnormal occlusion behavior chain record result, verifies the load change value and the object trajectory data extracted by the light curtain, screens the event chain with consistent trajectory behavior and load change, and generates the electric vehicle abnormal behavior record;

[0216] The abnormal event response module transmits the record data to the elevator control mechanism based on the electric vehicle abnormal behavior record, triggers the elevator door lock and issues a voice warning, links the building alarm mechanism, and generates the elevator abnormal state event response result.

[0217] As described above, it is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims described above.

Claims

1. A method for identifying elevator electric vehicles based on abnormality detection, characterized in that: The following steps are involved: S1: Based on the image data of the elevator camera and the infrared signal data of the light curtain, the edge contour of the object during the door opening is extracted frame by frame, the width-to-height ratio of the object is analyzed, the contour closure parameters are calculated, the width of the light curtain crossing path and the reflection intensity data are associated, the object morphological characteristics and signal path are identified and processed, and the object geometric characteristics determination result is generated; S2: Based on the determination result of the geometric characteristics of the object, the width-to-height ratio and the range of change of the continuous motion trajectory of the target object are analyzed, the object morphological characteristics and position offset data are combined, the light curtain signal crossing speed and reflection intensity fluctuation are counted, the geometric morphological data and speed fluctuation are compared, and the target object characteristic matching result is generated; S3: Based on the target object characteristic matching result, extract the elevator camera field of view occlusion time period and change area, count the light curtain triggering frequency and reflection change during the occlusion time, analyze the occlusion duration and frequency, calculate the occlusion duration and frequency overlap rate, combine the occlusion data for comparative analysis, and generate abnormal occlusion behavior chain record results; S4: Based on the abnormal occlusion behavior chain record results, analyze the load change before and after the door is opened, verify the load change value and the object trajectory data extracted by the light curtain, select the event chain whose trajectory behavior is consistent with the load change, analyze the behavior pattern and occlusion characteristics in the event chain, and generate an abnormal behavior record of the electric vehicle; S5: Based on the abnormal behavior record of the electric vehicle, the recorded data is transmitted to the elevator control mechanism, the elevator door is locked and a voice warning is issued, the data is uploaded to the remote monitoring mechanism, the building alarm mechanism is linked, and the elevator abnormal state event response result is generated; The specific steps of step S1 are: S101: Based on the image data of the elevator camera and the infrared signal data of the light curtain, the edge contours of the objects in the image during the door opening are extracted frame by frame, the object and the background area are separated by edge detection, the width and height ratio of the object edge is recorded, and the change characteristics of the edge morphology on the time axis are calculated to generate the object edge morphology data set; S102: Based on the object edge morphology data set, calculate the edge closure deviation value, compare the geometric positions of the edge start point and the edge end point, and judge the contour closure, analyze the spatial continuity of the closure change between frames, perform curve fitting on the edge closure degree, and record key feature points to generate edge contour closure feature data; S103: Based on the edge contour closure feature data, combined with the crossing path width and reflection intensity information of the light curtain infrared signal data, the spatial correlation between the signal path and the contour closure parameters is analyzed, and the object feature mapping relationship is constructed through a matrix of multi-dimensional characteristic data to generate the object geometric characteristic judgment result.

2. The elevator electric vehicle identification method based on abnormality detection according to claim 1 is characterized in that: The object geometric characteristic determination result includes object edge contour parameters, aspect ratio analysis results, contour closure parameters, crossing path width characteristics, and reflection intensity characteristics; the target object characteristic matching result includes object aspect ratio characteristics, motion trajectory range characteristics, position offset data, crossing speed characteristics, and reflection intensity fluctuation characteristics; the abnormal occlusion behavior chain record result includes occlusion time period data, occlusion area change characteristics, light curtain signal trigger frequency statistics, reflection change accumulation characteristics, and occlusion duration frequency overlap rate; the electric vehicle abnormal behavior record includes load change characteristics, trajectory behavior chain, event chain pattern characteristics, and occlusion characteristic analysis results; the elevator abnormal state event response result includes door lock trigger signal, voice warning command, remote monitoring upload data, and building alarm linkage mechanism.

3. The elevator electric vehicle identification method based on abnormality detection according to claim 1 is characterized in that: The edge closure deviation value calculation formula is specifically: Among them, C b represents the closure deviation value, Represents the horizontal coordinate of the starting point of the edge of the i-th frame, Represents the ordinate of the starting point of the edge of the i-th frame, Represents the horizontal coordinate of the end point of the edge of the i-th frame, Represents the ordinate of the end point of the edge of the i-th frame, A i Represents the area of ​​the closed region in the i-th frame, n represents the total number of frames, and π is the pi constant.

4. The elevator electric vehicle identification method based on abnormality detection according to claim 1 is characterized in that: Based on the determination result of the geometric characteristics of the object, the specific steps of analyzing the width-to-height ratio of the target object and the range of change of the continuous motion trajectory, combining the object morphological characteristics and position offset data, counting the light curtain signal crossing speed and reflection intensity fluctuations, and comparing the geometric morphological data with speed fluctuations to generate the target object characteristic matching result are as follows: S201: extracting the range of the target object's aspect ratio change based on the object geometric characteristic determination result, grouping the target object according to the width and height data between frames, calculating the ratio change of each group, statistically analyzing the frequency distribution of the ratio change interval, extracting the key points of the continuous change trend on the time axis, and generating the target object's aspect ratio change data; S202: Based on the target object aspect ratio change data, combined with the motion trajectory and position offset data, analyzing the change amplitude of each frame position in the trajectory, calculating the cumulative displacement in the change direction, screening key points with abnormal offset characteristics for the trajectory curve, and generating the target object trajectory and position feature data; S203: Based on the target object trajectory and position feature data, combined with the crossing speed and reflection intensity fluctuation range of the light curtain signal, the time nodes of the signal parameters and the trajectory change characteristics are compared, and the parameter correlation between the corresponding signal data and the geometric characteristics is statistically analyzed to generate a target object characteristic matching result.

5. The elevator electric vehicle identification method based on abnormality detection according to claim 1 is characterized in that: Based on the target object characteristic matching results, extract the elevator camera field of view occlusion time period and change area, count the light curtain trigger frequency and reflection change during the occlusion time, analyze the occlusion duration and frequency, calculate the occlusion duration and frequency overlap rate, combine the occlusion data for comparative analysis, and generate the abnormal occlusion behavior chain record results in the following specific steps: S301: Based on the target object characteristic matching result, extract the occlusion time period and change area in the elevator camera field of view, detect the boundary point position of the occlusion area frame by frame, record the area change value of the occlusion area on the time axis, count the trend of the occlusion area change and the boundary stability, and generate occlusion time and area change data; S302: Based on the shading time and area change data, extract the frequency and reflection intensity of the light curtain trigger signal within the shading time period, synchronously analyze the light curtain signal change amplitude and the shading time in the time interval, calculate the overlap ratio between the two, select the characteristic interval matching the shading and the signal, and generate the shading time and light curtain trigger frequency relationship data; S303: Based on the relationship data between the occlusion time and the light curtain triggering frequency, extract the change in the occlusion area and the light curtain signal triggering parameters, perform segmented comparison on the signal and occlusion data, record the correlation between the signal anomaly and the regional fluctuation at the time node, extract and classify the abnormal related behaviors, and generate the abnormal occlusion behavior chain record results.

6. The elevator electric vehicle identification method based on abnormality detection according to claim 5 is characterized in that: The calculation formula of the overlap intensity value of the shielding and light curtain signals is specifically as follows: Among them, R o Represents the overlapping intensity value of the occlusion and light curtain signal, ΔF k Represents the trigger frequency change of the light curtain signal during the kth period of time, T k represents the duration of the kth period, B k represents the area value of the kth occlusion area, π is the pi constant, d represents the total number of occlusion time periods, m represents the total number of signal triggering time periods, T j Represents the time value when the jth signal is triggered.

7. The elevator electric vehicle identification method based on abnormality detection according to claim 1 is characterized in that: Based on the abnormal occlusion behavior chain record results, analyze the load change before and after the door is opened, verify the load change value and the object trajectory data extracted by the light curtain, select the event chain with consistent trajectory behavior and load change, analyze the behavior pattern and occlusion characteristics in the event chain, and generate the specific steps of abnormal behavior record of electric vehicles as follows: S401: Based on the abnormal occlusion behavior chain recording results, extract the elevator load change data before and after the door is opened, statistically analyze the load change data of each frame on the time axis, record the load change value of each frame, calculate the load change trend, and mark the time point of significant change to generate load change and time series data; S402: Based on the load change and time series data, verify the load change value and the object trajectory data extracted by the light curtain, extract the synchronization of the load change event and the trajectory behavior in the time axis, filter and match the event chain according to the change range, and count the event correlation characteristics to generate load and trajectory event chain data; S403: Based on the load and trajectory event chain data, the corresponding relationship between the continuity of the behavior pattern in the event chain and the occlusion characteristics is analyzed, the trajectory change and occlusion behavior association parameters are calculated, the abnormal behavior characteristics are screened and sorted, and the abnormal behavior record of the electric vehicle is generated.

8. The elevator electric vehicle identification method based on abnormality detection according to claim 1 is characterized in that: Based on the abnormal behavior record of the electric vehicle, the recorded data is transmitted to the elevator control mechanism, the elevator door is locked and a voice warning is issued, the data is uploaded to the remote monitoring mechanism, the building alarm mechanism is linked, and the specific steps of generating the elevator abnormal state event response result are: S501: Based on the abnormal behavior record of the electric vehicle, the recorded data is transmitted to the elevator control mechanism, the abnormal behavior feature data is analyzed and the corresponding control parameters are obtained, the locking function of the elevator door is started, the voice warning device is triggered and the response status is recorded, and the elevator locking and voice warning response data are generated; S502: Based on the elevator locking and voice warning response data, upload abnormal behavior records and elevator operation status data to the remote monitoring mechanism, establish a data transmission network interface and monitor data transmission integrity, record feedback status information of the remote monitoring mechanism, and generate remote monitoring abnormal behavior upload results; S503: Based on the remote monitoring abnormal behavior upload result, the building alarm mechanism is linked, the abnormal behavior data is analyzed to trigger the alarm condition and synchronize the control parameters, the alarm device in the building is started and the alarm response time and execution status are recorded, and the elevator abnormal status event response result is generated.

9. An elevator electric vehicle identification system based on abnormality detection, characterized in that: According to the method for identifying an elevator electric vehicle based on abnormality detection according to any one of claims 1 to 8, the system comprises: The object edge extraction module extracts the edge contour of the object during the door opening frame by frame based on the image data of the elevator camera and the infrared signal data of the light curtain, calculates the contour closure parameters, associates the width of the light curtain crossing path with the reflection intensity data, and generates the object geometric characteristic determination result; The target characteristic analysis module analyzes the width-to-height ratio and the range of change of the continuous motion trajectory of the target object based on the determination result of the geometric characteristics of the object, counts the crossing speed of the light curtain signal and the fluctuation of the reflection intensity, compares the geometric shape data with the speed fluctuation, and generates the target object characteristic matching result; The abnormal occlusion analysis module extracts the occlusion time period and change area of ​​the elevator camera's field of view based on the target object characteristic matching result, counts the light curtain triggering frequency and reflection change during the occlusion time, calculates the occlusion duration and frequency overlap rate, and compares and analyzes the occlusion data to generate abnormal occlusion behavior chain record results; The load behavior verification module analyzes the load change before and after the door is opened based on the abnormal occlusion behavior chain record results, verifies the load change value and the object trajectory data extracted by the light curtain, selects the event chain with consistent trajectory behavior and load change, and generates an abnormal behavior record of the electric vehicle; The abnormal event response module transmits the recorded data to the elevator control mechanism based on the abnormal behavior record of the electric vehicle, triggers the elevator door to lock and issue a voice warning, links the building alarm mechanism, and generates an elevator abnormal state event response result.

Images