A substation pressboard misplacement identification method

Abstract

This invention relates to the field of image recognition technology and discloses a method for identifying substation pressure plate misalignment. This method addresses the problem in traditional methods of identifying pressure plate objects and logical misalignments where the apparent compliance of the pressure plate's activation / deactivation status is inconsistent with the current operating conditions. The method first establishes a pressure plate reference association model on a cabinet basis, binding the pressure plate's identity, physical slot, nameplate text, operating condition status, and combined constraint rules of the pressure plate's activation / deactivation status to the cabinet panel coordinate system. It then collects cabinet panel inspection images, reconstructs the cabinet panel coordinate system based on the cabinet's structural anchor points, and segments the pressure plate working area. Finally, it verifies the pressure plate object's identity, activation / deactivation status, combined constraint rules of activation / deactivation status, and operating condition association, generating pressure plate misalignment types and alarm records. This improves the accuracy of identifying both concealed and logical misalignments of the pressure plate.

Description

Technical Field

[0001] This invention relates to the field of image recognition technology, specifically a method for identifying misaligned pressure plates in substations. Background Technology

[0002] The activation / deactivation status of substation relay protection pressure plates corresponds to the operating mode of the protection device, directly affecting whether the protection device can operate correctly according to the predetermined logic. If the pressure plate is missed, mistakenly activated, or incorrectly installed, it can easily cause the protection to fail to operate or malfunction, or lead to the distortion of the verification results of secondary safety measures. For example, the published invention patent application CN110277833B discloses a method and terminal equipment for identifying the status of substation pressure plates. It identifies the current status of the pressure plates by collecting real-time monitoring information of the protection pressure plates and issues an alarm when the current status is inconsistent with the status stored in the database. Another example is the published invention patent application CN110827333B, which discloses a method, system, and medium for identifying the splicing of relay protection pressure plates. It obtains the pressure plate data in the cabinet by identifying, dividing, matching, and splicing the pressure plate objects and text data in the pressure plate image. The above-mentioned solutions can improve the efficiency of pressure plate inspection and the ability to identify the engagement / disengagement status to a certain extent. However, they mainly focus on pressure plate status recognition, text matching, or image stitching, without establishing verifiable benchmark relationships based on the unique identity of the pressure plate, its actual physical slot, nameplate text, expected status, logical relationship with other equipment in the same cabinet, and the current operating condition of the primary equipment. Therefore, in cases where pressure plates are mistakenly installed in adjacent slots after maintenance, where pressure plates are shifted entirely or partially after cabinet modifications, or where adjacent pressure plates are misattributed due to target frame offset, the existing methods may only make compliance judgments based on the pressure plate engagement / disengagement status, making it difficult to identify hidden misalignments where the pressure plate status itself is correct but the installation position is incorrect. Furthermore, issues such as positioning deviation of the inspection robot and gimbal angle accuracy also pose challenges. In situations such as descent, obstruction of reflective surfaces by cabinets, or damage to positioning markers, a single image reference or positioning reference is prone to failure, leading to a batch of false alarms or missed alarms. Furthermore, existing methods typically do not combine primary equipment operating modes, interval maintenance status, and secondary safety measures for condition coupling verification, making it difficult to identify logical misalignments where the pressure plate object and its activation / deactivation status appear compliant but are inconsistent with the current operating conditions. Therefore, there is an urgent need to propose a substation pressure plate misalignment identification method. This method establishes a pressure plate reference association model and performs dual verification of the pressure plate object's identity, physical slot location, measured status, and operating conditions to improve the detection capability of both concealed and logical misalignments, thereby enhancing the accuracy of pressure plate inspection and verification and the reliability of secondary safety measures implementation. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention provides a method for identifying substation pressure plate misalignment, which solves the problem in traditional methods of difficulty in identifying pressure plate objects and logical misalignment where the seemingly compliant activation / deactivation status is inconsistent with the current operating conditions.

[0004] To achieve the above objectives, the present invention provides the following technical solution: A method for identifying misalignment of substation pressure plates includes: S1. Establish a pressure plate reference association model based on the cabinet, and bind the pressure plate identity code, physical slot, nameplate text, working condition status, and pressure plate deployment / retraction status combination constraint rules with the cabinet panel coordinate system. The pressure plate deployment / retraction status combination constraint rules include mutual exclusion constraint rules, collaborative constraint rules, and deployment quantity constraint rules. S2. Obtain the inspection image of the cabinet panel, the acquisition station, the acquisition time, the operating conditions of the primary equipment and the interval maintenance status, and call the pressure plate reference association model corresponding to the acquisition station; S3. Extract the structural anchor points of the cabinet from the inspection images of the cabinet panel, reconstruct the coordinate system of the cabinet panel based on the structural anchor points, and divide each pressure plate working area according to the reference slot. S4. Perform pressure plate object identity verification and pressure plate loading / unloading status verification in each pressure plate working area to form a verification relationship between pressure plate identity, actual slot position, actual measured status and expected status. S5. Based on the constraint rules of the combination of the on / off states of the pressure plate on the same screen and the correlation between the working conditions of the primary and secondary equipment, the verification relationship is reviewed, and the pressure plate misalignment type and alarm record are generated.

[0005] Furthermore, S1 includes: When establishing the pressure plate reference association model, the panel coordinate system is established based on the fixed structural points of the cabinet; Based on the design drawings, secondary circuit configuration table, protection setting sheet and on-site verification record, determine the correspondence between pressure plate slot, identification code, nameplate area, geometric center, pixel reference point, handle envelope, state angle range, working condition status and pressure plate on / off status combination constraint rules, and encapsulate it into a cabinet-level reference package with version number.

[0006] Furthermore, S2 includes: When calling the pressure plate reference association model, the reference package index is queried according to the cabinet code, the workstation is confirmed, the shooting posture is adjusted and multi-frame images are acquired, and the working condition snapshot is obtained simultaneously and an acquisition message with time correspondence and spatial correspondence is formed. Verify the base package version and perform differential updates when versions are inconsistent; The acquired images are screened for exposure, sharpness, and occlusion quality. If the quality screening fails, re-acquisition or re-verification is performed, and a record to be verified is generated.

[0007] Furthermore, the baseline package version is validated, and differential updates are performed when versions are inconsistent, including: Before calling the local benchmark package, the consistency of the version number, effective version number, inspection task number and workstation code of the local cached benchmark package is verified. When the version number is inconsistent with the effective version number, the changed pressure plate record, working condition mapping entry and the combination constraint rules of the pressure plate activation and deactivation status on the same screen are obtained according to the field-level revision record and written to the local cache; When the workstation code and the inspection task are inconsistent, the base package index of the corresponding cabinet is rematched.

[0008] Furthermore, S3 includes: When reconstructing coordinates and dividing the work area, grayscale, edge enhancement and closing operations are performed on qualified inspection images. The frame, nameplate, pressure plate matrix or replacement structure anchor points of the cabinet are extracted and cross-validated according to the anchor point position relationship in the reference package. The effective anchor point pixel coordinates are paired with the panel millimeter coordinates to establish a geometric transformation relationship. After outlier anchor point removal and reprojection residual judgment, the image is reprojected. The working area is divided according to the geometric center of the pressure plate, the handle envelope, and the nameplate area. The integrity of the divided working area is marked to generate a sub-record to be verified.

[0009] Furthermore, the effective anchor point pixel coordinates are paired with the panel millimeter coordinates to establish a geometric transformation relationship. After outlier anchor point removal and reprojection residual determination, image reprojection is performed, including: When establishing geometric transformation relationships, the image pixel coordinates of the effective anchor points are registered with the panel millimeter coordinates in the reference package, and the geometric transformation matrix is ​​obtained by using planar homography transformation. Abnormal anchor points are identified based on the median of the reprojection residuals of each anchor point. After removing abnormal anchor points, the geometric transformation matrix is ​​recalculated. The reliability of the geometric transformation is marked based on the average reprojection residuals, and the inspection images of the cabinet panels are reprojected.

[0010] Furthermore, S4 includes: When generating verification relationships, after the pressure plate working area is complete and the geometric quality meets the verification conditions, an object identity fusion score is formed based on the geometric center and slot code, nameplate text similarity, and handle and base shape characteristics. After the object identity is verified, the correspondence between the measured states is determined based on the handle state angle, multi-frame consistency, working condition status table and transition time window. When the object identity is incorrect, the state angle is in the boundary range, or the state is undeterminable, the corresponding verification identifier will be written into the verification record.

[0011] Furthermore, when the object identity is inconsistent, the state angle is in the boundary range, or the state is undeterminable, the corresponding verification identifier will be written into the verification record, including: When writing the review identifier, the object is marked as abnormal based on the object identity fusion score, geometric deviation, text similarity and shape anomaly type, and the normal status response process of the corresponding pressure plate is no longer executed; Based on the state angle boundary marker, inter-frame consistency, image quality status, and backup image verification results, the status is marked as potentially abnormal or the status is undeterminable, and the verification marker is associated with the corresponding pressure plate identification code, slot code, and verification time.

[0012] Furthermore, S5 includes: When verifying the relationship, a combined constraint violation index is generated according to the mutual exclusion constraint rule, the collaborative constraint rule, and the input quantity constraint rule; The mutual exclusion constraint rule is used to verify that only one pressure plate in the associated pressure plate is in the input state. The coordination constraint rule is used to verify that the input and withdrawal states of the associated pressure plates are consistent. The input quantity constraint rule is used to verify that the number of associated pressure plates in the input state falls within a preset integer range.

[0013] Furthermore, the measured state vector is compared bit by bit with the expected state vector under the working condition, and inconsistencies are filtered out by combining work order permission and transition time window, including: When performing working condition correlation verification, the measured status of each pressure plate is arranged according to the pressure plate identification code, and the status of each pressure plate under the current primary equipment operating condition is read to generate a working condition inconsistency list. For each pressure plate in the list of inconsistent operating conditions, verify the interval maintenance status, work order number, permit time, work order permit content and permit execution status, and mark the permit transition items or operating condition logic misalignment items according to the allowable transition time window.

[0014] Compared with the prior art, the present invention provides a method for identifying misalignment of substation pressure plates, which has the following advantages: 1. This invention establishes a reference association model for pressure plates on a cabinet-by-cabinet basis, unifying the constraint rules for pressure plate identity, physical slot, nameplate text, expected working status, and the deployment / retraction status of pressure plates on the same cabinet into the cabinet panel coordinate system. Combined with structural anchor point reconstruction from inspection images, pressure plate working area segmentation, object identity verification, deployment / retraction status verification, deployment / retraction status combination constraint verification, and primary and secondary equipment working condition association verification, the correspondence between pressure plate objects, actual slots, measured status, and current working conditions is continuously verified, avoiding judgments based solely on deployment / retraction status in a single frame image. This method can identify anomalies such as misinstallation of adjacent similar pressure plates, nameplate misalignment, target frame offset, reference offset, and inconsistencies between pressure plate status and operating conditions. It reduces false alarms and missed alarms caused by robot positioning deviations, reflective obstruction, or asynchronous working condition information, improving the accuracy of detecting hidden and logical misalignments of pressure plates, the stability of inspection verification, and the reliability of secondary safety measures.

[0015] 2. This invention sets up a cabinet-level benchmark package, version number, field-level revision record, candidate version parallel verification, and event log in the pressure plate benchmark association model. During inspection, it records collected messages, image quality, geometric reliability, object identity fusion score, state consistency ratio, combined constraint violation index, and operating condition permission results. This ensures that each pressure plate identification and alarm judgment has a corresponding data source, model version, image evidence, and rule basis. This method enables incremental maintenance and verification of benchmark data after cabinet modification, pressure plate replacement, nameplate adjustment, or protection configuration change. It avoids batch misjudgments due to expired benchmark packages or local calibration offsets, and facilitates maintenance personnel in tracing alarm causes, verification processes, and handling results. This improves the standardization of pressure plate inspection data management, the traceability of alarm handling, and the reliability of benchmark model maintenance. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the process for identifying misaligned pressure plates in a substation according to the present invention. Figure 2 This is a schematic diagram of the pressure plate reference association model structure of the present invention; Figure 3 This is a scenario diagram of the inspection data collection and reference package invocation of the present invention; Figure 4 This is a schematic diagram of the anchor point extraction and pressure plate working area division of the cabinet structure in this invention; Figure 5 This is a dual verification diagram of the identity and deployment / retraction status of the pressure plate object in this invention; Figure 6 This is the alarm diagram for verifying the combination constraints of the activation and deactivation states and the correlation between the working conditions of this invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Example 1: Figure 1 - Figure 6 A method for identifying misalignment of substation pressure plates is presented, including: S1. Establish a pressure plate reference association model based on the cabinet, and bind the pressure plate identity code, physical slot, nameplate text, working condition status, and pressure plate deployment / retraction status combination constraint rules with the cabinet panel coordinate system. The pressure plate deployment / retraction status combination constraint rules include mutual exclusion constraint rules, collaborative constraint rules, and deployment quantity constraint rules. S2. Obtain the inspection image of the cabinet panel, the acquisition station, the acquisition time, the operating conditions of the primary equipment and the interval maintenance status, and call the pressure plate reference association model corresponding to the acquisition station; S3. Extract the structural anchor points of the cabinet from the inspection images of the cabinet panel, reconstruct the coordinate system of the cabinet panel based on the structural anchor points, and divide each pressure plate working area according to the reference slot. S4. Perform pressure plate object identity verification and pressure plate loading / unloading status verification in each pressure plate working area to form a verification relationship between pressure plate identity, actual slot position, actual measured status and expected status. S5. Based on the constraint rules of the combination of the on / off states of the pressure plate on the same screen and the correlation between the working conditions of the primary and secondary equipment, the verification relationship is reviewed, and the pressure plate misalignment type and alarm record are generated.

[0019] This method takes the secondary cabinets in the relay protection room of a smart substation as the application object. The input information includes at least the cabinet panel inspection images, acquisition station, acquisition time, primary equipment operating conditions, interval maintenance status, and a pre-established pressure plate reference association model. Based on the same cabinet panel coordinate system, it continuously verifies the pressure plate identity, physical slot, nameplate text, engagement / disengagement status, combined constraints of the engagement / disengagement status of the same panel, and the operating condition status. It generates recognition results that include at least the pressure plate misalignment type, misalignment location, misalignment object, corresponding operating condition, and traceable alarm records. This method does not use the pressure plate engagement / disengagement classification results in a single frame image as the final judgment basis. Instead, it first establishes a reference constraint relationship between the pressure plate object and the actual slot, and then maps the field-acquired images to this reference constraint relationship for identity verification, status verification, engagement / disengagement status combination constraint review, and operating condition association review. This ensures that each engagement / disengagement status recognition result corresponds to a clear pressure plate identity and physical slot, and is suitable for substation pressure plate misalignment recognition scenarios based on image recognition and video understanding.

[0020] Specifically, such as Figure 2 As shown: Before the cabinet is put into operation, after the overall cabinet renovation is completed, or after the pressure plate configuration is changed, the pressure plate reference association model establishment process is triggered, and a single cabinet is used as the smallest modeling unit for the reference record; the cabinet panel is set as a two-dimensional reference plane, and the fixed structural point at the upper left corner of the cabinet panel, the corner point of the outer frame of the cabinet, or the fixed installation point calibrated on site is selected as the origin of the coordinates. The horizontal coordinate is established horizontally to the right along the cabinet panel, and the vertical coordinate is established vertically downward along the cabinet panel. The coordinate unit is millimeters; in a typical relay protection cabinet, the width of the cabinet panel can be 800 mm and the height can be 2260 mm. These values ​​are determined according to the cabinet design drawings, manufacturer's data, or on-site measured data; for non-standard cabinets, the actual width, height, pressure plate matrix start and end range, and row and column spacing are written into the cabinet geometric parameter table for subsequent slot positioning, image mapping, and work area segmentation; When establishing the pressure plate reference association model, the cabinet design drawings, secondary circuit configuration table, protection setting sheet, typical operation ticket, and on-site verification record are read to form a basic data set of the cabinet. Row and column coordinates are established according to the actual arrangement order of the pressure plates on the cabinet panel, and each pressure plate is mapped to a unique physical slot. The physical slot is represented by a row and column code, for example, the 5th row and 8th column is recorded as R05C08, and the row and column code is converted into the geometric center coordinates under the cabinet panel coordinate system. The horizontal center distance of the pressure plates can be 40 mm to 50 mm, and the vertical center distance can be 30 mm to 40 mm. This range is obtained by converting the pressure plate matrix spacing in the cabinet layout drawing, or by determining the distance between the center points of adjacent pressure plates measured on-site. During on-site measurement, at least one set of adjacent pressure plates in the upper, middle, and lower parts of the cabinet are selected as measurement samples to obtain the horizontal center distance and the vertical center distance, and the minimum measured center distance in the same cabinet is taken as the conversion benchmark for allowable deviation. The allowable deviation of the geometric center coordinates can be selected as ±2.5 mm. This value is approximately equal to 6.25% of the minimum transverse center distance of 40 mm and 8.33% of the minimum longitudinal center distance of 30 mm, and is determined in conjunction with the cabinet installation tolerance, panel assembly deviation, and subsequent image reprojection residual. This deviation is less than half of the center distance between adjacent slots and is used to distinguish adjacent physical slots. When the actual center distance of the pressure plate is different from the above typical range, the allowable deviation can be determined according to 5% to 8% of the actual minimum transverse center distance and 6% to 10% of the actual minimum longitudinal center distance, and the allowable deviation should be less than half of the center distance between adjacent slots. The length of the pressure plate handle can be 30 mm to 40 mm, and the width can be 15 mm to 20 mm. This range is determined based on the actual size of commonly used relay protection pressure plates or the results of on-site calibration, and is used to determine the initial search area for handle contour extraction. For the shape of the pressure plate handle, the handle contour of the pressure plate in a stable posture is collected, and the outer contour of the handle is approximated as a hexagonal or octagonal geometric envelope. The millimeter coordinates of each vertex of the envelope in the panel coordinate system are recorded. For the nameplate text, the text content above the pressure plate or adjacent to the nameplate area is read, and the complete text string, number of characters, character height, character spacing, and coordinates of the nameplate outer rectangle are recorded. For multi-line nameplates, the text content, line number, and line spacing of each line are recorded separately, so that the nameplate recognition results in subsequent on-site images can be aligned and verified according to the character positions. The pixel reference point is generated from the front view of the cabinet panel image collected during the commissioning and calibration phase. The millimeter coordinates of the geometric center of the pressure plate in the cabinet panel coordinate system are converted into pixel coordinates through the coordinate mapping relationship established by the cabinet frame anchor points, nameplate anchor points and pressure plate matrix anchor points. When the camera resolution, shooting distance or cabinet installation position changes, the mapping relationship between the millimeter coordinates and pixel coordinates is updated again according to the cabinet structure anchor points to ensure that the subsequent inspection images correspond to the cabinet panel coordinate system. The pressure plate identification code is generated by combining the voltage level, cabinet number, bay number, and slot number. The code length can be selected to be no less than 12 digits to meet the combined expression requirements of voltage level, cabinet, bay, and slot. Among them, the first two digits can represent the voltage level, the middle six digits can represent the combined number of cabinet and bay, and the last four digits can represent the slot number. When there is already a unified coding rule in the station, it is generated according to the unified coding rule in the station, and the code is kept unique between different cabinets and different slots. After the pressure plate identification code is written into the pressure plate reference association model, it is used together with the physical slot and nameplate text to determine the pressure plate object, so that the subsequent status identification results correspond to the clear pressure plate identification and actual slot. For the pressure plate's engagement and disengagement states, the pressure plate reference association model records the geometric attitude range of the pressure plate handle under stable engagement, stable disengagement, and mechanical intermediate states. Optionally, the angle between the pressure plate handle and the horizontal reference of the cabinet panel is used as a state description quantity. An angle between 0 and 15 degrees is recorded as the first stable state, an angle between 75 and 90 degrees is recorded as the second stable state, and an angle between 15 and 75 degrees is recorded as the mechanical intermediate state. The first and second stable states are mapped to the engagement or disengagement states according to the commissioning calibration image, respectively. If the insertion and withdrawal directions of different types of pressure plates are opposite, the insertion and withdrawal postures recorded in the commissioning calibration image of the pressure plate shall be used as the reference, and the status codes corresponding to the first and second stable states shall be reversed. The above-mentioned included angle range is determined based on the mechanical stroke angle of common continuous pressure plates or small electromagnetic pressure plates, which is usually close to 85 to 90 degrees. It is reduced by about 10 degrees at both the insertion and withdrawal ends to cover the small attitude fluctuations caused by normal mechanical sway and running vibration, and to reserve the identification range for mechanical intermediate states such as half insertion and half withdrawal. A status table for operating conditions is established in the pressure plate reference association model. Normal operation, hot standby, cold standby, switch to maintenance, maintenance, fault handling, and test conditions are indexed, and the status for each pressure plate is written for each operating condition. For transitional operating conditions such as switch to maintenance, allowable transitional states and transitional time windows are set. The transitional time window can be selected to not exceed 300 seconds. This value is determined based on the historical time consumption statistics of primary equipment status switching, secondary safety measure execution, and pressure plate activation / deactivation confirmation in typical substation operation tickets, and is used to cover short-term transitional states during normal sequential control or manual confirmation processes. For intervals with longer maintenance processes or different on-site management requirements, transitional time windows are configured separately according to interval type in the status table for operating conditions. In the pressure plate reference correlation model, establish combination constraint rules for the on / off states of pressure plates in the same screen. These rules should include at least mutual exclusion groups, coordination groups, and quantity constraint groups. Mutual exclusion groups constrain that only one pressure plate in the same protection mode selection category is in the on / off state. Coordination groups constrain that related inlet and outlet pressure plates under the same protection function are simultaneously on / off. Quantity constraint groups constrain that the number of pressure plates in a related pressure plate group falls within a fixed integer range. The allowable range for the quantity constraint group is determined based on the protection configuration file, typical operation tickets, or pressure plate on / off rules confirmed by secondary professionals, and is defined as a closed interval. Write into the model, where and All values ​​are non-negative integers and are not greater than the total number of pressure plates in the associated pressure plate group; each pressure plate on-screen activation / deactivation status combination constraint rule includes at least the rule number, associated pressure plate identification code, applicable working conditions, status constraint conditions, and effective version number. After completing the above data processing, the reference information corresponding to each pressure plate is written into the same structured reference record. The structured reference record includes at least the pressure plate identification code, physical slot, geometric center, pixel reference point, handle geometric envelope, nameplate text, working condition status, and association rule index. All structured reference records of the same cabinet are encapsulated into a cabinet-level reference package. The cabinet-level reference package includes at least the version number, creation time, latest update time, applicable inspection station, and effective status identifier. If a pressure plate is added, removed, nameplate is replaced, or protection configuration is adjusted in the cabinet, only the changed fields are incrementally revised, and the unchanged fields retain their original values. The revised reference package serves as a candidate version and is verified in parallel with the effective version based on the same batch of inspection samples in the next inspection. The object verification pass rate for the candidate version is the ratio of the number of pallets that passed object identity verification to the total number of pallets participating in object identity verification in the same batch of inspection samples. The status verification pass rate is the ratio of the number of pallets that passed status verification to the total number of pallets participating in status verification. The stability of the associated pallet group verification is the ratio of the number of associated pallet groups that obtained the same associated pallet group verification conclusion in the same batch of inspection samples for the candidate version and the effective version to the total number of associated pallet groups participating in associated pallet group verification. When the candidate version is converted into the official effective version, the object verification pass rate and the status verification pass rate must not be lower than the effective version. The stability of the associated pallet group verification can be selected to be no less than 0.98, and no new abnormalities of the deployment and withdrawal status combination constraint that have not been manually confirmed should be added. The stability of the associated pallet group verification is set to 0.98 to allow a small number of individual differences caused by image quality, boundary calibration or on-site occlusion, while avoiding the introduction of batch same-screen pallet deployment and withdrawal status combination constraint rule offsets by the candidate version. If the candidate version has a verification rollback or adds an abnormality of the deployment and withdrawal status combination constraint that has not been confirmed, the effective version is retained and a record to be checked is generated.

[0021] Specifically, such as Figure 3 As shown: After the inspection task is started, the process of on-site image acquisition, time synchronization and reference model calling is executed; when the cabinet code and acquisition station code are issued in the inspection task, or when the local inspection plan reaches the inspection time of the corresponding cabinet, on-site acquisition is started; the acquired information includes at least the cabinet panel inspection image, acquisition station, acquisition time, shooting posture, supplementary light intensity, working status and reference package version information, etc., and is encapsulated into an acquisition message with time correspondence and spatial correspondence. Before the inspection equipment arrives at the cabinet, it queries the cabinet-level reference package index table according to the cabinet code in the inspection task to obtain the reference package version number, workstation code, shooting posture reference value, and working condition mapping table version corresponding to the cabinet. After the inspection equipment arrives at the preset workstation, the workstation is confirmed. Workstation confirmation can be completed through magnetic tag coding, visual-assisted positioning, or track mileage position. Optionally, when the magnetic tag code is consistent with the workstation code in the inspection task, and the attitude sensor outputs an attitude angle change of no more than 0.5 degrees or an angular velocity fluctuation of no more than 0.5 degrees per second within 500 milliseconds, the workstation is considered to have stopped successfully. The 500 milliseconds is determined based on the short-term mechanical vibration decay time and image acquisition trigger response time after the inspection equipment stops. The 0.5 degrees and 0.5 degrees per second are determined based on the conventional measurement accuracy of the inspection equipment's attitude sensor, the residual swing range after the gimbal is in position, and the allowable error of the pressure plate handle angle recognition, in order to reduce the impact of image jitter on image acquisition when the equipment just stops. After the workstation is confirmed, the acquisition device adjusts the shooting direction according to the shooting posture reference value in the cabinet-level reference package so that the effective area of ​​the cabinet panel enters the acquisition field of view; the effective area of ​​the cabinet panel is the area of ​​the cabinet panel that participates in the pressure plate recognition, and at least covers the pressure plate matrix area, the corresponding nameplate area and the cabinet structure anchor point area; the angle between the camera optical axis and the normal of the cabinet panel can be selected to be no more than 8 degrees, and this value is determined based on the multi-angle shooting test during the commissioning calibration phase; within this angle range, the residual of the pressure plate center after the structural anchor point reprojection can be controlled within the allowable deviation range of the geometric center. When the angle range is exceeded, the projection length and attitude angle error of the pressure plate handle will increase, which will reduce the stability of the commissioning and withdrawal status recognition; Image acquisition employs a multi-frame acquisition method; five frames of cabinet panel images can be acquired at the same workstation, with an optional 200-millisecond interval between adjacent frames. These five frames are used to retain at least three qualified images after quality screening. The 200-millisecond interval is determined based on the micro-vibration decay time of the inspection equipment, the indoor lighting flicker cycle, and the inspection cycle. For a 50Hz AC lighting environment, 200 milliseconds cover multiple lighting flicker cycles, reducing the impact of single-frame brightness fluctuations on recognition stability. A hardware timestamp is written to each image frame during acquisition, achieving millisecond-level time accuracy. The acquisition message includes at least the cabinet code, workstation code, image frame number, capture time, attitude angle, supplementary lighting level, image resolution, and acquisition device number. Auxiliary captured images are marked as backup frames and are retrieved when the quality of the main acquired image is insufficient or when verification is required. Within the same time window of image acquisition, the operating status and interval maintenance status of primary equipment are obtained. The operating status of primary equipment can be obtained from the monitoring backend, the dispatch automation system, or the one-click sequential control system. The status identifier should at least distinguish between normal operation, hot standby, cold standby, transfer for maintenance, maintenance, fault handling, and testing. The interval maintenance status can be obtained from the maintenance work ticket, operation ticket execution status, or local operation and maintenance confirmation records. The allowable deviation between the acquisition timestamp and the operating status snapshot timestamp can be selected to be no more than 2 seconds. This value is determined based on the status refresh cycle of the station monitoring backend or the one-click sequential control system, the network transmission delay, and the inspection processing cycle, so that the image data and the operating status data correspond to the same operation stage. If the above allowable deviation is exceeded, it is determined that the current working condition snapshot and image data are out of sync, and a new working condition snapshot is requested. When requesting a new working condition snapshot, a maximum of 3 requests can be made, with an interval of 1 second between each request. The 3 requests are used to cover short-term jitter in the station network and delays in working condition status refresh. The 1-second interval matches the common status refresh cycle and can avoid excessive waiting time for a single workstation. If a valid working condition snapshot cannot be obtained after three requests, the working condition snapshot is marked as unavailable, and object verification, status verification, and deployment / return status combination constraint verification are performed first, and the working condition coupling review is suspended. After image acquisition and working condition synchronization are completed, the pressure plate reference association model corresponding to the acquisition station is called. During the call, the cabinet code, station code, inspection task number, and reference package version number are verified simultaneously. If the locally cached reference package version number is consistent with the effective version number, the local reference package is called. If the two are inconsistent, the corresponding differential update data is obtained and written to the local cache. The differential update content is consistent with the field-level revision record in the reference association model. After the call is completed, a record to be verified is established for this cabinet inspection. The record to be verified is indexed by the cabinet code and acquisition time, and includes at least the following fields: image frame list, station confirmation status, working condition snapshot, interval maintenance status, reference package version number, and acquisition quality identifier. Basic quality checks are performed on the acquired images. For 8-bit grayscale images, if the average exposure value is below 30 grayscale or above 220 grayscale, it is considered an exposure anomaly. A grayscale value below 30 typically corresponds to too many underexposed areas, while a grayscale value above 220 typically corresponds to too many saturated highlight areas, both of which affect the stable extraction of the panel structure anchor points, nameplate text, and pressure plate handle edges. Image sharpness scoring can be determined using Laplacian variance, edge gradient mean, or pressure plate edge sharpness scoring. Sharpness scoring is preferentially calculated within the panel structure anchor point area, pressure plate matrix area, and nameplate area, and a weighted average is taken as the image sharpness score. The baseline sharpness threshold can be selected as 70% of the average sharpness score of the commissioning calibration images. When the on-site image sharpness score is below this threshold, it is considered a blur anomaly. The 70% percentage is used to allow for normal fluctuations caused by on-site lighting and slight vibrations, while excluding images that cannot meet the requirements for image positioning and status recognition. If the area of ​​obvious occlusion exceeds 20% of the effective area of ​​the cabinet panel, it is judged as an occlusion anomaly. Obvious occlusion areas can be determined based on low-texture continuous areas, abnormally bright reflective areas, human or tool outline areas, and areas where the difference from the commissioning calibration image exceeds the grayscale difference threshold. The 20% proportion is determined based on the distribution of the pressure plate matrix area in the effective recognition area of ​​the cabinet, and is used to distinguish between local occlusion and large-area occlusion that affects anchor point extraction and working area segmentation. When the number of abnormal images in 5 frames exceeds 2 frames, the re-import process is started to ensure that the number of qualified images entering subsequent processing is no less than 3 frames. During supplementary sampling, the supplementary lighting level is adjusted in increments of 10%, with a maximum of 3 adjustments. The 10% increment is used to avoid excessive brightness changes that could lead to overexposure or increased reflection. The 3 adjustments correspond to a maximum brightness correction range of approximately 30%, which can cover common indoor lighting fluctuations and limit the supplementary sampling time per station. If at least 3 qualified images cannot be obtained after supplementary sampling, the station is marked as a low-confidence acquisition state. In the low-confidence acquisition state, the conclusion of pressure plate misalignment is not directly output. Instead, the record of this station is sent to the auxiliary image verification or manual verification branch. The resulting record to be verified serves as the input for structural anchor point extraction, coordinate reconstruction, and pressure plate working area segmentation.

[0022] Specifically, such as Figure 4 As shown: When there are no fewer than 3 qualified cabinet panel inspection images in the record to be verified, the cabinet structure anchor points are extracted, coordinates are reconstructed, and the pressure plate working area is segmented; the input data includes at least the cabinet panel inspection images, the panel coordinate system in the cabinet-level reference package, the cabinet geometric parameters, the pressure plate reference slot, the pressure plate handle geometric envelope, and image quality records; by converting the field-acquired images to the cabinet panel coordinate system consistent with the pressure plate reference association model, an independent working area is generated for each pressure plate under the unified coordinate system; Anchor point extraction uses the inherent structure of the cabinet body as the positioning basis; the cabinet structure anchor points include at least the following types: cabinet outer frame anchor points, nameplate anchor points, and pressure plate matrix structure anchor points; the cabinet outer frame anchor points correspond to the four outer corner points of the cabinet panel or the intersection of the frame lines; the nameplate anchor points correspond to the four corner points of the nameplate on the top of the cabinet or the endpoints of the long side of the nameplate; the pressure plate matrix structure anchor points correspond to the four corner points of the outer rectangle of the pressure plate matrix and the geometric center of the outermost pressure plate; for cabinets with inconspicuous top nameplates, fixing screws, panel dividing lines, or door panel edges can be used as alternative anchor points, and the panel coordinates of the alternative anchor points are pre-written into the cabinet-level reference package; For each frame of image that passes quality inspection, grayscale conversion, edge enhancement, and morphological closing operations are performed to extract continuous straight line segments and rectangular candidate contours. For the outer frame anchor points of the cabinet, long straight line segments are searched in the image edge region, and outer rectangles are selected based on aspect ratio, edge closure, and proportional relationship with the cabinet's reference size. Edge closure can be expressed as the ratio of the closed side length of the candidate contour to the perimeter of the theoretical rectangle. For the nameplate anchor points, horizontal rectangular regions are searched in the upper region of the cabinet, and candidate nameplate regions are determined by combining the text texture density within the region. Text texture density can be expressed as the ratio of the number of character edge pixels in the candidate region to the area of ​​the candidate region. For the pressure plate matrix structure anchor points, small structural components that are repeatedly arranged are detected in the pressure plate candidate region, and the outer range of the pressure plate matrix is ​​selected by the consistency of row and column spacing. The consistency of row and column spacing is expressed as the ratio of the standard deviation to the mean of the distance between adjacent candidate center points. When this ratio does not exceed 15%, and the distance between adjacent candidate center points falls within the pixel conversion range corresponding to the horizontal and vertical center distances recorded in the reference package, the candidate area is used as a candidate anchor point for the pressure plate matrix structure. The 15% ratio is determined based on the installation tolerance of the pressure plate on the same screen, the perspective residual of the shooting, and the edge positioning error, and is used to distinguish between regular pressure plate matrices and non-pressure plate structures. In a typical screen cabinet, this pixel conversion range corresponds to a horizontal center distance of 40 mm to 50 mm and a vertical center distance of 30 mm to 40 mm. After obtaining various anchor point candidates, cross-validation is performed based on the relative positional relationships of the anchor points recorded in the baseline package. Cross-validation should at least verify the relative height relationship between the outer frame of the cabinet and the nameplate anchor point, the longitudinal distance relationship between the nameplate anchor point and the upper boundary of the pressure plate matrix, and the left and right margin relationship between the outer rectangle of the pressure plate matrix and the cabinet frame. When all three types of anchor points pass cross-validation, it is determined to be a complete and valid anchor point set. When only two types of anchor points pass the validation, a dual-type anchor point mode is adopted. When only one type of anchor point is available, a single-type anchor point auxiliary estimation mode is adopted. The initial value of geometric confidence can be 1, and it is adjusted according to the available anchor point categories, the average reprojection residual, and the anchor point cross-validation results. In the single-type anchor point auxiliary estimation mode, the geometric confidence is reduced by 15% to 30%, and the reduction ratio can be selected as 25%. This ratio is determined based on the decrease in the number of coordinate transformation constraints after the reduction of anchor point categories, and is used to reflect the increased uncertainty of single-type anchor point estimation relative to multi-type anchor point joint positioning. When all three types of anchor points are unavailable, an anchor point extraction failure record is generated, and the process is transferred to the acquisition anomaly handling procedure. After determining the effective anchor point set, the pixel coordinates of the anchor points in the current image are paired with the millimeter coordinates of the anchor points in the reference package to establish a geometric transformation relationship from image coordinates to cabinet panel coordinates; this geometric transformation relationship can be represented by planar homography transformation, denoted as: ; in, Let i be the pixel coordinates of the i-th anchor point in the current image. Here, H represents the millimeter coordinates of the corresponding anchor point in the coordinate system of the cabinet panel, H is the geometric transformation matrix, and π represents homogeneous coordinate normalization. The above relationship is used to determine the geometric transformation matrix that minimizes the reprojection error between the current image anchor point and the reference anchor point. To reduce the impact of false anchor points on coordinate reconstruction, an outlier anchor point removal mechanism is adopted when calculating the geometric transformation matrix. If the reprojection residual of an anchor point exceeds three times the median of all anchor point residuals, the anchor point is removed from the pairing set and the geometric transformation matrix is ​​recalculated. The median of three times is determined based on the judgment method of majority sample consistency and minority sample deviation in outlier point screening. It is used to identify outlier anchor points that are significantly inconsistent with the deviation of the majority anchor points and reduce the impact of local reflections, occlusions, or false anchor points on the overall coordinate reconstruction. After the transformation is completed, the rotation component, translation component, scale component, perspective component, and average reprojection residual are recorded. The average reprojection residual can be set to no more than 2.5 mm. This value is consistent with the allowable deviation of the geometric center of the pressure plate and is used to ensure that the coordinate reconstruction accuracy meets the requirements of slot positioning. When the average reprojection residual is greater than 2.5 mm but not greater than 5 mm, the coordinate reconstruction is marked as a low-confidence geometric transformation, and the low-confidence identifier is written into the geometric quality field of the sub-record to be verified. Subsequent object identity verification and deployment / return status verification need to be combined with backup images or manual review and confirmation. 5 mm is twice the allowable deviation of the geometric center and is used to distinguish between low-confidence geometric transformations that can be degraded and coordinate reconstruction results that are not suitable for continued use in slot identification. When the average reprojection residual is greater than 5 mm, the coordinate reconstruction is judged to have failed and the process is transferred to the acquisition anomaly handling procedure. After coordinate reconstruction, the inspection images of the cabinet panels are reprojected to ensure that the reprojected images are consistent with the panel reference plane in the reference package in terms of orientation, scale, and pressure plate matrix position. Then, the pressure plate working area is divided according to the geometric center, physical slot, and handle geometric envelope of each pressure plate in the reference package. The pressure plate working area covers the pressure plate handle area, the corresponding nameplate area, and the necessary edge allowance, while avoiding the main identification areas of adjacent pressure plates. The lateral boundary of the pressure plate working area can be taken as 0.45 times the lateral center distance to the left and right of the pressure plate center, and the longitudinal boundary can be taken as 0.55 times the longitudinal center distance to the top and bottom of the pressure plate center, and extended towards the nameplate direction by 1.2 times the nameplate height; the lateral 0.45 times makes the total width of the working area 0.9 times the lateral center distance, which is used to cover the pressure plate handle and reserve an isolation strip for adjacent slots; the longitudinal 0.55 times makes the total height of the working area 1.1 times the longitudinal center distance, which is used to cover the vertical projection changes of the pressure plate handle under the insertion, withdrawal and intermediate mechanical postures; the nameplate height of 1.2 times is used to reserve the top and bottom edge margins for text recognition, avoiding the top or bottom of the characters being cut off; for cabinets with irregular row and column spacing, the working area polygon boundary recorded in the reference package can be used for segmentation; After segmentation, a verification sub-record is created for each pressure plate. The verification sub-record includes at least the pressure plate identification code, slot code, working area image, reprojection coordinates, geometric transformation residual, geometric quality field, corresponding nameplate area, handle candidate area, and image quality status. If the effective pixel area of ​​a working area is less than 80% of the reference area, or the grayscale histogram shows a single-peak saturation state, the working area is marked as incomplete. The reference area is the pixel area of ​​the corresponding pressure plate working area polygon in the reference package converted at the current reprojection scale. The 80% ratio is determined based on the lowest identifiable area of ​​the pressure plate handle body, nameplate area, and slot center in the working area, which allows for local occlusion or slight clipping at the edge of the working area. A single-peak saturation state refers to a working area where more than 70% of the pixels are concentrated in the grayscale value range of 0 to 20, or more than 70% of the pixels are concentrated in the grayscale value range of 235 to 255. This is used to identify unusable working areas caused by severe underexposure or high saturation. The 70% ratio is determined based on the impact of underexposure and high saturation areas in the 8-bit grayscale image on edge extraction, text recognition, and handle contour segmentation. When the working area is incomplete, the pressure plate misalignment conclusion is not directly output. Instead, the working area is transferred to the abnormal branch, and the backup image is called first to re-extract anchor points, reconstruct coordinates, and segment the working area. If the backup image still cannot generate a complete working area, the corresponding pressure plate is marked as pending verification. All pressure plate working areas and pending verification sub-records are used as inputs for subsequent object identity verification and deployment / return status verification.

[0023] Specifically, such as Figure 5 As shown: After the pressure plate working area is divided and the sub-records to be verified are generated, each pressure plate is sequentially subjected to object identity verification and engagement / disengagement status verification. The input data includes at least the image of the pressure plate working area, the corresponding pressure plate reference record, the cabinet panel coordinate system, geometric quality fields, the current primary equipment operating condition, the interval maintenance status, and the status table to be in operation. The object identity verification is performed before the engagement / disengagement status verification, so that the engagement / disengagement status identification result corresponds to a clear pressure plate identity and the actual physical slot. Object identity verification is performed when the pressure plate working area has been marked as complete and the geometric transformation residual is no greater than 2.5 mm; when the geometric transformation residual is greater than 2.5 mm but no greater than 5 mm, the object identity verification of the pressure plate is marked as low confidence verification, and the backup image or manual review result is called to participate in the confirmation; when the geometric transformation residual is greater than 5 mm, object identity determination is not performed, and the corresponding pressure plate is recorded as pending review; 2.5 mm is consistent with the allowable deviation of the pressure plate geometric center, and 5 mm is twice the allowable deviation, used to distinguish between low confidence scenarios that can continue to be verified and scenarios that are not suitable for continuing to be used for slot assignment determination; Object identity verification is performed from at least three dimensions: geometric identity, textual identity, and shape identity. Geometric identity verification reads the geometric center, slot code, and pressure plate handle geometric envelope from the reference record, and compares the pressure plate center, estimated slot, and handle circumscribed contour extracted from the work area with the reference record. The geometric center deviation threshold can be selected as ±2.5 mm, which is consistent with the allowable deviation of the pressure plate geometric center and is used to determine whether the pressure plate center is still within the binding slot range. The handle envelope vertex deviation threshold can be selected as ±3.5 mm, which is determined by adding approximately 1 mm edge segmentation error to the 2.5 mm allowable slot center deviation. The edge segmentation error is determined based on the positioning fluctuation of the pressure plate handle edge under reflective, shadow, and partial occlusion conditions. The geometric identity score is denoted as G. G is set to 1 when the geometric center deviation is no greater than 2.5 mm, the slot codes are consistent, and the shank envelope vertex deviation is no greater than 3.5 mm. G is set to 0.5 when the geometric center deviation is greater than 2.5 mm but no greater than 3.5 mm and the slot codes are consistent, or when the shank envelope vertex deviation is greater than 3.5 mm but no greater than 4.5 mm and the slot codes are consistent. G is set to 0 when the slot codes are inconsistent, the geometric center deviation is greater than 3.5 mm, or the shank envelope vertex deviation is greater than 4.5 mm. 3.5 mm is used as the upper limit of the suspected geometric center range, determined by adding approximately 1 mm of boundary fluctuation to the 2.5 mm slot allowable deviation. 4.5 mm is used as the upper limit of the suspected shank envelope range, determined by adding approximately 1 mm of segmentation fluctuation to the 3.5 mm envelope vertex deviation, used to cover slight contour offsets caused by localized reflections, shadows, or missing edges. Text identity verification performs text recognition on the nameplate area in the work area. Before recognition, the text area is located based on the bounding rectangle of the nameplate in the baseline record, and the text area is binarized, tilted, and segmented. The text recognition result is aligned with the baseline nameplate text according to the character position, and the number of characters, character order, character height, and overall similarity are compared. The overall similarity can be determined according to the character-level edit distance, denoted as: ; Where D is the edit distance between the identified text string and the reference text string, and L is the number of characters in the longer text string between the identified text string and the reference text string; the text identity score is denoted as W; when the overall similarity T is greater than or equal to 0.92, and the number of characters and the character order are consistent with the reference nameplate text, W is 1; when T is greater than or equal to 0.85 and less than 0.92, and there is no whole string misalignment or adjacent nameplate substitution, W is 0.5; when T is less than 0.85, or the number of characters and the character order are significantly inconsistent with the reference nameplate text, W is 0; 0.92 is determined statistically based on the nameplate recognition results in the commissioning calibration image and historical qualified inspection images, and is used to allow for individual similar character recognition errors; 0.85 is determined based on the similarity distribution of adjacent nameplate substitution, multi-character misidentification, and character order disordered samples in historical inspections, and is used to exclude obvious text mismatches; When the nameplate text is less than 4 characters, the proportion of text identity in the fusion judgment is reduced, while the proportion of geometric identity in the fusion judgment is increased; the limit of 4 characters is determined based on the sensitivity of character-level edit distance. When the nameplate has less than 4 characters, a single character recognition error can cause the overall similarity to drop by more than 25%, which can easily amplify the occasional recognition error of short nameplates. The shape identity verification performs contour segmentation on the pressure plate handle and pressure plate base, and extracts features such as handle aspect ratio, number of contour corner points, local texture description and grayscale distribution; shape identity is used to determine whether there is a pressure plate model replacement, missing pressure plate or foreign object obstruction in the slot, but is not used to determine whether the pressure plate is put in or taken out; in order to reduce the influence of the current putting or taking out posture on the shape judgment, the handle contour can be normalized according to the principal axis direction and compared with the reference envelope, or rotation invariant shape description quantity can be used for comparison; The appearance identity score is denoted as A; when the shape consistency is greater than or equal to 0.85 and the distance between the texture feature and the baseline feature does not exceed twice the standard deviation of the mean distance of the baseline samples, A is 1; when the shape consistency is greater than or equal to 0.75 and less than 0.85, or the distance between the texture feature and the baseline feature is greater than twice the standard deviation but not more than three times the standard deviation, A is 0.5; when the shape consistency is less than 0.75, or the distance between the texture feature and the baseline feature exceeds three times the standard deviation, A is 0; 0.85 is determined based on the contour similarity distribution of the same type of pressure plate under different postures and different lighting conditions, and is used to preserve the normal imaging differences of the same type of pressure plate; 0.75 is determined based on the contour similarity distribution of pressure plate missing, model replacement, and foreign object occlusion samples, and is used to distinguish suspected shape deviations and clear shape anomalies; the baseline samples are extracted from commissioning calibration images and historical qualified inspection images, with twice the standard deviation used to cover the texture fluctuations of the same pressure plate under normal lighting changes, and three times the standard deviation used to identify abnormal appearances that significantly exceed the normal texture fluctuation range; Geometric identity, textual identity, and physical appearance identity are weighted to form an object identity fusion score F; when the nameplate text has no fewer than 4 characters, F is determined according to the following formula: ; When the nameplate text has fewer than 4 characters, F is determined according to the following formula: ; Wherein, G is the geometric identity score, W is the text identity score, and A is the shape identity score; the weights are determined based on the indicative strength of the three types of identity information in the identification of misaligned pressure plates. The geometric identity directly corresponds to the actual slot and has the strongest indicative power for misaligned slots; the text identity is used to verify the nameplate object and can identify nameplate substitution or inconsistent object identification; the shape identity is used to supplement the identification of pressure plate model replacement, missing pressure plates, and foreign object obstruction. When the object identity fusion score F is greater than or equal to 0.85, the object identity is considered passed; when F is greater than or equal to 0.65 and less than 0.85, the object identity is considered suspected to be inconsistent, and a backup image is called for secondary verification; when F is less than 0.65, the object identity is considered inconsistent; 0.85 and 0.65 are determined based on the fusion score distribution of the three types of samples in the historical calibration samples: correct object, suspected object, and incorrect object, and are used to form the segmented judgment of pass, suspected inconsistency, and inconsistency; when the object identity is inconsistent, the regular submission and withdrawal status correspondence is not performed, and the geometric deviation, textual inconsistency, or abnormal shape is written into the verification relationship and transferred to the misalignment type classification process; After the object's identity is verified, the engagement / disengagement status is checked. The engagement / disengagement status check is based on the posture of the pressure plate handle. The main line segment of the handle is extracted within the working area of ​​the pressure plate. The main line segment can be obtained through edge detection, line fitting, or skeleton extraction. The angle between the main line segment of the handle and the horizontal reference of the cabinet panel is recorded as the state angle θ. The measurement accuracy of the state angle can be selected to be no less than 2 degrees. This value is determined based on the visible length of the pressure plate handle, edge positioning error, and line fitting error under the conditions of 4096 pixels by 3072 pixels or equivalent resolution image, and is used to meet the requirements for distinguishing between stable engagement, stable disengagement, and mechanical intermediate state. The state angle θ is matched with the first stable state, second stable state, and mechanical intermediate state recorded in the pressure plate reference association model. When the state angle θ is between 0 and 15 degrees, it is recorded as the first stable state; when it is between 75 and 90 degrees, it is recorded as the second stable state; and when it is between 15 and 75 degrees, it is recorded as the mechanical intermediate state. The first and second stable states are then determined as the input state or the output state according to the state code mapping relationship in the commissioning calibration image. If the input direction and output direction of different pressure plate models are opposite, the state transition is performed according to the mapping relationship of that pressure plate model in the reference record. When the state angle is in the boundary range of 15 to 20 degrees between the first stable state and the mechanical intermediate state, or in the boundary range of 70 to 75 degrees between the second stable state and the mechanical intermediate state, a boundary state flag is written; the width of the boundary range is 5 degrees, which is 2 degrees greater than the state angle measurement accuracy, and is used to cover the boundary fluctuations caused by angle measurement errors and short-term mechanical oscillations. Consistency fusion is performed on the state recognition results of the same pressure plate in 5 frames of images. If the state is consistent in at least 4 frames, the consistent state is taken as the actual measured state of the pressure plate. If it is consistent in less than 4 frames, the backup image is called for verification. At least 4 frames of consistency corresponds to an 80% inter-frame consistency ratio, which can tolerate 1 frame of abnormality caused by jitter, occlusion or recognition error in 5 frames of acquisition, while avoiding the majority judgment of 3 frames being too sensitive to occasional misidentification. If neither the main acquisition image nor the backup image can achieve 4 frames of consistency, and the boundary state flag is true, the pressure plate is recorded as a suspected mechanical intermediate state. If the image quality is abnormal and the state angle cannot be obtained stably, it is recorded as an undeterminable state and written into the low confidence verification record. After obtaining the measured status, retrieve the current primary equipment operating condition and interval maintenance status, and look up the required status for the pressure plate under the current operating condition in the operating condition status table; when the measured status and the required status are consistent, generate a status corresponding pass record; when the measured status and the required status are inconsistent but the operating condition is in a permissible transitional state and the transition time window recorded in the operating condition status table is not exceeded, generate a transitional state record; when the measured status and the required status are inconsistent and do not belong to a permissible transitional state, generate a status misalignment record. Each verification relationship includes at least the following fields: pressure plate identity code, actual slot, geometric identity conclusion, textual identity conclusion, external shape identity conclusion, object identity fusion score, measured status, expected status, status angle, inter-frame consistency ratio, current operating condition, interval maintenance status, and verification time. This verification relationship also retains the object identity basis and status judgment basis, serving as input for the verification of the combination constraint of the activation / deactivation status and the verification of the operating condition association.

[0024] Specifically, such as Figure 6 As shown: After all pressure plate verification relationships are generated, the following steps are performed: verification of the combination constraints of the pressure plate activation / deactivation status on the same screen, verification of the correlation between primary and secondary equipment operating conditions, determination of misalignment types, and generation of alarm records. The input data includes at least all pressure plate verification relationships, the combination constraints of the pressure plate activation / deactivation status on the same screen, the operating condition status table, the current operating condition of the primary equipment, the interval maintenance status, the work order permission record, and the image quality label. By verifying the verification relationships of individual pressure plates at the pressure plate group level and the operating condition level, composite misalignments that pass the individual verification but have abnormal correlation between the pressure plate group relationships or abnormal operating condition relationships are identified. The combined constraint review of the activation / deactivation status reads the mutual exclusion group, coordination group, and quantity constraint group within the same cabinet. For mutual exclusion groups, the number N of pressure plates in the activation state within the group is counted. If N equals 1 and all other pressure plates are in the deactivation state, the mutual exclusion group passes. If N equals 0 or is greater than 1, a mutual exclusion group violation record is generated. The activation quantity target of the mutual exclusion group is determined by the protection mode selection rule and written into the combined constraint rule of the activation / deactivation status of the pressure plates in the same cabinet. For the coordination group, check whether the status of the pressure plates within the group is consistent. If all pressure plates in the group are engaged or disengaged, the coordination group passes. If there is a mixture of engaged and disengaged statuses within the group, a coordination group violation record is generated. The coordination group is used to verify the consistency of status between inlet pressure plates, outlet pressure plates, or related function pressure plates under the same protection function. For the quantity constraint group, the number of input states N within the group is counted and correlated with the allowable interval in the benchmark model. When comparing, If the quantity constraint group passes, then the quantity constraint violation record is generated; otherwise, a quantity constraint violation record is generated; the allowed range is... The rules for engaging and disengaging the pressure plate are determined by the protection configuration file, typical operation tickets, or rules confirmed by secondary professionals. and All are non-negative integers and are not greater than the total number of pressure plates in the corresponding associated pressure plate group; The severity of violations in the constraint rules for the simultaneous on / off states of pressure plates is determined according to the type of the associated pressure plate group; for mutually exclusive groups, the severity of violations is determined according to |N-1|; for quantity constraint groups, when N is less than 1... At that time, the difference in violations was When N is greater than At that time, the difference in violations was For coordinated groups, the severity of the violation is determined by the ratio of the number of inconsistent pressure plates within the group to the total number of pressure plates within the group; thus, mutually exclusive violations, coordinated violations, and quantity constraint violations are transformed into recordable combined constraint violation indicators. After completing the constraint verification of the commissioning and decommissioning status combination, a verification of the association between primary and secondary equipment operating conditions is performed; all pressure plates in the current cabinet are sorted according to their identification codes to form a measured state vector A, where each element represents the measured state of the corresponding pressure plate; the baseline state vector B under the current primary equipment operating condition is read from the operating condition status table, and compared bit by bit according to the same identification code order; if and If there is a discrepancy, the i-th pressure plate is added to the inconsistency list; then the interval maintenance status and work order permission record are read to filter the inconsistency list. A valid work order must meet at least the following conditions: the work order number is consistent with the current interval, the permitted time covers the current collection time, the permitted content includes the temporary state of the corresponding pressure plate or the pressure plate group associated with the corresponding pressure plate, and the work order is in the permitted execution state. If an inconsistent state has been permitted by a valid work order and is still within the allowed transition time window recorded in the work condition status table, then the item is marked as permitted transition. If it has not been permitted by a valid work order or has exceeded the allowed transition time window, then a work condition logic misalignment record is generated. The misalignment type is determined based on the results of object identity verification, activation / deactivation status verification, activation / deactivation status combination constraint review, and working condition association review. Misalignment types include at least position misalignment, identification misalignment, status misalignment, combination constraint misalignment, and working condition logic misalignment. Position misalignment is determined based on records of geometric identity deviation and inconsistencies between the actual slot and the reference slot. Identification misalignment is determined based on records of geometric identity matching but nameplate text mismatch. Status misalignment is determined based on records of object identity verification passing, actual status inconsistent with the current working condition's expected status, and not belonging to a permitted transition. Combination constraint misalignment is determined based on violation records of mutual exclusion groups, coordination groups, or quantity constraint groups. Working condition logic misalignment is determined based on records of passing individual status verification and activation / deactivation status combination constraint review, but inconsistent working condition association review. If multiple misalignment types are triggered simultaneously on the same pressure plate or the same associated pressure plate group, the main alarm type is determined in the order of position misalignment, identifier misalignment, status misalignment, combined constraint misalignment, and working condition logic misalignment. The remaining misalignment types are written as additional types into the same alarm record and pushed out at once with the main alarm type. Alarm logs should include at least the following fields: pressure plate identification code, cabinet code, slot code, misalignment type, main alarm type, supplementary alarm type, misalignment occurrence time, current operating condition, interval maintenance status, trigger rule number, evidence image index, geometric deviation, text similarity, status angle, inter-frame consistency ratio, associated pressure plate group violation number, reason for inconsistency in operating conditions, and review status. The reliability of misalignment is jointly determined by object attribution, status stability, consistency of constraint rules for the combination of pressure plate activation and deactivation status in the same screen, operating condition synchronization, and work order permission result. For state misalignment, confidence level Determined according to the following formula: ; in, For object identity fusion score, The percentage of frames with consistent state. The validity score is given for the snapshot time of the working condition. The value is set to 1 when the time difference between the data collection timestamp and the operating condition snapshot timestamp is no more than 2 seconds; 0.5 when the time difference is greater than 2 seconds but no more than 5 seconds; and 0 when the operating condition snapshot is unavailable or the time difference is greater than 5 seconds. The 2-second deviation is consistent with the allowable deviation of operating condition synchronization, and the 5-second deviation is determined based on the maximum acceptable delay after the sum of three re-requests of operating condition snapshots, each with a 1-second interval, and the network response time within the station. The values ​​of 0.45, 0.35, and 0.20 are determined based on the degree of influence of object attribution, state stability, and operating condition synchronization on the reliability of state misalignment alarms. Among these, incorrect object attribution directly leads to incorrect state attribution and has the highest weight. The consistency of multiple state frames reflects the stability of the measured state and has the second highest weight. The validity of the operating condition snapshot time is used to eliminate misjudgments caused by operating condition asynchrony and serves as an auxiliary weight. For positional misalignment, confidence level Determined according to the following formula: ; in, For geometric deviation score, Points will be awarded for textual discrepancies. Score for abnormal appearance; Take 1 when the slot code is inconsistent or the geometric center deviation is greater than 3.5 mm; take 0.5 when the geometric center deviation is greater than 2.5 mm but not greater than 3.5 mm; and take 0 when the geometric center deviation is not greater than 2.5 mm. The identity score W is determined in reverse order based on the textual identity score. When W is 0... When W is 1 and W is 0.5 When W is 0.5 and W is 1 Set to 0; The identity score A is determined in reverse order based on physical appearance score A; when A is 0... When A is 1 and A is 0.5 When A is 0.5 and A is 1 The values ​​are 0; 0.60, 0.25 and 0.15 are determined based on the degree of influence of the geometric deviation of the slot and the nameplate markings not conforming to the shape on the object ownership judgment in the position misalignment. Among them, the geometric deviation directly reflects the binding relationship between the pressure plate and the slot and has the highest weight. For misaligned combined constraints, the confidence level is determined by the identity fusion score of each pressure plate object within the group, the consistency ratio of state frames within the group, and the degree of violation of the combined constraint rules for pressure plate deployment and withdrawal states on the same screen; for mutually exclusive groups, the degree of violation of the combined constraint rules for pressure plate deployment and withdrawal states on the same screen is determined according to... Normalization yields; for the quantity constraint group, the violation degree of the constraint rule for the combination of the same screen pressure plate deployment and deployment states is determined according to N and the allowable interval. The difference in violations between them is normalized to obtain the result; for collaborative groups, the degree of violation of the constraint rule for the combination of the on-screen pressure plate deployment and withdrawal states is obtained according to the ratio of the number of pressure plates with inconsistent states within the group to the total number of pressure plates within the group; the normalization process is based on the total number of pressure plates within the corresponding associated pressure plate group to make the degree of violation between associated pressure plate groups of different sizes comparable. For operational condition logic misalignment, the confidence level is determined by the operational condition matching score, work order permission score, and time validity score. The operational condition matching score is 1 when the measured state is inconsistent with the expected state, 0 when it is in an allowed transition state and has not exceeded the transition time window, and 0.5 when the operational condition state is unavailable or the state source is incomplete. The work order permission score is 1 when there is no valid work order permission, 0 when the content of the valid work order permission covers the corresponding temporary state of the pressure plate, and 0.5 when the work order content partially matches the pressure plate identity or operational condition state. The time validity score is based on the operational condition snapshot time validity score. Confirmed; the combined weight of the working condition matching score and the work order permission score shall not be less than 0.70, which is used to highlight the decisive role of the current working condition and permission status on the working condition logic misalignment; the confidence weight of different misalignment types can be written into the cabinet-level benchmark package and updated with the benchmark package version; Abnormal branch processing and alarm output are executed synchronously; when the working condition snapshot is unavailable, no confirmation alarm is output for working condition logic misalignment, only a working condition supplementation identifier is output, and the object identity verification, deployment / retraction status verification, and deployment / retraction status combination constraint review results are retained; when the coordinate reconstruction is in a low confidence state, a low confidence geometric identifier is added to the alarm record, and the position misalignment alarm is output after verification by backup image or manual review; when the object identity fusion score is in the suspected range, the alarm type is first marked as suspected position misalignment or suspected identifier misalignment, and then converted to a confirmation alarm after the next inspection cycle or manual verification; When at least three pressure plates in the same row or column simultaneously exhibit geometric deviations in the same direction, and the difference between the deviation of each pressure plate and the average deviation of that row or column does not exceed 1 mm, or the coefficient of variation of the deviation does not exceed 20%, it is determined that the overall offset of the reference package is suspected. The three values ​​are used to rule out isolated anomalies caused by misinstallation of a single pressure plate or partial obstruction. The 1 mm value is consistent with the edge segmentation fluctuation of the pressure plate. The 20% value is used to judge the consistency of the offset in the same direction. When it is determined that the overall offset of the reference package is suspected, the formal misalignment alarm for that row or column is suspended, a reference review work order is generated, and the relevant pressure plate record is marked as a reference pending review status. When the processing delay exceeds the time budget for a single workstation, a phase record of the completed steps is output, and the incomplete steps are marked as pending review. The time budget for a single workstation can be selected as 60 seconds. This value is determined based on the inspection cycle of a single cabinet, image processing time, waiting time for working condition snapshot requests, and abnormal re-sampling time, in order to avoid the long-term occupation of a single workstation affecting the entire round of inspection tasks. The phase record should include at least the following fields: completed verification items, incomplete verification items, currently discovered anomalies, low confidence markers, and suggestions for continued review. After alarm records are generated, they are output according to the misalignment type and handling priority. Location misalignment and identification misalignment are directly related to the ownership of the pressure plate object and generate instant push records. Status misalignment and combined constraint misalignment generate maintenance pending records. Operating condition logic misalignment generates operation mode verification records or maintenance permit verification records. Each alarm record is also written back to the event log of the benchmark association model. The event log includes at least the fields of alarm record number, associated benchmark package version number, trigger rule number, review status, handling result and update time, which are used for subsequent field-level updates, threshold correction and benchmark package version review.

[0025] Example 2: Based on Example 1, the specific application process of a substation pressure plate misalignment identification method is further explained: Using a 110kV line protection cabinet as the identification object, the cabinet code is P110-L01, and the inspection station code is W03. The cabinet is equipped with relay protection pressure plates such as line protection inlet pressure plate, line protection outlet pressure plate, reclosing function pressure plate, maintenance status pressure plate, and standby function pressure plate. When the cabinet is in normal operation, the line protection inlet pressure plate and line protection outlet pressure plate should be in the engaged state, and the maintenance-related pressure plates should be in the disengaged state. When the line bay is switched to maintenance mode, the status of the outlet pressure plate and maintenance status pressure plate should change according to the maintenance safety measures. If adjacent pressure plates of the same model are swapped after maintenance is restored, the nameplate mounting position is offset, or the operator only performs the engagement / disengagement operation based on the currently visible position, it is difficult to determine whether the pressure plate is located in the correct physical slot by only identifying whether the pressure plate handle is in the engaged or disengaged state. Before the switchgear is put into operation, a reference association model for the switchgear's pressure plates is generated based on the switchgear's secondary circuit configuration table, protection setting sheet, switchgear layout diagram, and on-site verification records. For example, the physical slot of the pressure plate in row 5, column 8 is recorded as R05C08, the pressure plate identification code is recorded as 110L01CK0508, the nameplate text is recorded as line protection output, and the normal operating condition should be in the "engaged" state. The physical slot of the pressure plate in row 5, column 9 is recorded as R05C09, the pressure plate identification code is recorded as 110L01CK0509, the nameplate text is recorded as reclosing engaged, and the normal operating condition should be in the "disengaged" state. The pressure plate reference association model also records the geometric center coordinates, nameplate area, handle geometric envelope, state angle range, cooperative group or mutual exclusion group to which the pressure plate belongs, and applicable operating conditions of the above pressure plates. This ensures that each pressure plate is simultaneously subject to the association constraints of the identification code, physical slot, nameplate text, required state, and the combined constraints of the pressure plate engagement and disengagement states in the same switchgear. After the inspection robot moves to workstation W03 according to the inspection plan, it reads the current cabinet code P110-L01 and calls the baseline package for the cabinet to be in an active state. After confirming the workstation, the inspection robot collects multiple frames of cabinet panel images, and simultaneously obtains the collection time, shooting posture, supplementary lighting level, primary equipment operating conditions, and interval maintenance status. If the monitoring backend returns that the line interval is currently in normal operating condition, and the operating condition snapshot time meets the synchronization requirements with the image collection time, then the operating condition is used as the basis for the status judgment of this identification. If there is a delay in the operating condition snapshot, the operating condition snapshot is requested again. If a valid operating condition snapshot still cannot be obtained, the operating condition logic judgment is suspended, and the object identity verification, activation / deactivation status verification, and activation / deactivation status combination constraint review are continued. When performing structural positioning on the acquired cabinet panel images, structural anchor points such as the cabinet's outer frame, top nameplate area, and pressure plate matrix boundary are extracted from the images. These anchor points are then paired with millimeter coordinates from the reference package to obtain the geometric transformation relationship between the current image coordinates and the cabinet panel coordinates. If the reprojection residual meets the slot positioning requirements, the inspection image is reprojected onto the cabinet panel coordinate system, and each pressure plate working area is divided according to reference slots such as R05C08 and R05C09. If a working area is incomplete due to occlusion or severe reflection, the misalignment result is not directly output for that working area; instead, a backup image is called to re-divide the area. If the backup image still cannot meet the recognition requirements, the corresponding pressure plate is marked as pending verification. Within the R05C08 working area, object identity verification is first performed. The identification processing end extracts the geometric center, handle outline, and nameplate text of the pressure plate within this working area and compares them with the baseline record corresponding to R05C08. If the handle of the pressure plate at the R05C08 position in the field image is in the engaged state, but the identified nameplate text is close to the reclosing engagement state, and its shape characteristics and geometric center position are closer to the baseline record of the adjacent R05C09 pressure plate, then the object identity of the pressure plate is judged as suspected mismatch or mismatch, rather than directly judging it as compliant based on the slot being in the engaged state. Correspondingly, within the R05C09 working area, if the identified nameplate text is close to the line protection outlet, and the geometric position deviation relationship forms an adjacent exchange relationship with R05C08, then verification evidence of adjacent slot exchange is generated. Thus, even if both pressure plates are currently in an identifiable state, the hidden position misalignment can still be identified based on the correspondence between the object identity and the actual slot. After the object's identity is verified or confirmed, the activation / deactivation status is checked. For each pressure plate, the main line segment of the handle is extracted, the state angle of the main line segment relative to the horizontal reference of the cabinet is calculated, and the state mapping relationship registered in the reference package is used to determine whether the pressure plate is in the activated state, deactivated state, or mechanical intermediate state. For multiple frames of the same pressure plate, inter-frame consistency fusion is performed to reduce the risk of misjudgment caused by single-frame reflection, jitter, or short-term obstruction. If the measured state of the line protection output pressure plate corresponding to R05C08 is deactivated, but the state table of the operating conditions under normal operation records its state as activated, a state misalignment record is generated. If the pressure plate is in an allowable transition state and has not exceeded the transition time window, a transition state record is generated, and no confirmation alarm is output. After completing the object identity verification and on / off status verification of a single pressure plate, the combined constraints of on / off status on the same screen are reviewed. Taking a coordinated group consisting of a line protection inlet pressure plate and a line protection outlet pressure plate as an example, if the inlet pressure plate is in the on state while the outlet pressure plate is in the off state, the coordinated group rules do not meet the synchronous on-state requirements, and a coordinated group violation record is generated. Taking a protection mode selection mutual exclusion group as an example, if two mode selection pressure plates in the same functional group are simultaneously in the on state, a mutual exclusion group violation record is generated. For quantity constraint groups, the number of pressure plates in the on state within the group is counted and compared with the allowable range in the benchmark package. If the number exceeds or falls below the allowable range, a quantity constraint violation record is generated. When performing operational condition correlation verification, each pressure plate in the current cabinet is sorted according to its identity code to form a measured state vector, which is then compared bit by bit with the baseline state vector under normal operating conditions. If the measured state of a pressure plate is inconsistent with its expected state, the interval maintenance status and work order permission record are read again. If a valid work order exists, and the permission content covers the temporary state of the pressure plate or the pressure plate group associated with the pressure plate, and the current time is within the permission time range and transition time window, the inconsistency is marked as a permission transition. If there is no valid permission, or the permission content does not correspond to the pressure plate identity, it is determined to be an operational condition logical misalignment. Through the above processing, logical misalignments can be identified where the pressure plate object, slot, and local state all pass the single-item verification, but are inconsistent with the current primary equipment operating mode or maintenance permission status. Finally, misalignment types and alarm records are generated based on various verification results. If R05C08 and R05C09 form an adjacent exchange relationship, the main alarm type is determined to be position misalignment, and the identifier misalignment or status misalignment is written as an additional type into the same alarm record. If only the activation / deactivation status is inconsistent with the required status, it is determined to be status misalignment. If the status verification of a single pressure plate passes but the combination constraint rules of the activation / deactivation status of pressure plates on the same screen are not met, it is determined to be combination constraint misalignment. If the pressure plate status is inconsistent with the operating condition requirements or maintenance permission, it is determined to be operating condition logic misalignment. The alarm log should include at least the following fields: pressure plate identification code, cabinet code, slot code, misalignment type, evidence image index, geometric deviation, text similarity, status angle, inter-frame consistency ratio, trigger rule number, current working condition, and review status, and should be written back to the event log of the benchmark association model. For cases where multiple pressure plates in the same row or column simultaneously exhibit geometric deviation in the same direction, priority should be given to determining whether it is a suspected overall offset of the benchmark package. If it is suspected of being a suspected overall offset of the benchmark package, a benchmark review work order should be generated to avoid misjudging expired benchmarks as batch pressure plate misalignment. Through the above operation process, this method completes the following in a single inspection task: reference model invocation, image acquisition, coordinate reconstruction, work area segmentation, object identity verification, engagement / disengagement status verification, engagement / disengagement status combination constraint review, working condition association review, and alarm output. This ensures that the pressure plate identification result not only reflects whether the pressure plate is in an engaged or disengaged state, but also whether the state corresponds to the correct pressure plate, the correct slot, and the correct operating condition. It is suitable for identifying complex misalignment events such as misinstallation of adjacent slots, mismatched nameplates, half-engaged / half-disengaged status, violations of engagement / disengagement status combination constraints, and inconsistent operating conditions.

[0026] It should be noted that this invention can be deployed on the device itself to realize embedded applications, or it can run on a PC or other terminal with a user interface, thereby meeting various hardware environments and usage requirements.

[0027] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination. When implemented in software, the above embodiments can be implemented in whole or in part by a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions of the embodiments of this application are implemented in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted wirelessly or wiredly from one website, computer, server, or data center to another website, computer, server, or data center. Wired methods include optical fiber, twisted pair, coaxial cable, etc. Wireless methods include infrared, microwave, etc. Available media include any available media that can be accessed by a computer or data storage devices such as servers and data centers that contain one or more sets of available media. Available media can be magnetic media (floppy disks, hard disks, magnetic tapes), optical media (DVDs), or semiconductor media. Semiconductor media can be solid-state drives.

[0028] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0029] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for identifying misaligned pressure plates in a substation, characterized in that, include: S1. Establish a pressure plate reference association model based on the cabinet, and bind the pressure plate identity code, physical slot, nameplate text, working condition status, and pressure plate deployment / retraction status combination constraint rules with the cabinet panel coordinate system. The pressure plate deployment / retraction status combination constraint rules include mutual exclusion constraint rules, collaborative constraint rules, and deployment quantity constraint rules. S2. Obtain the inspection image of the cabinet panel, the acquisition station, the acquisition time, the operating conditions of the primary equipment and the interval maintenance status, and call the pressure plate reference association model corresponding to the acquisition station; S3. Extract the structural anchor points of the cabinet from the inspection images of the cabinet panel, reconstruct the coordinate system of the cabinet panel based on the structural anchor points, and divide each pressure plate working area according to the reference slot. S4. Perform pressure plate object identity verification and pressure plate loading / unloading status verification in each pressure plate working area to form a verification relationship between pressure plate identity, actual slot position, actual measured status and expected status. S5. Based on the constraint rules of the combination of the on / off states of the pressure plate on the same screen and the correlation between the working conditions of the primary and secondary equipment, the verification relationship is reviewed, and the pressure plate misalignment type and alarm record are generated.

2. The method for identifying misalignment of substation pressure plates according to claim 1, characterized in that, S1 includes: When establishing the pressure plate reference association model, the panel coordinate system is established based on the fixed structural points of the cabinet; Based on the design drawings, secondary circuit configuration table, protection setting sheet and on-site verification record, determine the correspondence between pressure plate slot, identification code, nameplate area, geometric center, pixel reference point, handle envelope, state angle range, working condition status and pressure plate on / off status combination constraint rules, and encapsulate it into a cabinet-level reference package with version number.

3. The method for identifying misalignment of substation pressure plates according to claim 1, characterized in that, S2 includes: When calling the pressure plate reference association model, the reference package index is queried according to the cabinet code, the workstation is confirmed, the shooting posture is adjusted and multi-frame images are acquired, and the working condition snapshot is obtained simultaneously and an acquisition message with time correspondence and spatial correspondence is formed. Verify the base package version and perform differential updates when versions are inconsistent; The acquired images are screened for exposure, sharpness, and occlusion quality. If the quality screening fails, re-acquisition or re-verification is performed, and a record to be verified is generated.

4. The substation pressure plate misalignment identification method according to claim 3, characterized in that, The baseline package version is validated, and differential updates are performed when versions are inconsistent, including: Before calling the local benchmark package, the consistency of the version number, effective version number, inspection task number and workstation code of the local cached benchmark package is verified. When the version number is inconsistent with the effective version number, the changed pressure plate record, working condition mapping entry and the combination constraint rules of the pressure plate activation and deactivation status on the same screen are obtained according to the field-level revision record and written to the local cache; When the workstation code and the inspection task are inconsistent, the base package index of the corresponding cabinet is rematched.

5. The method for identifying misalignment of substation pressure plates according to claim 1, characterized in that, S3 includes: When reconstructing coordinates and dividing the work area, grayscale, edge enhancement and closing operations are performed on qualified inspection images. The frame, nameplate, pressure plate matrix or replacement structure anchor points of the cabinet are extracted and cross-validated according to the anchor point position relationship in the reference package. The effective anchor point pixel coordinates are paired with the panel millimeter coordinates to establish a geometric transformation relationship. After outlier anchor point removal and reprojection residual judgment, the image is reprojected. The working area is divided according to the geometric center of the pressure plate, the handle envelope, and the nameplate area. The integrity of the divided working area is marked to generate a sub-record to be verified.

6. The method for identifying misalignment of substation pressure plates according to claim 5, characterized in that, Geometric transformation relationships are established by pairing effective anchor point pixel coordinates with panel millimeter coordinates. After outlier anchor point removal and reprojection residual determination, image reprojection is performed, including: When establishing geometric transformation relationships, the image pixel coordinates of the effective anchor points are registered with the panel millimeter coordinates in the reference package, and the geometric transformation matrix is ​​obtained by using planar homography transformation. Abnormal anchor points are identified based on the median of the reprojection residuals of each anchor point. After removing abnormal anchor points, the geometric transformation matrix is ​​recalculated. The reliability of the geometric transformation is marked based on the average reprojection residuals, and the inspection images of the cabinet panels are reprojected.

7. The method for identifying misalignment of substation pressure plates according to claim 1, characterized in that, S4 includes: When generating verification relationships, after the pressure plate working area is complete and the geometric quality meets the verification conditions, an object identity fusion score is formed based on the geometric center and slot code, nameplate text similarity, and handle and base shape characteristics. After the object identity is verified, the correspondence between the measured states is determined based on the handle state angle, multi-frame consistency, working condition status table and transition time window. When the object identity is incorrect, the state angle is in the boundary range, or the state is undeterminable, the corresponding verification identifier will be written into the verification record.

8. The method for identifying misalignment of substation pressure plates according to claim 7, characterized in that, When the object identity is incorrect, the state angle is in the boundary range, or the state is undeterminable, the corresponding verification identifier will be written into the verification record, including: When writing the review identifier, the object is marked as abnormal based on the object identity fusion score, geometric deviation, text similarity and shape anomaly type, and the normal status response process of the corresponding pressure plate is no longer executed; Based on the state angle boundary marker, inter-frame consistency, image quality status, and backup image verification results, the status is marked as potentially abnormal or the status is undeterminable, and the verification marker is associated with the corresponding pressure plate identification code, slot code, and verification time.

9. The method for identifying misalignment of substation pressure plates according to claim 1, characterized in that, S5 includes: When verifying the relationship, a combined constraint violation index is generated according to the mutual exclusion constraint rule, the collaborative constraint rule, and the input quantity constraint rule; The mutual exclusion constraint rule is used to verify that only one pressure plate in the associated pressure plate is in the input state. The coordination constraint rule is used to verify that the input and withdrawal states of the associated pressure plates are consistent. The input quantity constraint rule is used to verify that the number of associated pressure plates in the input state falls within a preset integer range.

10. A method for identifying misalignment of substation pressure plates according to claim 9, characterized in that, The measured state vector is compared bit by bit with the expected state vector under the working condition, and inconsistencies are filtered out by combining work order permission and transition time window, including: When performing working condition correlation verification, the measured status of each pressure plate is arranged according to the pressure plate identification code, and the status of each pressure plate under the current primary equipment operating condition is read to generate a working condition inconsistency list. For each pressure plate in the list of inconsistent operating conditions, verify the interval maintenance status, work order number, permit time, work order permit content and permit execution status, and mark the permit transition items or operating condition logic misalignment items according to the allowable transition time window.

Images