A method for false displacement identification and data validity determination in structural visual monitoring

Abstract

The present application belongs to the technical field of optical metrology and image processing, and particularly relates to a pseudo displacement identification and data validity determination method in structural visual monitoring, comprising the following steps: S1: acquiring a monitoring image sequence, a stable reference area and auxiliary state information; S2: extracting monitoring point displacement, image quality and multi-source state characteristics; S3: constructing a multi-dimensional consistency index; S4: judging whether the multi-dimensional consistency index meets a true structure deformation criterion; S5: generating a data validity level according to credibility and abnormal type, and controlling whether the data participates in safety early warning; S6: storing original images, pseudo displacement labels, disposal processes, final displacement and validity levels, and returning to continuous monitoring. The present scheme can identify pseudo displacement sources and execute differentiated disposal by jointly analyzing images, space, time, structure, wave bands and equipment state information, and finally outputs structure deformation data with a validity level.

Description

Technical Field

[0001] This invention belongs to the field of optical metrology and image processing technology, and particularly relates to a method for identifying false displacement and determining data validity in structural visual monitoring. Background Technology

[0002] Camera-based structural displacement, crack, rotation, and vibration measurement technologies have advantages such as being non-contact, having a large coverage area, and being easy to deploy remotely. They have been widely used in long-term health monitoring of bridges, buildings, slopes, tunnels, and large equipment. This type of technology involves setting natural textures, artificial targets, or actively emitting targets on the structural surface, and using algorithms such as digital image correlation, feature matching, or optical flow tracking to analyze the pixel motion of the target area in the image sequence. Then, combined with camera intrinsic parameters and field scale factors, the pixel motion is converted into structural deformation data in physical space. With the development of high-resolution cameras, high-speed acquisition equipment, and embedded processing platforms, visual measurement has become one of the important technical means in the field of structural health monitoring, and has achieved a large number of application results in laboratory verification and long-term field monitoring.

[0003] However, the actual engineering monitoring environment is far more complex than laboratory conditions. During long-term continuous monitoring, the pixel motion of the target in the image sequence acquired by the camera does not all come from the actual deformation of the measured structure. Specifically, diurnal and seasonal changes in ambient light may cause fluctuations in the overall brightness and contrast of the image, affecting the stability of feature matching; atmospheric turbulence and temperature stratification may cause changes in refractive index along the line of sight, resulting in visual motion of the target in the image; rain, snow, dust, and insect attachment may cause partial lens occlusion or image blurring; wind loads or vibrations from nearby construction may cause high-frequency shaking of the camera bracket; loosening of camera mounting bolts or temperature deformation during long-term service may cause slow drift of the camera optical axis; power supply fluctuations or LED aging of the active light-emitting target may cause changes in the target brightness and shape, affecting the accuracy of target positioning; in addition, seasonal changes or local damage to the surface texture of the measured structure may also lead to feature loss and mismatch.

[0004] All of the above factors can produce pseudo-displacements that are highly similar to the actual structural displacements in terms of numerical magnitude and temporal characteristics. Once these pseudo-displacements are mixed into the monitoring data, they may be misjudged as dangerous structural deformations and trigger false alarms, or they may cover up the true structural anomalies and cause missed alarms, seriously threatening the reliability of structural safety assessments.

[0005] To address the aforementioned errors and interference issues in visual measurements, various processing solutions have been proposed in the existing technology: The first type of solution is the motion compensation method based on a stable reference point. This involves setting a stationary reference object or selecting a static area in the image background within the camera's field of view. The overall rigid motion of the camera or the common motion component caused by environmental vibration is estimated by calculating the pixel motion of the reference point. This common component is then subtracted from the displacement of the target measurement point. Some systems also use an auxiliary inertial sensor or a second fixed camera to independently estimate the camera's self-motion to enhance the reliability of the compensation. This solution is suitable for camera rigid body jitter, but when the camera mounting interface is loose, the reference area itself changes, or local atmospheric disturbances exist, simply subtracting the common motion may cause incorrect compensation.

[0006] The second type of approach is a measurement reliability judgment method based on image quality evaluation. This method calculates quality factors such as correlation peak coefficient, matching residual, image sharpness index, number of feature points and distribution uniformity to score the measurement results of each frame of the image. When the score is lower than a set threshold, the data at the corresponding time is marked as unreliable or removed. This type of method can detect low-quality images or obvious mismatches, but it usually cannot explain whether the anomaly comes from lighting, defocus, occlusion, target anomaly or real structural mutation.

[0007] The third type of approach uses a multi-sensor cross-validation method to enhance reliability. This method compares visual displacement with measurement data from contact sensors such as accelerometers, inclinometers, strain gauges, or displacement gauges deployed at the same location. When the consistency of the outputs from different sensors exceeds the allowable range, the visual data is marked or corrected. This type of approach can improve monitoring reliability, but it usually does not establish a unified judgment framework between global-local motion, spatial continuity, structural constraints, dual-band consistency, and equipment status.

[0008] However, the existing solutions mentioned above still have the following shortcomings in practical applications: 1. Single processing target and lack of unified identification framework: Existing methods are usually designed only for a single source of error. Reference point compensation mainly deals with the overall camera motion, image quality evaluation mainly finds blur and mismatch, and multi-sensor cross-validation mainly finds inconsistencies between sensors. There is a lack of a technical framework for unified identification and differentiation of multiple types of pseudo-displacement sources. At the same time, existing dual-band or multi-channel systems usually only perform image fusion and do not use the consistency of positioning results from different bands to judge target anomalies and environmental interference.

[0009] 2. Insufficient criteria information, prone to misjudgment and lack of physical interpretability: Judging the credibility of data based solely on a single indicator such as correlation coefficient, image clarity or outlier threshold may, on the one hand, misjudge real structural mutations as invalid data and delete them, and on the other hand, may miss systematic pseudo-displacements with high correlation coefficients. More importantly, the above methods can only output a binary judgment of valid or invalid, credible or unreliable, and cannot explain whether the abnormal data comes from illumination, defocus, occlusion, target anomalies or real structural mutations.

[0010] 3. The handling methods are simplistic, lacking differentiated strategies and tiered output: Simply using reference points to deduct common motion cannot handle combined situations such as reference area failure, loose installation, local atmospheric disturbances, and simultaneous movement of the structure and camera. Existing algorithms generally do not utilize the spatial continuity of adjacent measurement points, the allowable deformation direction of the structure, and mechanical boundary conditions, making it difficult to distinguish between single-point mismatches and real local damage. Abnormal data handling methods are simplistic, mostly involving only deletion or filtering, lacking mechanisms to perform compensation, weight reduction, replacement, resampling, or invalidation for different sources of pseudo-displacement. Monitoring results generally only output displacement values, lacking clear indicators of reliability, pseudo-displacement labels, data validity levels, and whether data is allowed to participate in safety early warning systems.

[0011] To address this, this application proposes a method for identifying false displacements and determining data validity in structural visual monitoring. By jointly analyzing image, spatial, temporal, structural, band, and equipment status information, the method identifies the source of false displacements and selects appropriate handling strategies based on the source. This avoids misjudging false displacements as dangerous structural deformations and also avoids simply deleting real sudden deformations as outliers. Summary of the Invention

[0012] In view of this, the present invention aims to provide a method for identifying false displacements and determining data validity in structural visual monitoring, so as to solve the problems of unidentifiable sources of false displacements, insufficient physical interpretability of anomaly criteria, and single handling strategies and output formats in existing visual measurement data processing methods.

[0013] To achieve the above objectives, the technical solution created by this invention is implemented as follows: A method for identifying false displacements and determining data validity in structural visual monitoring includes the following steps: S1: Acquire monitoring image sequences, stable reference areas, and auxiliary status information; auxiliary status information includes equipment status information, environmental status information, image status information, and structural constraint information; S2: Extract displacement of monitoring points, image quality, and multi-source state features; S3: Constructing multidimensional consistency indicators; S4: The pseudo-displacement recognition processor determines whether the multidimensional consistency index meets the criteria for real structural deformation; the pseudo-displacement recognition processor sequentially performs image quality and feature reliability evaluation, global-local motion consistency analysis, spatial-temporal-structural constraint analysis, pseudo-displacement source identification, data validity grading and handling; and outputs real structural deformation, pseudo-displacement type, measurement credibility, data validity level, handling instructions, early warning and storage results; If the conditions are met, output the actual structural deformation result; If the conditions are not met, the source of the false displacement and the range of abnormal measurement points are identified, and relevant handling strategies are implemented based on the identified source of the false displacement; the displacement results and measurement reliability are recalculated until the true structural deformation criteria are met. S5: Generate data validity levels based on credibility and anomaly type, and control whether data participates in security alerts; S6: Store the original image, pseudo-displacement labels, processing procedures, final displacement, and validity level, and return it for continuous monitoring.

[0014] Furthermore, the monitoring image sequence in S1 includes images of the structure under test acquired at preset time intervals. Each image of the structure under test contains at least one monitoring point and at least one stable reference region. The monitoring image sequence includes images acquired at different image scales, under different exposure conditions, at different wavelengths, or through different imaging channels. The stable reference region is a stable reference target, a verified stable region in the structure, a camera base reference target, or a stable background region acquired by an auxiliary camera.

[0015] Furthermore, the equipment status information in S1 includes camera installation status and equipment operation status; environmental status information includes ambient temperature, light intensity, rainfall, snowfall, wind speed, and thermal disturbance information along the line of sight path; image status information includes image clarity, exposure status, occlusion status, focus status, and number of effective features; and structural constraint information includes topological relationships of measurement points, allowable deformation directions, boundary conditions, rigid body regions, and beam-slab continuity information.

[0016] Furthermore, the extraction of monitoring point displacement, image quality, and multi-source state features in S2 specifically includes: Perform image registration, template matching, digital image correlation, feature point tracking or target decoding on the monitored image sequence to obtain the displacement vector of each monitoring point and obtain the reference motion vector for the stable reference region. Image quality and positioning reliability characteristics are calculated for each monitoring point. These characteristics include sharpness, exposure saturation ratio, local contrast, number of effective features, correlation peak, peak width, peak sidelobe ratio, subpixel fitting residual, template deformation degree, and occlusion ratio. When the system has multiple scales, multiple algorithms, or dual-band channels, displacement estimates of different scales, different algorithms, and different bands are obtained respectively, and the differences, orientation angles, and positioning variances between the estimation results are calculated.

[0017] Furthermore, the multidimensional consistency metrics in S3 include at least: Global-local motion consistency index: Identifies overall camera shake by comparing the motion direction, amplitude, and temporal correlation of multiple measurement points, a stable reference area, and a single target area; Multi-scale consistency index or multi-algorithm consistency index: Identify texture degradation and mismatches by comparing the outputs of different image scales, different template sizes, or different displacement algorithms; Measurement point spatial continuity index: Calculate displacement gradient, curvature or rigid body consistency based on the topological relationship of neighboring measurement points, and identify isolated jumps and local occlusions; Structural constraint consistency index: Determines whether the displacement conforms to the allowable direction of the structure, boundary conditions, rigid body region constraints, beam deflection tendency, or slope principal slip direction; Temporal continuity index: By analyzing the trend, period, abrupt change, duration and frequency characteristics of displacement, it distinguishes between slow deformation, structural vibration, camera loosening step and random mismatch; Band or channel consistency index: Identify target failure, illumination interference, and single-channel anomalies by comparing visible light, near-infrared, or wide and narrow field-of-view measurement results; Equipment and image status consistency index: Identify equipment status anomalies by comparing the relationship between visual displacement and camera posture, mount strain, preload status, focus status, and environmental changes.

[0018] Furthermore, in S4, measurement reliability is generated based on pseudo-displacement type, number of remaining valid measurement points, various consistency indices, compensation residuals, and historical stability, and the data is divided into the following validity levels: Level A: Directly participates in structural safety early warning through all key consistency constraints; Level B: Slight interference exists or reliable compensation has been completed. It is used for trend analysis and participates in early warning when additional conditions are met. Level C: There is a clear anomaly, requiring re-data collection and confirmation at backup measurement points. High-level warnings must not be triggered alone. Class D: Reference datum failure, loose installation, severe obfuscation, or unrecoverable data; the data is invalid and must not be included in structural state calculations. The effectiveness level controls subsequent security alerts, trend analysis, data storage, and maintenance alarm logic.

[0019] Furthermore, in S4, the measurement reliability is obtained based on a combination of image quality, geometric consistency, structural consistency, temporal continuity, band consistency, and equipment status. At the same time, rejection conditions must be set for critical anomalies that cannot be compensated. When the installation is loose, the reference benchmark fails, or the target encoding is abnormal, the data is directly judged as the lowest level, regardless of the scores of other indicators.

[0020] Furthermore, in S4, when multiple measuring points and reference areas move synchronously in the same direction and the camera attitude changes synchronously, it is determined to be camera shake; when visual displacement shows a step and is accompanied by preload, strain, or attitude abnormalities, it is determined to be loose installation; when multiple local areas show high-frequency random non-rigid body shake, periodic fluctuations in sharpness, and stable equipment status, it is determined to be atmospheric disturbance; when brightness changes significantly, geometric edge positions lack consistent changes, and dual-band results are inconsistent, it is determined to be illumination interference; when sharpness decreases, correlation peaks broaden, and positioning variance increases, it is determined to be out of focus or blurry; when the number of effective features decreases or the target outline is missing, it is determined to be occlusion; when a single measuring point jumps abruptly and violates spatial and structural constraints, it is determined to be mismatch; when the active target encoding is incomplete or the dual-band point positions are inconsistent, it is determined to be target abnormality.

[0021] Furthermore, the handling strategies in S4 include compensation, downweighting, replacement, resampling, and invalidation. Among them, camera shake is a compensable category, which uses reference area and attitude information for global motion compensation; atmospheric disturbance is a downweightable category, which uses spatiotemporal filtering, multi-frame fusion, or reducing the weight of the current data; illumination interference is a resampling or feature replacement category, which adjusts the exposure or switches to a stable feature; occlusion and mismatch are replaceable categories, which switch to a backup measurement point or remove abnormal points; and defocus, reference benchmark failure, target encoding abnormality, and loose installation are non-compensable categories, which suspend output and mark the data as invalid.

[0022] Furthermore, the pseudo-displacement source identification in S4 is implemented using physical rule-based decision trees, probabilistic graphical models, Bayesian classifiers, support vector machines, or calibrated neural networks; when multiple abnormal conditions are met simultaneously, a composite pseudo-displacement label is output and the confidence level of each type is retained.

[0023] Compared with the prior art, the present invention can achieve the following beneficial effects: 1. A unified identification framework based on multi-dimensional consistency constraints is constructed to achieve physical classification and identification of pseudo-displacement sources: This invention jointly constructs a three-dimensional consistency discrimination framework by combining information from seven dimensions: global-local motion consistency, multi-scale measurement consistency, measurement point spatial continuity, structural deformation direction consistency, time series continuity, dual-band measurement consistency, and camera installation state stability. This overcomes the shortcomings of existing methods that only target a single error source or only perform binary anomaly judgment. Based on this framework, this invention can set camera shake, loose installation, atmospheric disturbance, lighting interference, defocus blur, occlusion, mismatch, and target anomaly as categories with different physical criteria for separate identification.

[0024] 2. Introducing spatial and structural constraints to effectively distinguish between real local damage and single-point mismatch: This invention makes full use of spatial and structural constraint information such as the spatial continuity of displacement of adjacent measuring points, the allowable deformation direction of the structure, and mechanical boundary conditions. When there is a significant deviation between the displacement of a certain measuring point and that of adjacent measuring points, the structural stress mode and allowable deformation direction can be combined to determine whether the deviation conforms to the laws of structural mechanical behavior, thereby accurately distinguishing between single-point feature mismatch and real local structural damage.

[0025] 3. Employing source-related differentiated processing strategies to achieve refined data quality control: This invention applies compensation, weight reduction, replacement, resampling, or invalidation strategies to different pseudo-displacement sources, avoiding erroneous correction of uncompensable data and complete deletion of recoverable data. After processing, the displacement and reliability are recalculated.

[0026] 4. Establish a tiered validity output and mandatory veto mechanism to ensure interpretable and traceable results: This invention generates four levels of data validity: A, B, C, and D. These levels control the data's usage permissions in safety warnings, trend analysis, and maintenance alarms. Mandatory veto conditions are set for critical anomalies such as loose installation or reference benchmark failure. All pseudo-displacement tags, handling paths, and validity levels are stored synchronously, and the pseudo-displacement type, abnormal measurement point range, reliability, and handling process can be output. Attached Figure Description

[0027] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments and descriptions of the invention are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings: Figure 1 This is an overall flowchart of the pseudo-displacement identification and data validity determination method in an embodiment of the present invention; Figure 2 This is a framework diagram of pseudo-displacement identification and data validity determination in an embodiment of the present invention; Figure 3 This is a schematic diagram illustrating pseudo-displacement classification in an embodiment of the present invention; Figure 4 This is a schematic diagram illustrating the execution of different handling strategies based on different pseudo-displacement sources and the generation of data validity levels in an embodiment of the present invention. Detailed Implementation

[0028] To make the purpose, technical solution, and advantages of this invention clearer, the following description is provided in conjunction with the appendix. Figure 1-4 The present invention will be further described in detail below with reference to specific embodiments. It should be understood that the specific embodiments described herein are only for explaining the present invention and do not constitute a limitation thereof.

[0029] This invention provides a method for identifying false displacements and determining data validity in structural visual monitoring, comprising the following steps: S1: Acquire monitoring image sequences, stable reference areas, and auxiliary status information; auxiliary status information includes equipment status information, environmental status information, image status information, and structural constraint information.

[0030] The monitoring image sequence includes images of the structure under test acquired at preset time intervals. Each image contains at least one monitoring point and at least one stable reference region. In this embodiment, multiple monitoring points are included. The monitoring image sequence may include images acquired at different image scales, under different exposure conditions, at different wavelengths, or through different imaging channels. The stable reference region may be a stable reference target set on an independent stable foundation, a stable region in the structure that has been verified over a long period of time, a camera base reference target, or a stable background region acquired by an auxiliary camera.

[0031] Equipment status information includes camera installation status and equipment operation status; environmental status information includes ambient temperature, light intensity, rainfall, snowfall, wind speed, and thermal disturbance information along the line of sight path; image status information includes image clarity, exposure status, occlusion status, focus status, and number of effective features; structural constraint information includes topological relationships of measurement points, allowable deformation directions, boundary conditions, rigid body regions, and beam-slab continuity information.

[0032] S2: Extract the displacement, image quality, and multi-source state features of the monitoring points, specifically including: Image registration, template matching, digital image correlation, feature point tracking or target decoding are performed on the monitored image sequence to obtain the displacement vector of each monitoring point, and the reference motion vector is obtained for the stable reference region.

[0033] Image quality and positioning reliability characteristics are calculated for each monitoring point. These characteristics include sharpness, exposure saturation ratio, local contrast, number of effective features, correlation peak, peak width, peak sidelobe ratio, subpixel fitting residual, template deformation degree, and occlusion ratio.

[0034] When the system has multiple scales, multiple algorithms, or dual-band channels, displacement estimates of different scales, different algorithms, and different bands are obtained respectively, and the differences, orientation angles, and positioning variances between the estimation results are calculated.

[0035] S3: Construct a multidimensional consistency metric, which should include at least the following: Global-local motion consistency index: Identifies overall camera shake by comparing the motion direction, amplitude, and temporal correlation of multiple measurement points, a stable reference area, and a single target area; Multi-scale measurement consistency index or multi-algorithm consistency index: Identify texture degradation and mismatches by comparing the outputs of different image scales, different template sizes, or different displacement algorithms; Measurement point spatial continuity index: Calculate displacement gradient, curvature or rigid body consistency based on the topological relationship of neighboring measurement points, and identify isolated jumps and local occlusions; Structural constraint consistency index or structural allowable direction consistency index: to determine whether the displacement conforms to the structural allowable direction, boundary conditions, rigid body region constraints, beam deflection trend or slope principal slip direction; Time series continuity index: By analyzing the trend, period, abrupt change, duration and frequency characteristics of displacement, it distinguishes between slow deformation, structural vibration, camera loosening step and random mismatch; Band or channel consistency index: Identify target failure, illumination interference, and single-channel anomalies by comparing visible light, near-infrared, or wide and narrow field-of-view measurement results; Equipment and image status consistency index: Identify equipment status anomalies by comparing the relationship between visual displacement and camera posture, mount strain, preload status, focus status, and environmental changes.

[0036] The above indicators are combined into a multidimensional consistency feature vector, and each indicator can be represented by normalized error, correlation coefficient, probability value, binary label or confidence interval.

[0037] S4: The pseudo-displacement recognition processor determines whether the multidimensional consistency index meets the true structural deformation criterion. The true structural deformation criterion is composed of image quality criterion, global-local motion criterion, measurement point spatial continuity criterion, structural constraint criterion, temporal continuity criterion, band or channel consistency criterion, and equipment status criterion. When the displacement to be determined meets the preset key consistency constraints and there are no rejection conditions such as reference benchmark failure, camera installation looseness, or target coding abnormality, the displacement is determined to be a true structural deformation or an effective displacement that can be used for structural status analysis.

[0038] The pseudo-displacement recognition processor sequentially performs image quality and feature reliability evaluation, global-local motion consistency analysis, spatial-temporal-structural constraint analysis, pseudo-displacement source identification, and data validity grading and processing; and outputs the real structural deformation, pseudo-displacement type, measurement credibility, data validity level, processing instructions, early warning, and storage results.

[0039] If the conditions are met, the actual structural deformation result will be output.

[0040] If the conditions are not met, the source of the false displacement and the range of abnormal measurement points are identified, and relevant handling strategies are implemented based on the identified source of the false displacement. The displacement results and measurement reliability are then recalculated until the true structural deformation criteria are met.

[0041] Measurement reliability is generated based on pseudo-displacement type, number of remaining valid measurement points, various consistency indices, compensation residuals, and historical stability, and the data is classified into the following validity levels: Level A: Directly participates in structural safety early warning through all key consistency constraints; Level B: Slight interference exists or reliable compensation has been completed. It is used for trend analysis and participates in early warning when additional conditions are met. Level C: There is a clear anomaly, requiring re-data collection and confirmation at backup measurement points. High-level warnings must not be triggered alone. Class D: Reference datum failure, loose installation, severe obfuscation, or unrecoverable data; the data is invalid and must not be included in structural state calculations. The effectiveness level controls subsequent security alerts, trend analysis, data storage, and maintenance alarm logic.

[0042] The measurement reliability is obtained by combining image quality, geometric consistency, structural consistency, temporal continuity, band consistency, and equipment status. At the same time, a rejection condition must be set for critical anomalies that cannot be compensated. When the installation is loose, the reference benchmark fails, or the target encoding is abnormal, the data is directly judged as the lowest level, regardless of the scores of other indicators. For different structures and scenarios, the threshold or classification model can be determined by normal operation data, controlled camera disturbance, illumination change, artificial occlusion, defocus, atmospheric thermal disturbance, and installation looseness test.

[0043] When multiple measuring points and reference areas move synchronously in the same direction and the camera attitude changes synchronously, it is judged as camera shake; when visual displacement shows a step and is accompanied by preload, strain or attitude abnormalities, it is judged as loose installation; when multiple local areas show high-frequency random non-rigid body shake, periodic fluctuations in sharpness and stable equipment status, it is judged as atmospheric disturbance; when brightness changes significantly, geometric edge positions lack consistent changes and dual-band results are inconsistent, it is judged as illumination interference; when sharpness decreases, correlation peaks broaden and positioning variance increases, it is judged as out of focus or blur; when the number of effective features decreases or the target outline is missing, it is judged as occlusion; when a single measuring point jumps abruptly and violates spatial and structural constraints, it is judged as mismatch; when the active target encoding is incomplete or the dual-band point positions are inconsistent, it is judged as target abnormality.

[0044] The handling strategies include compensation, weight reduction, replacement, resampling, and invalidation. Among them, camera shake is compensable and global motion compensation is performed using reference area and attitude information. Atmospheric disturbance is weighted and spatiotemporal filtering, multi-frame fusion, or reducing the weight of the current data is used. Illumination interference is resampling or feature replacement and exposure is adjusted or stable features are switched. Occlusion and mismatch are replaceable and backup measurement points are switched or abnormal points are removed, and mismatch is classified as single-point mismatch. Defocus, reference benchmark failure, target encoding abnormality, and loose installation are not directly compensable and output is paused and the data is marked as invalid. After processing, the displacement result and reliability are recalculated to avoid introducing new systematic errors during the compensation process.

[0045] The identification of pseudo-displacement sources is achieved using decision trees based on physical rules, probabilistic graphical models, Bayesian classifiers, support vector machines, or calibrated neural networks. When multiple abnormal conditions are met simultaneously, composite pseudo-displacement labels are output and confidence levels of each type are retained.

[0046] S5: Generate data validity levels based on credibility and anomaly type, and control whether data is included in security alerts.

[0047] S6: Store the original image, pseudo-displacement labels, processing procedures, final displacement, and validity level, and return it for continuous monitoring.

[0048] Figure 1 This is a flowchart illustrating the overall process for pseudo-displacement identification and data validity determination. Figure 2 This is a framework diagram for pseudo-displacement identification and data validity determination. Figure 3 This is a schematic diagram for pseudo-displacement classification. Figure 4 This diagram illustrates how different handling strategies are implemented based on different pseudo-displacement sources, and how data validity levels are generated.

[0049] Figure 4In the process, once the pseudo displacement identification processor completes the anomaly source determination, that is, after identifying the pseudo displacement type, it selects the corresponding handling strategy based on the identification result, generates handling instructions, determines the data validity level, outputs warnings, and stores the results.

[0050] Specifically, for real structural deformations that satisfy all consistency constraints, the displacement result is directly output and used for safety early warning, and the data validity level is judged as Grade A; for overall camera shake, global motion compensation is performed using stable reference areas and attitude information, and the displacement is recalculated and judged as Grade A or B; for slight atmospheric disturbances, spatiotemporal filtering, multi-frame fusion, or data weight reduction are used for processing, and the data is judged as Grade B; for changes in illumination, corrections are made by adjusting exposure, replacing stable features, or re-acquiring images, and the data is judged as Grade B or C based on the correction effect; for partial occlusion or feature mismatch, the displacement is recalculated after switching to backup measurement points or removing abnormal measurement points, and the data is judged as Grade C; for out-of-focus or severe blur, the current result is paused and data is re-acquired, and the data is judged as Grade C or D; for abnormal target encoding, the target is marked as invalid, and the data is judged as Grade D; for abnormalities that cannot be directly compensated, such as loose camera installation, the data is marked as invalid and a maintenance alarm is triggered, and the data is judged as Grade D.

[0051] Finally, the processor outputs the pseudo displacement type, measurement confidence level, handling instructions, data validity level, and warning and storage results in association. Data of different levels participate in security warnings, trend analysis and historical data storage according to preset permissions, so as to realize data validity management corresponding to different handling strategies for different pseudo displacement sources.

[0052] Among them, pseudo displacement type refers to the source of pseudo displacement or the actual deformation situation identified; measurement reliability refers to the degree of measurement reliability evaluated by the algorithm, which is affected by multiple factors; handling instructions refer to the data processing or handling measures taken for different sources; data validity level refers to the data availability classification, with A being the best and D being the worst; early warning and storage results refer to the early warning issuance and data storage based on the validity level.

[0053] In practical applications, cameras can be used to monitor the long-term displacement of multiple marker points on the facade of bridges or buildings. At the same time, a stable reference target is set within the field of view, and the camera tilt angle, mounting strain, ambient temperature and image sharpness are acquired. The processor uses digital image correlation to calculate the displacement of each measuring point, and repeats the calculation on both the original scale and the downsampled scale.

[0054] Meanwhile, the specific implementation methods in the above steps can be replaced or combined according to the actual monitoring scenario and hardware conditions. Those skilled in the art will understand that the following alternative methods can all be used in conjunction with the foregoing embodiments.

[0055] Displacement calculation is not limited to digital image correlation; it can also be achieved through template matching, optical flow, feature point tracking, deep learning matching, or active target decoding. The appropriate algorithm can be selected based on the surface texture features of the structure being measured and the required monitoring accuracy.

[0056] The selection of a stable reference region can be achieved by means of external reference targets, structurally stable regions, auxiliary cameras, internal optical references, or multi-level reference regions, so as to provide a reliable reference benchmark under different monitoring scenarios.

[0057] Structural constraints can be implemented using rigid body transformation, beam and plate bending curves, finite element modes, slope principal slip direction, or data-driven relationships, to be applicable to the mechanical behavior characteristics of different types of structures.

[0058] Band consistency can be achieved by using methods such as dual cameras, wide and narrow fields of view, multiple exposures, multiple focal lengths, or consistency verification using different algorithms.

[0059] Credibility can be calculated using methods such as weighted scoring, probability, confidence interval, fuzzy evaluation, or evidence theory to suit the needs of representing uncertainty in different scenarios.

[0060] Data validity levels can be divided into two, three, or more levels according to actual monitoring needs, with different levels corresponding to different data usage permissions.

[0061] The handling strategy can be implemented by means of local filtering, model prediction replacement, switching of backup measurement points, manual verification, etc., so as to flexibly select the handling method according to the specific source and severity of the pseudo displacement.

[0062] Equipment status information can include information such as focus status, lens temperature, power supply status, and communication status, to more comprehensively reflect the equipment's operating conditions.

[0063] The method in this application is not limited to structural displacement measurement, but can also be used for crack width, rotation angle, vibration, settlement and other visual quantitative measurement scenarios, adapting to the needs of different structural health monitoring tasks.

[0064] It should be understood that the various forms of processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this invention disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this invention can be achieved, and this is not limited herein.

[0065] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method for identifying false displacements and determining data validity in structural visual monitoring, characterized in that, Includes the following steps: S1: Acquire monitoring image sequences, stable reference areas, and auxiliary status information; auxiliary status information includes equipment status information, environmental status information, image status information, and structural constraint information; S2: Extract displacement of monitoring points, image quality, and multi-source state features; S3: Constructing multidimensional consistency indicators; the multidimensional consistency indicators in S3 should include at least: Global-local motion consistency index: Identifies overall camera shake by comparing the motion direction, amplitude, and temporal correlation of multiple measurement points, a stable reference area, and a single target area; Multi-scale consistency index or multi-algorithm consistency index: Identify texture degradation and mismatches by comparing the outputs of different image scales, different template sizes, or different displacement algorithms; Measurement point spatial continuity index: Calculate displacement gradient, curvature or rigid body consistency based on the topological relationship of neighboring measurement points, and identify isolated jumps and local occlusions; Structural constraint consistency index: Determines whether the displacement conforms to the allowable direction of the structure, boundary conditions, rigid body region constraints, beam deflection tendency, or slope principal slip direction; Temporal continuity index: By analyzing the trend, period, abrupt change, duration and frequency characteristics of displacement, it distinguishes between slow deformation, structural vibration, camera loosening step and random mismatch; Band or channel consistency index: Identify target failure, illumination interference, and single-channel anomalies by comparing visible light, near-infrared, or wide and narrow field-of-view measurement results; Equipment and image status consistency index: Identify equipment status anomalies by comparing the relationship between visual displacement and camera posture, mount strain, preload status, focus status, and environmental changes. S4: The pseudo-displacement recognition processor determines whether the multidimensional consistency index meets the criteria for real structural deformation; the pseudo-displacement recognition processor sequentially performs image quality and feature reliability evaluation, global-local motion consistency analysis, spatial-temporal-structural constraint analysis, pseudo-displacement source identification, data validity grading and handling; and outputs real structural deformation, pseudo-displacement type, measurement credibility, data validity level, handling instructions, early warning and storage results; If the conditions are met, output the actual structural deformation result; If the conditions are not met, the source of the false displacement and the range of abnormal measurement points are identified, and relevant handling strategies are implemented based on the identified source of the false displacement; the displacement results and measurement reliability are recalculated until the true structural deformation criteria are met. S5: Generate data validity levels based on credibility and anomaly type, and control whether data participates in security alerts; S6: Store the original image, pseudo-displacement labels, processing procedures, final displacement, and validity level, and return it for continuous monitoring.

2. The method for identifying false displacements and determining data validity in structural visual monitoring according to claim 1, characterized in that, The monitoring image sequence in S1 includes images of the structure under test acquired at preset time intervals. Each image of the structure under test contains at least one monitoring point and at least one stable reference region. The monitoring image sequence includes images acquired at different image scales, under different exposure conditions, at different wavelengths, or through different imaging channels. The stable reference region is a stable reference target, a verified stable region in the structure, a camera base reference target, or a stable background region acquired by an auxiliary camera.

3. The method for identifying false displacements and determining data validity in structural visual monitoring according to claim 1, characterized in that, The equipment status information in S1 includes camera installation status and equipment operation status; environmental status information includes ambient temperature, light intensity, rainfall, snowfall, wind speed, and thermal disturbance information along the line of sight path; image status information includes image clarity, exposure status, occlusion status, focus status, and number of effective features; structural constraint information includes topological relationships of measurement points, allowable deformation directions, boundary conditions, rigid body regions, and beam-slab continuity information.

4. The method for identifying false displacements and determining data validity in structural visual monitoring according to claim 1, characterized in that, The extraction of monitoring point displacement, image quality, and multi-source state features in S2 specifically includes: Perform image registration, template matching, digital image correlation, feature point tracking or target decoding on the monitored image sequence to obtain the displacement vector of each monitoring point and obtain the reference motion vector for the stable reference region. Image quality and positioning reliability characteristics are calculated for each monitoring point. These characteristics include sharpness, exposure saturation ratio, local contrast, number of effective features, correlation peak, peak width, peak sidelobe ratio, subpixel fitting residual, template deformation degree, and occlusion ratio. When the system has multiple scales, multiple algorithms, or dual-band channels, displacement estimates of different scales, different algorithms, and different bands are obtained respectively, and the differences, orientation angles, and positioning variances between the estimation results are calculated.

5. The method for identifying false displacements and determining data validity in structural visual monitoring according to claim 1, characterized in that, In S4, measurement reliability is generated based on pseudo-displacement type, number of remaining valid measurement points, various consistency indices, compensation residuals, and historical stability, and the data is classified into the following validity levels: Level A: Directly participates in structural safety early warning through all key consistency constraints; Level B: Slight interference exists or reliable compensation has been completed. It is used for trend analysis and participates in early warning when additional conditions are met. Level C: There is a clear anomaly, requiring re-data collection and confirmation at backup measurement points. High-level warnings must not be triggered alone. Class D: Reference datum failure, loose installation, severe obfuscation, or unrecoverable data; the data is invalid and must not be included in structural state calculations. The effectiveness level controls subsequent security alerts, trend analysis, data storage, and maintenance alarm logic.

6. The method for identifying false displacements and determining data validity in structural visual monitoring according to claim 1, characterized in that, In S4, the measurement reliability is obtained by combining image quality, geometric consistency, structural consistency, temporal continuity, band consistency, and equipment status. At the same time, a veto condition must be set for critical anomalies that cannot be compensated. When the installation is loose, the reference benchmark fails, or the target encoding is abnormal, the data is directly judged as the lowest level, regardless of the scores of other indicators.

7. The method for identifying false displacements and determining data validity in structural visual monitoring according to claim 1, characterized in that, In S4, when multiple measuring points and reference areas move synchronously in the same direction and the camera attitude changes synchronously, it is judged as camera shake; when visual displacement shows a step and is accompanied by preload, strain, or attitude abnormalities, it is judged as loose installation; when multiple local areas show high-frequency random non-rigid body shake, periodic fluctuations in sharpness, and stable equipment status, it is judged as atmospheric disturbance; when brightness changes significantly, geometric edge positions lack consistent changes, and dual-band results are inconsistent, it is judged as illumination interference; when sharpness decreases, correlation peaks broaden, and positioning variance increases, it is judged as out of focus or blur; when the number of effective features decreases or the target outline is missing, it is judged as occlusion; when a single measuring point jumps abruptly and violates spatial and structural constraints, it is judged as mismatch; when the active target encoding is incomplete or the dual-band point positions are inconsistent, it is judged as target abnormality.

8. The method for identifying false displacements and determining data validity in structural visual monitoring according to claim 1, characterized in that, The handling strategies in S4 include compensation, downweighting, replacement, resampling, and invalidation. Among them, camera shake is a compensable category, which uses reference area and attitude information for global motion compensation; atmospheric disturbance is a downweightable category, which uses spatiotemporal filtering, multi-frame fusion, or reducing the weight of the current data; illumination interference is a resampling or feature replacement category, which adjusts the exposure or switches to a stable feature; occlusion and mismatch are replaceable categories, which switch to a backup measurement point or remove abnormal points; and out-of-focus, reference benchmark failure, target encoding abnormality, and loose installation are non-compensable categories, which pause output and mark the data as invalid.

9. The method for identifying false displacements and determining data validity in structural visual monitoring according to claim 1, characterized in that, The pseudo-displacement source identification in S4 is implemented using physical rule-based decision trees, probabilistic graphical models, Bayesian classifiers, support vector machines, or calibrated neural networks; when multiple abnormal conditions are met simultaneously, a composite pseudo-displacement label is output and the confidence level of each type is retained.

Images