Elevator door lock detection method and system based on high-definition camera

Abstract

The application relates to an elevator door lock detection method and system based on a high-definition camera, in particular to the technical field of elevator safety detection, the method comprises the following steps: after receiving an installation acceptance instruction, image collection is performed on the elevator door lock according to preset detection parameters, a plurality of door lock initial images are obtained, acceptance compliance detection is performed according to the door lock initial images, and an acceptance detection result is obtained; a standard reference library and initial full-life-cycle reference data are established based on the acceptance reference length corresponding to each floor door; after receiving a daily detection instruction, full-life-cycle linkage type daily detection and trend analysis are performed, a daily detection result and a full-life-cycle safety trend report are obtained; the application realizes non-contact and automatic high-precision detection of the elevator door lock through visual detection, effectively solves the low-efficiency problem existing in the current manual detection, and can give early warning on safety hazards through the construction of a full-life-cycle dynamic monitoring system.

Description

Technical Field

[0001] This invention relates to the field of elevator safety inspection technology, and in particular to an elevator door lock detection method and system based on a high-definition camera. Background Technology

[0002] As a core safety component of elevator operation, the elevator landing door lock's locking reliability directly affects personal safety during elevator use. Elevator landing door locks typically consist of a movable hook installed on the active door and a fixed hook installed on the driven door. The hook is equipped with an electrode spring or conductive needle. When the hook is fully engaged with the fixed hook, the electrode spring / conductive needle conducts electricity with the conductive component inside the fixed component, allowing the elevator to start. The length of the hook's engagement with the fixed component is a key indicator of locking reliability. Industry standards require this engagement length to be at least 7 millimeters to effectively prevent unauthorized personnel from prying open the landing door from the outside, thus avoiding accidents such as falls and people being trapped.

[0003] During elevator installation and daily use, factors such as installation errors, long-term wear and tear, and mechanical deviations can easily lead to situations where the locking hook is not fully engaged with the fixing component, but the electrode spring / conductive needle has already contacted the conductive component and conducted electricity. In this case, the elevator can start normally, but the landing door locking strength is insufficient, posing a significant safety hazard. Currently, the main method for detecting this problem is manual inspection. Based on the aforementioned manual inspection requirements, some manufacturers have made some basic optimizations: most elevator landing door locks on the market are open-top designs, while some locks are equipped with transparent protective covers to allow inspectors to directly observe the internal locking hook structure. In this case, the inspector must stand on the top wall of the elevator car, manually pry open the locking hook, and observe and measure the length of the locking hook engaged with the fixing component at the moment the electrode spring / conductive needle conducts electricity. This inspection method has many drawbacks:

[0004] First, manual operation poses safety risks due to working at heights, and the inspection process requires frequent starting and stopping of the elevator, which will affect its normal use. Second, the judgment results of manual measurement are greatly affected by subjective factors, and the accuracy is difficult to guarantee. Third, the inspection efficiency is extremely low, as each door lock on each floor of the elevator needs to be manually inspected, which is time-consuming and labor-intensive. Summary of the Invention

[0005] To improve the stability of current manual inspection methods for elevator door locks in detecting the daily operational safety of elevators, this application provides an elevator door lock inspection method and system based on a high-definition camera.

[0006] Firstly, this application provides an elevator door lock detection method based on a high-definition camera, employing the following technical solution:

[0007] An elevator door lock detection method based on a high-definition camera, the elevator door lock detection method being based on an elevator door lock detection system, the system including a camera for capturing images of the door lock, the method comprising:

[0008] Upon receiving the installation and acceptance instruction, images of the elevator door locks are acquired according to the preset testing parameters to obtain several initial images of the door locks. Acceptance compliance testing is then conducted based on these initial images to obtain the acceptance test results.

[0009] A standard benchmark library and initial full life cycle benchmark data are established based on the acceptance benchmark length corresponding to each door layer.

[0010] Upon receiving a routine inspection instruction, the system performs a full lifecycle-linked routine inspection and trend analysis to obtain routine inspection results and a full lifecycle safety trend report.

[0011] By adopting the above technical solution, this application breaks through the current limitations of accuracy and efficiency of manual inspection, realizes the full-process automated inspection of elevator door locks from installation and acceptance to daily maintenance, and provides a unified quantitative reference for door lock inspection by establishing a standard benchmark library and full life cycle benchmark data, breaks the data silos of single inspections, provides a compliant data foundation for subsequent trend analysis, and greatly improves the continuity and standardization of elevator door lock inspection.

[0012] In one specific implementation scheme, the system specifically includes a main-view camera and a secondary-view camera installed based on a preset dual-view coordinate transformation rule. The initial image of the door lock includes a first initial image corresponding to the main view and a second initial image corresponding to the secondary view. The second initial image is used to calibrate and correct the first initial image by using the projection relationship between a pixel in the second initial image and a corresponding pixel in the first initial image. The step of acquiring images of the elevator door lock according to preset detection parameters to obtain several initial door lock images specifically includes:

[0013] The elevator door lock is synchronously photographed from two perspectives according to preset detection parameters to obtain the first initial image and the second initial image. The detection parameters are used to limit and unify the shooting of the main perspective camera and the secondary perspective camera at different consecutive moments.

[0014] By adopting the above technical solution, this application ensures the temporal consistency of dual-view images at the same moment through the synchronous shooting configuration of primary and secondary dual-view cameras. At the same time, the secondary view image is used as the calibration basis for the primary view image, providing a hardware and data foundation for subsequent elimination of perspective distortion in the primary view image. The unified shooting parameters also ensure the comparability of images acquired at different times, effectively avoiding detection errors caused by fluctuations in shooting parameters.

[0015] In one specific implementation, the synchronous shooting is performed multiple times within a continuous period after the elevator door opens, with each continuous period corresponding to one first initial image and one second initial image; the acceptance and compliance inspection based on the initial door lock image, to obtain the acceptance and inspection results, includes:

[0016] The initial actual length of the door lock hook corresponding to each of the consecutive time moments is obtained based on the first initial image and the second initial image.

[0017] The acceptance benchmark length is obtained based on the initial actual length at all consecutive moments, and the acceptance benchmark length is subjected to acceptance compliance testing to obtain the acceptance test results.

[0018] By adopting the above technical solution, this application covers the core detection window when the door lock hook just disengages from the locking state by synchronously capturing multiple frames at continuous moments after the elevator door opens, ensuring that image data reflecting the actual locking performance of the door lock can be captured; at the same time, the acceptance benchmark length is calibrated based on the initial actual length at multiple continuous moments, avoiding the random errors of single-frame images and improving the accuracy and reliability of the acceptance benchmark length.

[0019] In one specific implementation, obtaining the initial actual length of the door lock hook corresponding to each of the consecutive time moments based on the first initial image and the second initial image includes:

[0020] Valid pixels are filtered based on the first initial image and the second initial image at the same consecutive time.

[0021] After initial screening, the effective pixels are analyzed and merged to obtain the set of effective pixels from the main viewpoint and the set of effective pixels from the sub-viewpoint at the same continuous time. The analysis and merging of the effective pixels are achieved by structuring the coordinates of the effective pixels and constructing a grid index.

[0022] Based on the effective pixel set of the main view and the effective pixel set of the secondary view, all key feature points corresponding to each consecutive time moment and key coordinates corresponding to each key feature point are obtained. The key feature points are used to calculate the hook insertion length of the door lock.

[0023] The initial actual length of the door lock hook at each consecutive moment is calculated based on the key coordinates.

[0024] By adopting the above technical solution, this application eliminates invalid interference pixels such as noise and reflection in the image through effective pixel point screening. Then, through coordinate structuring and grid indexing to construct connected region analysis and merging, the effective pixel area corresponding to the door lock component is accurately located. Isolated noise points and background interference pixels are eliminated as much as possible, providing a clean and reliable pixel data foundation for the subsequent accurate positioning of key feature points and length calculation, effectively improving the accuracy of length calculation.

[0025] In a specific implementation scheme, the key feature points include the end point of the hook head and the reference point on the inner side of the fastener, and the key coordinates include the key hook head coordinates; obtaining all key feature points corresponding to each consecutive time moment and the key coordinates corresponding to each key feature point based on the effective pixel set of the main view and the effective pixel set of the secondary view includes:

[0026] The continuous contour of the door lock component within the effective pixel set of the main viewpoint is extracted by the edge detection algorithm to obtain the first lock hook contour pixel subset;

[0027] The continuous contour of the door lock component within the effective pixel set of the secondary viewpoint is extracted by the edge detection algorithm to obtain the second lock hook contour pixel subset;

[0028] Locate the endpoint of the hook head in the first hook contour pixel subset to obtain the reference hook head coordinates;

[0029] Based on the dual-view coordinate transformation rule, the projected hook head coordinates corresponding to the second hook contour pixel subset are obtained according to the reference hook head coordinates;

[0030] Locate the actual hook head endpoint corresponding to the projected hook head coordinates in the second hook contour pixel subset to obtain the actual hook head coordinates;

[0031] Calculate the Euclidean distance between the projected hook head coordinates and the actual hook head coordinates, and determine whether to trigger the coordinate correction mechanism based on the Euclidean distance to obtain the key hook head coordinates.

[0032] By adopting the above technical solution, this application accurately separates the hook contour pixels through edge detection algorithm. Using the projection coordinates of the main view reference coordinates onto the secondary view as a reference, and combining the real coordinates of the actual contour of the secondary view, the coordinate deviation is quantified by Euclidean distance. This achieves the deviation verification of the coordinates of key feature points in the main view, providing an accurate basis for subsequent correction of perspective distortion and further ensuring the accuracy of the coordinates of key feature points.

[0033] In one specific implementation scheme, in the first subset of lock hook contour pixels, the effective pixel point farthest from the fastener contour is taken as the lock hook head endpoint, and the midpoint of the consecutive effective pixels with the smallest abscissa in the fastener contour is taken as the inner reference point of the fastener.

[0034] In one specific implementation scheme, the detection parameters include calculating the allowable error; the step of determining whether to trigger the coordinate correction mechanism based on the Euclidean distance to obtain the coordinates of the key lock head includes:

[0035] If the Euclidean distance between the projected hook head coordinates and the actual hook head coordinates is less than or equal to the allowable error, then the reference hook head coordinates are directly used as the key hook head coordinates.

[0036] Otherwise, a coordinate correction mechanism is triggered, and the actual hook head coordinates corresponding to the second hook contour pixel subset are converted into the corresponding precise coordinates of the first hook contour pixel subset through the pre-calibrated inverse homography matrix H-1, which are then used as the corrected key hook head coordinates.

[0037] By adopting the above technical solution, this application sets the trigger threshold for coordinate correction by pre-setting the allowable calculation error, and realizes the hierarchical processing of coordinate deviation. When the deviation is controllable, the reference coordinate is directly used to ensure calculation efficiency. When the deviation exceeds the limit, the coordinate is accurately corrected by the pre-calibrated inverse homography matrix, which greatly eliminates the perspective distortion error of the main view image, while taking into account both detection efficiency and calculation accuracy.

[0038] In a specific feasible implementation, upon receiving a routine inspection instruction, the process of performing a full lifecycle-linked routine inspection and trend analysis to obtain routine inspection results and a full lifecycle safety trend report includes:

[0039] Receive daily inspection instructions and load the target elevator's full lifecycle historical dataset, which includes the actual hook length and the acceptance benchmark length for all historical cycles.

[0040] Perform standardized image acquisition to obtain several daily images of door locks, and calculate the actual hooking length of the door lock hook in the current period based on all the daily images of door locks to obtain the daily detection results;

[0041] Based on the historical dataset of the entire life cycle, perform a full life cycle trend analysis to obtain target feature parameters that reflect the degree of decay in the locking performance of the door lock, and generate a full life cycle security trend report based on the target feature parameters.

[0042] By adopting the above technical solution, this application loads a full lifecycle historical dataset containing acceptance benchmark length and historical test data, ensuring the consistency of testing standards between daily testing and installation acceptance. Based on the full lifecycle trend analysis of historical data, it realizes the quantitative characterization of the degree of door lock locking performance degradation, providing data support for the dynamic assessment of door lock security status, and breaking through the limitation that a single test can only determine the current compliance.

[0043] Secondly, this application provides an elevator door lock detection system based on a high-definition camera, employing the following technical solution:

[0044] An elevator door lock detection system based on a high-definition camera includes:

[0045] The data acquisition module is used to acquire door lock images at several consecutive moments after the elevator door opens;

[0046] The data processing module is used to receive the door lock image and perform compliance detection on the door lock image using the elevator door lock detection method described in the first aspect above, so as to obtain the detection result.

[0047] Thirdly, this application provides a computer-readable storage medium, which adopts the following technical solution:

[0048] The readable storage medium stores at least one instruction, at least one program, code set, or instruction set, wherein the at least one instruction, at least one program, code set, or instruction set is loaded and executed by a processor to implement an elevator door lock detection method based on a high-definition camera as described in any of the first aspects.

[0049] In summary, this application includes at least one of the following beneficial technical effects:

[0050] 1. This application uses the lock hook stereo image acquired from the side of the secondary view as the calibration benchmark. It achieves perspective correction of key feature points of the main view through the pre-calibrated homography matrix, effectively eliminating perspective distortion and coordinate offset caused by camera installation errors and elevator start-stop jitter. This greatly improves the detection accuracy of the lock hook insertion length and effectively solves the problems of contour distortion and large measurement error that are easy to occur in traditional single vision detection and manual detection.

[0051] 2. This application uses the acceptance benchmark length determined during the installation and acceptance phase as the initial anchor point to construct a benchmark data system for the entire life cycle of elevator door locks. In daily testing, the degree of decay of door lock locking performance is quantified through target feature parameters, and dynamic trend analysis of door lock locking performance is completed. This breaks through the limitation of traditional testing, which can only complete a single static compliance judgment, and realizes the upgrade from "post-fault rectification" to "pre-fault hazard prediction". It helps to identify hidden safety risks such as door lock mechanical wear and loose installation in advance, and significantly improves the safety redundancy of elevator door lock operation. Attached Figure Description

[0052] Figure 1 This is a schematic diagram of the structure of an elevator door lock detection system based on a high-definition camera, according to an embodiment of this application.

[0053] Figure 2 This is a flowchart illustrating another embodiment of an elevator door lock detection method based on a high-definition camera. Detailed Implementation

[0054] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be further described in detail below with reference to the accompanying drawings.

[0055] The following describes in further detail, with reference to the accompanying drawings, an embodiment of an elevator door lock detection method and system based on a high-definition camera, according to this application.

[0056] One embodiment of this application discloses an elevator door lock detection system based on a high-definition camera. This detection system uses a high-definition camera to build a detection device, realizing automated and non-contact detection of the hook insertion length of the door lock on each floor of the elevator. The detection process is non-destructive, highly accurate, and efficient. It is suitable for elevator door locks on the market where the hook can be directly observed, and can be widely used in elevator installation and acceptance, daily maintenance, periodic inspection and other scenarios.

[0057] Reference Figure 1 An elevator door lock detection system based on a high-definition camera includes a data acquisition module, a data processing module, and a display output module, wherein:

[0058] The data acquisition module is the core of the entire system's image acquisition, used to acquire door lock images at specific moments, referring to several consecutive moments after the elevator door opens. Specifically, the data acquisition module is used to simultaneously acquire door lock images from two perspectives, and includes a detachable telescopic bracket, a high-definition industrial camera, a positioning and calibration component, and a power supply unit.

[0059] Two identical high-definition industrial cameras are fixed to the end of a detachable telescopic bracket, defined as a main-view camera and a secondary-view camera respectively. The two cameras are synchronously triggered to continuously capture images of the door lock from different angles the moment the elevator door opens. In this embodiment, the high-definition industrial cameras have a resolution of ≥2K, a frame rate of ≥30fps, and are equipped with fixed-focus lenses with no deviation after the focus is locked. The main-view camera is responsible for facing the door lock's mating plane from the front, capturing the frontal outline of the lock hook and fixing parts for subsequent length calculation. The secondary-view camera is responsible for shooting the door lock hook from the side, capturing the three-dimensional shape of the lock hook as a reference for the main-view coordinate calibration, which greatly eliminates the single-view perspective distortion error.

[0060] The detachable telescopic bracket is used to adjust the camera of the high-definition industrial camera towards the door lock structure of the elevator landing door through the telescopic drive component to complete the shooting. The telescopic drive component is an example of the electric push rod type in the prior art. In this embodiment, the detachable telescopic bracket adopts a magnetic and snap-on composite installation structure. It is installed non-destructively on the top wall of the elevator car near the landing door at a preset position. The telescopic stroke and telescopic accuracy of the bracket are adjustable. It can flexibly adjust the extension length according to the door lock installation height of different brands and specifications of elevators, so that the camera is at the optimal shooting distance (i.e., it can clearly shoot the complete door lock structure). The telescopic stroke of the bracket is based on meeting the detection requirements. In this embodiment, it is 0-50cm as an example.

[0061] The positioning calibration component is integrated into the base of the high-definition industrial camera. It limits the camera's shooting angle and focal length until the camera's mating plane with the door lock hook and fixing components forms a 90° angle. This effectively avoids perspective distortion causing contour distortion or shooting parameter shifts due to elevator vibration. Specifically, the positioning calibration component in this embodiment includes an angle limiting component and a focal length locking component. The angle limiting component consists of five parts: a rotating adjustment base, an angle scale, a 90° positioning block, a locking knob, and a spirit level. This is existing technology and will not be described in detail here. The specific operation process is as follows:

[0062] For the main view camera: First, loosen the horizontal and tilt locking knobs, rotate the rotary adjustment base to align the camera with the door lock mating plane, slowly adjust the tilt angle until the rotating end fits against the 90° positioning block, at which point the pointer points to the 90° tilt scale, observe the level bubble and center the bubble, and finally tighten the two locking knobs to complete the hard locking of the main view and the door lock mating plane at a 90° perpendicularity.

[0063] For the secondary view camera: Using the same set of calibration components and following the parallel calibration logic, adjust the secondary view camera to a position that is 90° parallel to the plane of the door lock. After confirming the horizontal position with the level bubble calibration component, tighten the locking knob to complete the hard locking of the angle. After the angle of both machines is locked, there is no adjustment margin, and the shaking of the elevator will not cause the angle to shift.

[0064] The focus locking mechanism consists of four parts: a focus adjustment dial, a scale ring, a locking clamp, and an anti-loosening screw. This is existing technology and will not be described in detail here. The specific operating procedure is as follows:

[0065] When establishing the standard reference library, adjust the focus adjustment dial to clearly image the outline of the door lock of the standard sample, record the corresponding scale value of the scale ring, then tighten the bolt of the locking clamp, lock the focus adjustment dial, and finally tighten the anti-loosening screw to complete the anti-loosening treatment; when inspecting door locks on each floor in the future, there is no need to adjust the focus, the lens focus is always locked at the scale value of the reference shooting, ensuring that the imaging ratio and magnification of all captured images are completely consistent, effectively eliminating the pixel size error caused by focus deviation.

[0066] The power supply unit powers the telescopic drive components of the detachable telescopic bracket and the high-definition industrial camera. It uses a rechargeable lithium battery pack, supports fast charging and long battery life, and is suitable for testing environments in elevators where there is no external power source.

[0067] The data processing module is electrically connected to the data acquisition module. It is used to acquire the door lock image in the data acquisition module and perform image algorithm processing based on the preset standard reference image and the door lock image to complete the quantitative solution of the hook insertion length and compliance judgment, and obtain the acceptance test results.

[0068] The display output module is used to realize real-time monitoring of the testing process and visualization of the acceptance test results. It includes an on-site display terminal and a report printing output unit. Specifically, it can use hardware carriers such as touch screens and industrial displays, and supports wired / wireless bidirectional communication with the data processing module.

[0069] The on-site display terminal can display the images captured by the data acquisition module, the inspection progress of door locks on each floor, and the judgment results of the inspected floors in real time. It also supports manual interactive operation, including fine-tuning of inspection parameters, selection of inspection floors, magnification of images, and marking of abnormal results, so that on-site inspection personnel can keep abreast of the inspection status. The report printing and output unit can automatically generate a standardized "Elevator Floor Door Lock Hook Engagement Length Inspection Report" based on the calculation results of the data processing module. The report includes basic information of the inspected elevator, inspection equipment parameters, actual engagement length values ​​of door locks on each floor, and compliance judgment results (qualified / unqualified). For unqualified floors, the abnormality reasons and corresponding inspection images are marked simultaneously. The report supports local printing, PDF export, and cloud sharing, providing formal inspection basis for elevator maintenance and acceptance.

[0070] The implementation of the method will be explained in detail below with reference to the above system:

[0071] Reference Figure 2 Another embodiment of this application provides an elevator door lock detection method based on a high-definition camera, including:

[0072] S100, after receiving the installation and acceptance instruction, performs dual-view image acquisition on the elevator door lock according to the preset detection parameters, obtains several initial images of the door lock from different perspectives, and performs acceptance compliance detection based on the initial images of the door lock to obtain the acceptance detection results;

[0073] The initial image of the door lock includes a first initial image corresponding to the main view and a second initial image corresponding to the secondary view; specifically, S100 includes:

[0074] S110 receives the installation acceptance command and loads the preset test parameters;

[0075] S110 is mainly used to complete the instruction parsing and standardized parameter configuration before testing to ensure the uniformity of testing parameters throughout the process and provide a compliance basis for subsequent benchmark establishment. Specifically, it receives and parses the issued installation and acceptance instructions to obtain target information such as the target elevator ID, total number of floors, door lock model, and acceptance execution standard. Then, it loads the preset testing parameters matching the elevator door lock model from the storage unit of the data processing module and performs initial filing of the target elevator using the target information and preset testing parameters. The testing parameters include, but are not limited to, the resolution, frame rate, focal length of the fixed-focus lens, and sampling trigger mode of the main / secondary view high-definition industrial camera (for the convenience of subsequent image comparison, this embodiment takes dual-camera synchronous continuous triggering as an example), the judgment benchmark parameters for the length of the door lock hook (minimum safe hook length, calculation allowable error, etc.), and the telescopic stroke, telescopic step accuracy, and telescopic action response time of the detachable telescopic bracket. It is mainly used to limit the shooting of the unified main view camera and the secondary view camera at different continuous moments.

[0076] It should be noted that in this implementation, the main process involves loosening the horizontal and tilt locking knobs, rotating the rotary adjustment base to align the main-view camera with the door lock mating plane, slowly adjusting the tilt angle until the rotating end is in contact with the 90° positioning block, at which point the dial pointer points to the 90° tilt mark, observing the bubble level to ensure the bubble is completely centered, and finally tightening the locking knob to complete the hard locking of the 90° vertical angle. Simultaneously, using the same calibration logic as the main view camera, the secondary view camera is adjusted to a position 90° parallel to the door lock mating plane. After confirming the horizontal position using the bubble level, the locking knob is tightened to complete the hard locking of the angle.

[0077] S120: Based on the detection parameters, continuously and synchronously capture images of the elevator door lock under dual perspectives to obtain the first initial image and the second initial image corresponding to each continuous moment.

[0078] The relationship between several consecutive time points and the first initial image and the second initial image is shown in Table 1 below:

[0079] Table 1:

[0080] <![CDATA[t1]]> <![CDATA[P 1,1 ]]> <![CDATA[P 1,2 ]]> <![CDATA[t2]]> <![CDATA[P 2,1 ]]> <![CDATA[P 2,2 ]]> <![CDATA[t3]]> <![CDATA[P 3,1 ]]> <![CDATA[P 3,2 ]]> ... ... ... <![CDATA[t n ]]> <![CDATA[P n,1 ]]> <![CDATA[P n,2 ]]>

[0081] S130, based on the first initial image and the second initial image, obtain the initial actual length of the door lock hook corresponding to each consecutive moment;

[0082] S130 is mainly used to filter effective pixels, perform dual-view stereo calibration, and quantize the initial dual-view images at each consecutive moment to accurately determine the initial actual length of the door lock hook engaging the fixing component at that moment; specifically, S130 includes:

[0083] S131, Select effective pixels based on the first initial image and the second initial image at the same consecutive time.

[0084] This step utilizes the 3σ principle to remove outliers and connected component analysis to filter valid data. It selects valid pixels from the original image that can be used for length calculation, aiming to eliminate invalid pixels caused by noise, reflections, and background interference as much as possible. Specifically, it includes the following sub-steps:

[0085] S1311, perform image preprocessing on the first initial image and the second initial image at the same consecutive time, respectively, to obtain the main view preprocessed grayscale image and the secondary view preprocessed grayscale image corresponding to the corresponding consecutive time.

[0086] In this embodiment, the image preprocessing steps include median filtering for noise reduction, 8-bit grayscale conversion, adaptive binarization, and morphological closing operation, in order to eliminate salt-and-pepper noise and repair contour breaks. The grayscale value range is 0-255.

[0087] S1312, based on the 3σ principle, performs initial screening of effective pixels for the main view preprocessed grayscale image and the secondary view preprocessed grayscale image respectively.

[0088] Among them, the grayscale image G is preprocessed from the main perspective. t,1 For example, the initial screening formula for effective pixels is as follows:

[0089]

[0090] Where, N total μ is the total number of pixels in the grayscale image. G1 σ represents the global grayscale mean. G1 G represents the standard deviation. t,1 (i) represents the gray value of the i-th pixel;

[0091] Based on the 3σ principle, grayscale values ​​within the range [μ] are selected. G1- 3σ G1 ,μ G1 +3σ G1 Within the specified range, remove abnormal pixels (such as overexposed reflections, completely black shadows, and background interference pixels); secondary view grayscale image G t,2 The same screening logic was used to complete the initial screening.

[0092] S132, Perform connected region analysis and merging on the effective pixels after initial screening to obtain the set of effective pixels from the main viewpoint and the set of effective pixels from the secondary viewpoint at the same continuous time.

[0093] Specifically, this step is achieved through the structuring of effective pixel coordinates and the construction of a grid index. First, the unordered initial screening of effective pixel coordinates is transformed into regular grid coordinates. Then, an efficient grid index table is constructed based on this, facilitating rapid neighborhood retrieval in the subsequent process.

[0094] Taking the set of effective pixels initially screened from the main viewpoint at the same continuous time t as an example (the secondary viewpoint uses the exact same logic), let n effective pixels be obtained after initial screening, and the original image coordinates of each point be (x... i ,y i (i=1,2,...,n, coordinate unit: pixels, origin is the bottom left corner of the image):

[0095] S1321, change the original image coordinates (x i ,y iConvert the coordinates to integer grid coordinates (a) according to the preset grid step size s (in this embodiment, s = 1 pixel, i.e., each pixel corresponds to one grid) according to the preset grid step size s. i ,b i );

[0096] The conversion formula is as follows: , , This is a floor function used to ensure that each pixel corresponds to a unique grid coordinate.

[0097] S1322, according to horizontal grid number a i All initially screened valid pixels are hierarchically grouped to obtain group sets, and a three-dimensional index table is constructed based on the group sets;

[0098] Where the grouping set G = {G0, G1, ..., G} m}( (where W is the maximum horizontal grid number and W is the image width), and each group set G a Pixels within the grid are ordered by vertical grid number b i Sort in ascending order; use a three-dimensional index table Index[a][b]=k, where k is the original index of the pixel corresponding to the grid coordinate in the initial set of valid pixels (if a grid has no valid pixels, then Index[a][b]=-1), to achieve fast pixel location and retrieval.

[0099] S1323, perform neighborhood binding on the effective pixels in the 3D index table to initially divide the contour region and obtain the initial contour region;

[0100] In this embodiment, S1323 is implemented as follows:

[0101] Set the region tag counter cnt=0, the tag array Tag (length n, initial value -1, -1 indicates an unassigned region), the region point set dictionary Area (key is region ID, value is the original index list of pixels in the region), and the list of regions to be merged MergeList (initially empty).

[0102] Traverse the group set G from 0 to m according to the horizontal grid number a, and for each unassigned region of pixel P(a) within each group i ,b i (Original index k):

[0103] If Tag[k] = -1, then cnt += 1, set Tag[k] to cnt, create the key cnt in Area, and add k to Area[cnt].

[0104] Traverse the 8-neighbor grid coordinates (a′, b′) of the pixel (satisfying |a′-a)i |≤1 and |b′-b i |≤1), if a′ is within the valid grid range and Index[a′][b′]=-1 (corresponding to the original index k′):

[0105] If Tag[k′] = -1, set Tag[k′] to cnt and add it to Area[cnt].

[0106] If Tag[k′] = -1 and Tag[k′] = cnt, store the region pair (cnt, Tag[k′]) in MergeList to avoid duplicate records.

[0107] S1324, merge and correct the initial contour region to obtain the effective pixel set of the main view and the effective pixel set of the secondary view at each consecutive time step.

[0108] In this embodiment, S1324 is implemented as follows:

[0109] Create a parent node mapping table Parent (key is region ID, value is the merged main region ID, initial value is itself), and iterate through the region pairs (P) in MergeList. A ,P B Let Parent[min(P)] A ,P B )]=max(P A ,P B This process merges the pixels of the sub-region into the main region and removes the sub-region from the Area.

[0110] For the merged main region, extract its coordinate range [a min ,a max ]×[b min ,b max It iterates through all grids within the range and adds unmarked points adjacent to pixels in the main region to the main region to complete the cross-layer association correction;

[0111] Preset minimum effective area pixel threshold N min =50, and iterate through all main regions of Area, counting the number of pixels N in each region. C If N C <N min If so, the area is marked as an invalid noise area and deleted from the Area;

[0112] The minimum effective area pixel threshold needs to be set according to the door lock outline size to ensure that the minimum outline of the lock hook or fastener is covered.

[0113] Merge the pixels of all remaining areas in the Area to obtain the effective pixel set S from the main viewpoint. t,1 The secondary viewpoint uses the exact same connected component analysis logic to obtain the effective pixel set S. t,2 .

[0114] S133, based on the effective pixel set of the main viewpoint and the effective pixel set of the secondary viewpoint, obtain all key feature points corresponding to each consecutive time moment and the key coordinates corresponding to each key feature point.

[0115] Among them, key feature points include the end point of the locking hook head and the reference point on the inner side of the fixing component, and key coordinates include the coordinates of the key locking hook head and the coordinates of the key inner side of the fixing component; specifically, S133 includes:

[0116] S1331, for the effective pixel set S of the main viewpoint t,1 The Canny edge detection algorithm is used to extract the continuous contour of the door lock component, and the contour tracking algorithm is used to separate the lock hook contour from the fastener contour to obtain the first lock hook contour pixel subset S. hook,1 and the first fixing element contour pixel subset S fix,1 ;

[0117] Among them, the same logic is used to extract the effective pixel set S from the secondary viewpoint. t,2 The corresponding second hook side profile pixel subset S hook,2 and the first fixing element contour pixel subset S fix,2 All contour pixel subsets contain only the valid pixels of the corresponding component.

[0118] Since the calculation principle for the inner coordinates of the critical fastener is the same as that for the head coordinates of the critical locking hook, the following explanation will only focus on the calculation of the head coordinates of the critical locking hook:

[0119] S1332, locate the endpoint of the hook head in the first hook contour pixel subset to obtain the reference hook head coordinates;

[0120] In this embodiment, the effective pixel point farthest from the outline of the fastener is taken as the endpoint of the hook head, and the coordinates of the hook head (x) are referenced. h ,y h The calculation formula for ) is as follows:

[0121]

[0122] in, This represents the minimum Euclidean distance from the hook profile point to the fastener profile.

[0123] The midpoint of the consecutive valid pixels with the smallest x-coordinate in the outline of the fastener is taken as the inner reference point of the fastener, and the inner coordinates (x) of the fastener are referenced. f,y f The calculation formula for ) is as follows:

[0124]

[0125] Among them, (x i ',y i ') represents the coordinates of the k consecutive valid pixels with the smallest horizontal coordinate in the outline of the fixed part. In this embodiment, k is set to 10 to ensure the stability of the reference point.

[0126] It should be noted that, in this embodiment, the main-view camera is primarily responsible for capturing images of the door lock's hook area from the front to calculate its length. However, slight angular deviations in the camera bracket installation are inevitable, and the slight vibrations after the elevator stops can cause slight perspective distortion in the main-view pixel coordinates, resulting in a discrepancy between the actual and measured coordinates. Direct use of this information would affect the accuracy of the length calculation. Therefore, this embodiment utilizes a secondary viewpoint—captured from the side of the lock hook—as a calibration benchmark to correct any potential deviations in the main-view data. This ensures that the captured data more accurately matches the actual door lock structure, significantly eliminating minor installation errors in the bracket and camera, as well as coordinate shifts caused by slight elevator vibrations, thereby improving detection accuracy.

[0127] S1332, based on the preset dual-view coordinate transformation rules, the projected hook head coordinates of the corresponding second hook contour pixel subset are obtained according to the reference hook head coordinates;

[0128] S1332 can also be: based on a preset dual-view coordinate transformation rule, the projected inner coordinates of the corresponding second hook contour pixel subset are obtained according to the inner coordinates of the reference fastener; perspective correction between the two coordinates only requires one pair of data to achieve, and the calculation principles of the projected hook head coordinates and the projected inner coordinates of the fastener are the same, so this embodiment takes the hook head endpoint as an example for explanation:

[0129] After the detection system is deployed and before formal testing, a metrologically calibrated planar calibration board, such as a checkerboard calibration board, can be used to perform homography matrix calibration on the main and secondary viewpoint cameras to obtain the 3×3 homography matrix H from the main viewpoint to the secondary viewpoint. A homogeneous coordinate transformation matrix is ​​generally used, and the matrix form is as follows:

[0130]

[0131] Where h 33 =1, used for normalization constraints, and the remaining elements are obtained by solving the homogeneous coordinates of at least 4 pairs of corresponding points on the calibration board. These are pre-calibrated fixed parameters and stored in the standard reference library.

[0132] Since the primary and secondary viewpoint cameras are fixed and orthogonally mounted, based on the preset dual-viewpoint coordinate transformation rules, namely the 3×3 homography matrix H, it acts as a "coordinate translator" between the two views. This allows for the precise conversion of any pixel coordinate in the primary viewpoint into the corresponding projected coordinates in the secondary viewpoint, and vice versa. In this embodiment, the homogeneous coordinates of the reference hook head are set as follows: The homogeneous coordinates of the verification point corresponding to the secondary viewpoint are: The two are related through the homography matrix H, and the conversion formula is:

[0133]

[0134] in, The non-zero scaling factor, after expansion, is:

[0135]

[0136] Right now:

[0137]

[0138] S1333, locate the actual hook head endpoint corresponding to the projected hook head coordinates in the second hook contour pixel subset, and obtain the actual hook head coordinates.

[0139] The actual hook head coordinates are the positions of the actual effective pixels that are closest to the projected hook head coordinates, found from the second hook contour pixel subset.

[0140] S1334, calculate the Euclidean distance between the projected lock hook head coordinates and the actual lock hook head coordinates, and determine whether to trigger the coordinate correction mechanism based on the Euclidean distance to obtain the key lock hook head coordinates and the inner coordinates of the key fastener.

[0141] Among them, the coordinates of the projected lock hook head (x h2 ,y h2 ) and the actual hook head end point (x ′ h2 ,y ′ h2 The Euclidean distance formula is as follows:

[0142] ;

[0143] If d h ≤The preset calculation allowable error, this embodiment takes 2 pixels as an example, to determine the validity of the reference coordinates, the reference lock head coordinates are directly used as the key lock head coordinates, and the corresponding reference key fastener inner coordinates are directly used as the key fastener inner coordinates.

[0144] If d h>The preset allowable calculation error determines that perspective distortion exists in the main view reference coordinates, triggering a coordinate correction mechanism. At this time, the pre-calibrated inverse homography matrix H is used. -1 The actual hook head coordinates of the second hook outline pixel subset are converted inversely to the corresponding precise coordinates in the main view, which are then used as the corrected key hook head coordinates (x). ′ h ,y ′ h ), inner coordinates of key fasteners (x ′ f ,y ′ f Similarly;

[0145] Among them, the inverse homography matrix H -1 The pre-calibrated 3×3 matrix is ​​denoted as... Where i,j=1,2,3 and h ′ 33 =1 is the normalization constraint, and the inverse transformation formula is as follows:

[0146] .

[0147] S134, calculate the initial actual length of the door lock hook at each consecutive moment based on the inner coordinates of the key fastener and the head coordinates of the key lock hook;

[0148] Since the main-view camera has been calibrated to be 90° perpendicular to the door lock mating plane, and the hook insertion direction is the horizontal direction of the image, the effective pixel distance corresponding to the insertion length is the absolute value of the horizontal pixel difference between the corrected inner reference point of the fixing component and the end point of the hook head. The corresponding calculation formula is:

[0149] , where P t x represents the effective pixel distance corresponding to the hook engagement length of the door lock at continuous time t, in pixels; ′ f x is the abscissa of the inner coordinate of the key fastener. ′ h The x-coordinate of the key hook head coordinates is used, and the absolute value is taken to ensure that the pixel distance is non-negative, which matches the actual hook insertion direction.

[0150] The effective pixel distance is converted into the actual length of the door lock hook engaging the fixing component based on the preset physical length-pixel distance linear mapping model.

[0151] The formula for the physical length-pixel distance linear mapping model is as follows:

[0152]

[0153] Among them, Lt P represents the actual length of the door lock hook engaging the fixing component at continuous time t, where k is the scaling factor of the physical length-pixel distance mapping model, representing the actual physical length corresponding to a single pixel. In this embodiment, this data is obtained from a calibrated 7mm standard door lock template; t is the effective pixel distance calculated at continuous time t, and b is a fixed correction value of the mapping model used to compensate for minor systematic errors in camera installation, which is also calibrated by a standard door lock template;

[0154] It should be noted that the calibration method in this embodiment uses three sets of templates with different standard hook lengths of 3mm, 7mm, and 10mm. The corresponding effective pixel distances P1, P2, and P3 are photographed and calculated respectively. k and b are solved by least squares fitting to ensure that the accuracy of the mapping model is ≤0.1mm.

[0155] S140, based on the initial actual length at all consecutive moments, the acceptance benchmark length is obtained, and the acceptance benchmark length is subjected to acceptance compliance testing to obtain the acceptance test results;

[0156] The calculation principle of the acceptance benchmark length in S140 is the same as that in S134. It mainly uses the 3σ principle to remove outliers from the initial actual length sequence corresponding to each consecutive time step output by S134, and selects the valid data that can truly reflect the hook length in the door lock's on state. Finally, the acceptance benchmark length of the door lock is calibrated, which is consistent with the pixel filtering logic mentioned above. The specific operation is as follows:

[0157] S1401, based on the 3σ principle, effectively filters all initial actual lengths to obtain an effective length sequence;

[0158] The screening principle in this step is the same as that in S131 above, and will not be repeated here; the effective length sequence refers to the initial actual length that excludes the abnormality caused by mechanical vibration and sudden change in light at the moment the elevator door opens.

[0159] S1402, perform weighted average calibration on the effective length sequence to obtain the acceptance benchmark length of the door lock on this floor;

[0160] The weight of each effective length sequence is inversely proportional to the time difference between the corresponding moment and the moment the landing door opens. That is, the closer the length value is to the moment of opening, the more accurately it reflects the engagement state when the elevator is in operation. The weighted calculation formula is as follows:

[0161]

[0162] Among them, L base L is the acceptance reference length. valid For a valid length sequence, m is the number of valid length values, and m ≤ n, w jThe weight parameter corresponding to the j-th effective length sequence is calculated using the following formula: And the sum of the weights is 1, Δt j Let T be the time difference between the time corresponding to the j-th valid length value and the time when the door opens. max This represents the maximum time difference corresponding to all valid length values.

[0163] Furthermore, if the discrete coefficients of the effective length sequence are CV=σ L / μ L If the value is ≤2%, it indicates that the length value fluctuates very little, and the arithmetic mean of the effective length sequence can be directly used as the acceptance benchmark length.

[0164] S1403, Perform acceptance compliance testing based on the acceptance benchmark length and generate the acceptance test results for the door locks on this floor;

[0165] In this embodiment, the mandatory safety requirement of a minimum engagement depth of 7mm for elevator landing door locks in GB7588-2003 "Safety Code for Elevator Manufacturing and Installation" is used as the core judgment criterion. This is combined with a preset detection accuracy threshold to complete the compliance judgment. Specifically:

[0166] The compliance assessment rules set a preset threshold L for the minimum safe hook length of the door lock. th =7mm, and at the same time set the critical warning range [7mm, 7.5mm]. According to the numerical range of the acceptance benchmark length, the acceptance test results are divided into three levels: qualified, critical warning and unqualified.

[0167] When L base When the thickness is greater than 7.5mm, the door lock installation on that floor is deemed to be qualified and the locking strength meets the safety specifications.

[0168] When 7mm≤L base When the error is ≤7.5mm, the door lock on that floor is determined to be in a critical safety state, triggering an early warning. The door lock installation accuracy needs to be manually checked, and the corresponding target elevator should be listed as a key monitoring object. The detection frequency can be appropriately increased in the future.

[0169] When L base If the thickness is less than 7mm, the door lock installation on that floor is deemed unqualified, the locking strength does not meet safety standards, and an abnormal alarm is triggered.

[0170] S200 establishes a standard benchmark library and initial full life cycle benchmark data based on the acceptance benchmark length corresponding to each door layer.

[0171] The standard benchmark library and initial full lifecycle benchmark data are both bound to the unique ID of the target elevator, stored in the local storage unit of the data processing module, and simultaneously encrypted and uploaded to the elevator maintenance cloud management platform. This ensures they are tamper-proof and provides a unified, traceable, and quantitative reference for subsequent full lifecycle maintenance inspections, trend analysis, and hazard warnings.

[0172] This embodiment adopts a hierarchical management logic of "general benchmark + dedicated benchmark", which takes into account both the universality of elevators of the same model and the dedicated adaptability of individual elevators. The general standard benchmark sub-library is mainly used to store the general benchmark data that has been calibrated by metrology. It is applicable to the testing of door locks of all elevators of the same model, including: the standard pixel size corresponding to the 7mm safety hook length, the physical length-pixel distance linear mapping model parameters calibrated by the 3mm / 7mm / 10mm standard door lock template, such as the proportional coefficient k, correction value b, etc., the primary and secondary viewpoint homography matrix H and inverse homography matrix H-1 pre-calibrated by S133, the standard contour effective pixel feature set of the lock hook and fastener, and the safety threshold parameters corresponding to the acceptance execution standard (GB7588-2003 "Safety Code for Elevator Manufacturing and Installation").

[0173] The elevator-specific benchmark sub-library is used to store exclusive data from historical testing and calibration, achieving precise adaptation of "one benchmark per floor". This includes: the acceptance benchmark length of all floor door locks of the elevator, the internal and external parameters of the main and secondary view cameras after on-site calibration, the initial effective images of the door locks on each floor from both perspectives, the contour feature parameters of the corresponding floors, coordinate correction coefficients, etc., in order to eliminate benchmark errors caused by installation deviations of door locks on different floors as much as possible.

[0174] For maintenance throughout the entire lifecycle, this embodiment mainly establishes a structured initial full lifecycle benchmark dataset with "target elevator ID-floor number-acceptance time" as the unique primary key to accurately anchor the initial safety status of the new elevator. The core of the dataset includes, but is not limited to, the unique ID of the target elevator, the total number of floors, the door lock model, the installation and acceptance time, the acceptance benchmark length of different floors, the initial actual length sequence of a single floor, the preliminary compliance judgment result, the correction parameters of the physical-pixel mapping model calibrated on site, the primary and secondary view calibration parameters, and a series of other detection process data.

[0175] After receiving the daily inspection instruction, the S300 performs a full lifecycle-linked daily inspection and trend analysis to obtain daily inspection results and a full lifecycle safety trend report.

[0176] Specifically, the S300 includes:

[0177] S310 receives daily inspection commands and loads the target elevator's dedicated benchmark library and full lifecycle historical dataset.

[0178] This step mainly involves data retrieval, and the corresponding data can be seen in S200, which is existing technology and will not be described in detail here.

[0179] The S320 performs standardized dual-view synchronous image acquisition to obtain daily images of the door lock at each continuous moment;

[0180] This step strictly follows the acquisition rules of the S100 acceptance test results. The detachable telescopic bracket is extended to a preset position that is completely consistent with the acceptance test. The main and secondary view cameras are started with the same resolution, frame rate, and synchronous trigger mode. Continuous shooting is performed in the core window of 0.1-0.3 seconds when the elevator door just opens. The main view daily test images and secondary view daily test images of each floor are obtained at the corresponding continuous time, ensuring that the acquisition conditions are completely consistent with the acceptance benchmark images.

[0181] S330, calculates the actual hook insertion length of the door lock hook in the current cycle based on all daily images of the door locks;

[0182] In particular, step S330 fully reuses the standardized algorithm logic of S131-S134 to process the daily inspection images to ensure that the calculation caliber and the acceptance benchmark length are completely consistent: preprocessing is performed on the dual-view daily inspection images in sequence, the effective pixels are initially screened based on the 3σ principle, and connected region analysis is performed to obtain the daily effective pixel point set of the main and secondary views; the dual-view coordinate perspective correction is completed based on the pre-calibrated homography matrix, key feature points are extracted and the effective pixel distance is calculated, and the actual hooking length of the door lock hook on each floor in the current inspection cycle is obtained through the physical length-pixel distance mapping model.

[0183] S340, based on the actual hooking length of the door lock hook in the current cycle, obtain the daily inspection results;

[0184] The compliance assessment logic of S340 is the same as that of S1403 mentioned above, mainly by directly comparing with preset compliance parameters, which will not be elaborated here.

[0185] S350 performs full lifecycle trend analysis based on the actual hook-in length and acceptance benchmark length corresponding to all historical cycles.

[0186] Specifically, the S350 includes:

[0187] S351, construct a full life cycle length sequence based on the actual hooking length and acceptance benchmark length corresponding to all historical cycles;

[0188] Specifically, using the acceptance benchmark length of the door lock on the target floor as the initial value of the sequence, the actual insertion lengths obtained from previous daily / periodic inspections are superimposed according to the order of inspection time to construct the original lifecycle length sequence L of the door lock on that floor. life =[L base ,L1,L2,...,L k ,L current ], where k is the number of historical detection cycles, L current This is the actual hook length obtained in this test.

[0189] S352, calculate the target feature parameters of the corresponding floor door locks based on the full life cycle length sequence;

[0190] The target characteristic parameters include cumulative degradation rate, periodic degradation rate, and performance stability coefficient. The cumulative degradation rate reflects the overall performance degradation of the door lock from installation and acceptance to the present. The periodic degradation rate reflects the rate of degradation within a unit testing cycle, helping to predict future security risks. The performance stability coefficient reflects the fluctuation of the door lock's locking performance throughout its entire lifecycle, identifying hidden risks such as abnormal wear and loose installation. Specifically, the calculation formulas for the target characteristic parameters are as follows:

[0191]

[0192]

[0193]

[0194] Among them, L base The standard length for accepting door locks on this floor; L current η is the actual hook-in length obtained in this test. total This indicates the cumulative degradation rate; a higher value means a more severe overall decline in the locking performance of the door lock. last The actual hooking length in the previous detection cycle; ΔT is the interval between two detections, which can be in days or months, or in the cumulative number of elevator runs; v deter This represents the periodic degradation rate; a higher value indicates a faster rate of wear and misalignment of the door lock in the near future, leading to a faster increase in security risks. life The standard deviation of the effective trend series; μ life The arithmetic mean of the effective trend series; CV life This is the performance stability coefficient. The higher the value, the greater the fluctuation in the lock's engagement length, and the higher the hidden risk of mechanical loosening and abnormal wear.

[0195] S353, Generate a full lifecycle security trend report based on target feature parameters;

[0196] This step, based on the three target characteristic parameters calculated by S352—cumulative degradation rate, periodic degradation rate, and performance stability coefficient—combined with the compliance judgment result of the actual hook-in length in the current period, establishes a multi-level risk warning system to complete the security level assessment, and finally generates a standardized full life cycle security trend report; the risk warning classification rules are shown in Table 2 below:

[0197] Table 2

[0198] Normal level 1. Acceptance test results: Current actual insertion length > 7.5mm; 2. Acceptance test results: Cumulative degradation rate < 10%, periodic degradation rate < 0.1mm / month, performance stability coefficient < 3%. Without prior warning, subsequent testing will be carried out according to the established routine cycle. Attention level The current actual hook-in length is >7.5mm, and the cumulative degradation rate acceptance test result is 10%~15%, or the periodic degradation rate acceptance test result is 0.1~0.3mm. The on-site terminal marks the floors of interest, prompting a focus on review during the next testing cycle. Warning level 1. Acceptance test result: Actual hook-in length 7mm~7.5mm; 2. Acceptance test result: Cumulative degradation rate ≥15%; 3. Acceptance test result: Periodic degradation rate ≥0.3mm / month; 4. Acceptance test result: Performance stability coefficient ≥5% A yellow alert is triggered, the on-site terminal highlights the alert, and a maintenance reminder is pushed. It is recommended that the acceptance test results be manually reviewed and potential hazards identified within 7 days. Alarm level 1. Acceptance test result: Current actual indentation length < 7mm; 2. Acceptance test result: Periodic degradation rate ≥ 0.5mm / Acceptance test result month A red audible and visual alarm is triggered, the abnormal floor detection result is identified, an emergency maintenance work order is sent, requiring immediate shutdown for rectification and completion of a re-inspection.

[0199] It should be noted that after completing the risk level assessment, this embodiment will automatically generate an "Elevator Door Lock Full Life Cycle Safety Trend Report" based on the results of this daily inspection, full-cycle deterioration characteristic parameters, and risk level classification results. The report includes basic information of the target elevator, details of the hook length of each floor in this inspection and compliance results, full-cycle hook length trend curve, deterioration characteristic parameters and risk levels of each floor, hidden danger warning information, and targeted maintenance and rectification suggestions. The report supports local printing, export of PDF acceptance inspection results, and synchronization with the cloud maintenance platform. At the same time, all data and analysis results of this inspection will be updated and archived to the target elevator's full life cycle historical dataset to complete the full life cycle closed loop of inspection and acceptance inspection results - inspection and acceptance inspection result analysis - inspection and acceptance inspection result warning - inspection and acceptance inspection result archiving.

[0200] Based on the same inventive concept described above, this application also discloses a smart terminal, which includes a memory and a processor. The memory stores at least one instruction, at least one program, code set, or instruction set. The at least one instruction, at least one program, code set, or instruction set is loaded and executed by the processor to implement an elevator door lock detection method based on a high-definition camera as provided in the above method embodiments.

[0201] Based on the same inventive concept described above, this application also discloses a computer-readable storage medium storing at least one instruction, at least one program, code set, or instruction set, wherein the at least one instruction, at least one program, code set, or instruction set can be loaded and executed by a processor to implement the elevator door lock detection method based on a high-definition camera provided in the above method embodiments.

[0202] It should be understood that "multiple" as used in this article refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. The character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0203] Those skilled in the art will understand that all or part of the steps of the above embodiments can be implemented by hardware or by a program instructing related hardware. The program can be stored in a computer-readable storage medium, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, and other media capable of storing program code.

[0204] The above description is merely an optional embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for detecting elevator door locks based on a high-definition camera, characterized in that, The elevator door lock detection method is based on an elevator door lock detection system, the system including a camera for capturing images of the door lock, and the method including: Upon receiving the installation and acceptance instruction, images of the elevator door locks are acquired according to the preset detection parameters to obtain several initial images of the door locks. Acceptance compliance tests are then performed based on these initial images to obtain the acceptance test results. A standard benchmark library and initial full life cycle benchmark data are established based on the acceptance benchmark length corresponding to each door layer. Upon receiving a routine inspection instruction, the system performs a full lifecycle-linked routine inspection and trend analysis to obtain routine inspection results and a full lifecycle safety trend report.

2. The elevator door lock detection method according to claim 1, characterized in that, The system specifically includes a main-view camera and a secondary-view camera installed based on a preset dual-view coordinate transformation rule. The initial images of the door lock include a first initial image corresponding to the main view and a second initial image corresponding to the secondary view. The second initial image is used to calibrate and correct the first initial image by using the projection relationship between a pixel in the second initial image and a corresponding pixel in the first initial image. The step of acquiring images of the elevator door lock according to preset detection parameters to obtain several initial door lock images specifically includes: The elevator door lock is synchronously photographed from two perspectives according to preset detection parameters to obtain the first initial image and the second initial image. The detection parameters are used to limit and unify the shooting of the main perspective camera and the secondary perspective camera at different consecutive moments.

3. The elevator door lock detection method according to claim 2, characterized in that, The synchronous shooting is performed multiple times within a continuous period of time after the elevator door opens, and each continuous period of time corresponds to one first initial image and one second initial image. The acceptance and compliance inspection based on the initial image of the door lock yields the following results: The initial actual length of the door lock hook corresponding to each of the consecutive time moments is obtained based on the first initial image and the second initial image. The acceptance benchmark length is obtained based on the initial actual length at all consecutive moments, and the acceptance benchmark length is subjected to acceptance compliance testing to obtain the acceptance test results.

4. The elevator door lock detection method according to claim 3, characterized in that, The step of obtaining the initial actual length of the door lock hook corresponding to each of the consecutive time moments based on the first initial image and the second initial image includes: Valid pixels are filtered based on the first initial image and the second initial image at the same consecutive time. After initial screening, the effective pixels are analyzed and merged to obtain the set of effective pixels from the main viewpoint and the set of effective pixels from the sub-viewpoint at the same continuous time. The analysis and merging of the effective pixels are achieved by structuring the coordinates of the effective pixels and constructing a grid index. Based on the effective pixel set of the main view and the effective pixel set of the secondary view, all key feature points corresponding to each consecutive time moment and key coordinates corresponding to each key feature point are obtained. The key feature points are used to calculate the hook insertion length of the door lock. The initial actual length of the door lock hook at each consecutive moment is calculated based on the key coordinates.

5. The elevator door lock detection method according to claim 4, characterized in that, The key feature points include the end point of the hook head and the reference point on the inner side of the fastener; the key coordinates include the key hook head coordinates; obtaining all key feature points corresponding to each consecutive moment and the key coordinates corresponding to each key feature point based on the effective pixel set of the main view and the effective pixel set of the secondary view includes: The continuous contour of the door lock component within the effective pixel set of the main viewpoint is extracted by the edge detection algorithm to obtain the first lock hook contour pixel subset; The continuous contour of the door lock component within the effective pixel set of the secondary viewpoint is extracted by the edge detection algorithm to obtain the second lock hook contour pixel subset; Locate the endpoint of the hook head in the first hook contour pixel subset to obtain the reference hook head coordinates; Based on the dual-view coordinate transformation rule, the projected hook head coordinates corresponding to the second hook contour pixel subset are obtained according to the reference hook head coordinates; Locate the actual hook head endpoint corresponding to the projected hook head coordinates in the second hook contour pixel subset to obtain the actual hook head coordinates; Calculate the Euclidean distance between the projected hook head coordinates and the actual hook head coordinates, and determine whether to trigger the coordinate correction mechanism based on the Euclidean distance to obtain the key hook head coordinates.

6. The elevator door lock detection method according to claim 5, characterized in that, In the first subset of lock hook contour pixels, the effective pixel point farthest from the fastener contour is taken as the head endpoint of the lock hook, and the midpoint of the continuous effective pixel point with the smallest abscissa in the fastener contour is taken as the inner reference point of the fastener.

7. The elevator door lock detection method according to claim 5, characterized in that, The detection parameters include calculating the allowable error; the determination of whether to trigger the coordinate correction mechanism based on the Euclidean distance, and obtaining the coordinates of the key lock head, includes: If the Euclidean distance between the projected hook head coordinates and the actual hook head coordinates is less than or equal to the allowable error, then the reference hook head coordinates are directly used as the key hook head coordinates. Otherwise, a coordinate correction mechanism is triggered, and the actual hook head coordinates corresponding to the second hook contour pixel subset are converted into the corresponding precise coordinates of the first hook contour pixel subset through the pre-calibrated inverse homography matrix H-1, which are then used as the corrected key hook head coordinates.

8. The elevator door lock detection method according to claim 1, characterized in that, Upon receiving the routine inspection instruction, the system performs a full lifecycle-linked routine inspection and trend analysis to obtain routine inspection results and a full lifecycle safety trend report, including: Receive daily inspection instructions and load the target elevator's full lifecycle historical dataset, which includes the actual hook length and the acceptance benchmark length for all historical cycles. Perform standardized image acquisition to obtain several daily images of door locks, and calculate the actual hooking length of the door lock hook in the current period based on all the daily images of door locks to obtain the daily detection results; Based on the historical dataset of the entire life cycle, perform a full life cycle trend analysis to obtain target feature parameters that reflect the degree of decay in the locking performance of the door lock, and generate a full life cycle security trend report based on the target feature parameters.

9. An elevator door lock detection system based on a high-definition camera, characterized in that, include: The data acquisition module is used to acquire door lock images at several consecutive moments after the elevator door opens; The data processing module is used to receive the door lock image and perform compliance detection on the door lock image using the elevator door lock detection method of any one of claims 1 to 8 to obtain the detection result.

10. A computer-readable storage medium, characterized in that, The readable storage medium stores at least one instruction, at least one program, code set, or instruction set, wherein the at least one instruction, at least one program, code set, or instruction set is loaded and executed by a processor to implement an elevator door lock detection method based on a high-definition camera as described in any one of claims 1 to 8.

Images