Deep vision perception method for lining structure deterioration features

Abstract

The application discloses a deep visual perception method for lining structure deterioration characteristics, relates to the technical field of artificial support structure, collects continuous visual images in the tunnel inspection process, extracts light spot moving track, brightness change amplitude and irradiation range parameters based on the continuous visual images, generates dynamic light spot interference characteristic information according to the light spot moving track, brightness change amplitude and irradiation range parameters; according to the dynamic light spot interference characteristic information, the brightness fluctuation of the continuous visual images is separated, the area range affected by the dynamic light spot in the continuous visual images is determined, and the texture change range corresponding to the area range affected by the dynamic light spot is obtained. Through dynamic light spot interference characteristic extraction, brightness fluctuation separation and continuous visual feature correlation analysis, the application reduces the influence of light interference on deterioration identification, and improves the lining structure disease perception accuracy and continuous tracking capability in a complex environment.

Description

Technical Field

[0001] This application relates to the field of artificial support structure technology, specifically to a depth visual perception method for the deterioration characteristics of lining structures. Background Technology

[0002] Lining structures are artificial support structures installed inside tunnels, underground engineering projects, slope caverns, hydraulic structures, and other engineering projects to support the surrounding rock, bear external loads, maintain structural stability, and improve the working environment. They are typically composed of materials such as concrete, reinforced concrete, shotcrete, and precast segments. Their main functions are to constrain surrounding rock deformation, share surrounding rock pressure, prevent soil and rock loosening and failure, and form a smooth, sealed inner wall for the engineering structure. The deterioration characteristics of lining structures mainly manifest as cracks, leaks, spalling, flaking, steel reinforcement corrosion, surface weathering, and deformation cracking caused by long-term exposure to groundwater erosion, freeze-thaw cycles, load changes, material aging, and construction defects.

[0003] Deep visual perception of lining structure deterioration characteristics refers to the automatic collection, analysis, and understanding of visual information on the lining surface using high-definition imaging equipment, 3D vision technology, deep learning algorithms, image recognition models, and intelligent detection systems. This enables refined identification of the location, shape, size, severity, and development trend of deteriorated areas. The specific visual perception process is as follows: By acquiring continuous images or 3D point cloud data of the lining surface using camera equipment, and employing techniques such as image enhancement, feature extraction, target detection, and semantic segmentation, we can identify crack width, length, distribution patterns, and abnormal features such as water seepage marks and concrete defects. Combined with spatial positioning and data analysis methods, we can establish a lining structure deterioration state model, thereby realizing the transformation from traditional manual visual inspection to digital, intelligent, and quantitative detection methods, and providing a reliable basis for tunnel and underground engineering structure safety assessment, disease early warning, and maintenance decisions.

[0004] In the process of deep visual perception of the deterioration characteristics of lining structures, visual acquisition equipment typically acquires image information of the lining surface through continuous shooting and uses image recognition algorithms to extract features from abnormal areas such as cracks, spalling, and leakage. However, in actual inspection environments, dynamic factors such as the movement of vehicle lights inside the tunnel, changes in the illumination angle of the inspection equipment, and changes in reflected light from water accumulation areas can cause constantly changing local light spot areas to appear on the lining surface. These dynamically migrating light spots alter the brightness distribution of the original texture of the lining surface, causing local areas in the visually acquired images to exhibit abnormal visual features resembling crack edges or material defects. This makes it difficult for the deep visual perception model to accurately distinguish between genuine deterioration features and false features created by changes in illumination. When moving light spots cover micro-cracks or shallow spalling areas, the high-brightness reflection from the light spots weakens the boundary features of the defects, resulting in incomplete defect texture information extracted by the visual model and a decrease in the accuracy of identifying minor deterioration areas. Meanwhile, the dynamic light spot causes the image features of the same area to change continuously during the continuous detection process, making it difficult for the model to establish a stable correlation between defect features, which further leads to hidden cracks and early erosion areas being incorrectly judged as normal.

[0005] The aforementioned problems can lead to significant deviations in the collection of information on the deterioration of the lining structure, resulting in the omission of high-risk defect areas and affecting the results of subsequent structural safety assessments. In severe cases, it may cause potential hazardous areas to go undetected, posing a risk to the safety of the project's operation. The information disclosed in the background section is only for enhancing the understanding of the background of this disclosure and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] The purpose of this application is to provide a depth visual perception method for the deterioration characteristics of lining structures, so as to solve the problems in the background art mentioned above.

[0007] To achieve the above objectives, this application provides the following technical solution: a depth visual perception method for the deterioration characteristics of lining structures, comprising the following steps: Continuous visual images are collected during tunnel inspection. Based on the continuous visual images, the movement trajectory of the light spot, the amplitude of brightness change, and the illumination range parameters are extracted. Dynamic light spot interference feature information is generated based on the movement trajectory of the light spot, the amplitude of brightness change, and the illumination range parameters. Based on the dynamic light spot interference feature information, brightness fluctuation separation is performed on continuous visual images to determine the range of areas affected by dynamic light spots in continuous visual images and obtain the texture change range corresponding to the range of areas affected by dynamic light spots. Based on the range of texture changes, the edge capture accuracy of the perceived area in the continuous visual image is adjusted. Feature extraction is performed on the lining structure area in the continuous visual image to obtain the crack contour continuity feature, the spalling texture difference feature and the abnormal pixel distribution feature. Based on the crack contour continuity feature, the spalling texture difference feature and the abnormal pixel distribution feature, the basis for identifying stable diseases is determined. Using stable disease identification criteria, temporal feature correlation analysis is performed on continuous visual images to construct feature correlation rules between continuous visual images. Based on the feature correlation rules, false anomaly information generated by the movement of dynamic light spots is filtered out, and continuous visual features corresponding to the real deterioration area of ​​the lining structure are retained. Based on continuous visual features, the defect localization parameters in the continuous visual image detection process are updated, and the deterioration perception information of the lining structure, which is resistant to light spot interference, is output.

[0008] Preferably, dynamic light spot interference feature information is generated based on the light spot movement trajectory, brightness change amplitude, and illumination range parameters. The specific steps are as follows: The brightness change region in the continuous visual image frame is continuously tracked. Based on the positional change relationship of the brightness change region at different acquisition times, the movement direction, movement distance and movement range of the brightness change region in the continuous visual image are determined, and the movement trajectory of the dynamic light spot in the continuous visual image is obtained. The visual area traversed by the dynamic light spot is determined based on the light spot's movement trajectory. Corresponding analysis is performed on the continuous visual image frames before and after the dynamic light spot passes through to obtain the brightness difference information of the same visual area at different acquisition times. The brightness change amplitude caused by the dynamic light spot is determined based on the brightness difference information. For the visual region corresponding to the light spot movement trajectory, the boundary changes of the dynamic light spot coverage area are obtained. The illumination range parameters are determined based on the position change range of the dynamic light spot coverage area at different acquisition times. Dynamic light spot interference feature information is generated by combining the light spot movement trajectory, brightness change amplitude and illumination range parameters.

[0009] Preferably, the brightness change amplitude is obtained by corresponding analysis of continuous visual image frames before and after the dynamic light spot passes through, to obtain the brightness difference information of the same visual area at different acquisition times, and to determine the degree of brightness increase, the degree of brightness decrease, and the duration of brightness change during the dynamic light spot's action based on the brightness difference information.

[0010] Preferably, the texture change range corresponding to the area affected by the dynamic light spot is obtained, and the specific steps are as follows: Based on the light spot movement trajectory, brightness change amplitude, and illumination range parameters in the dynamic light spot interference characteristic information, brightness fluctuation separation is performed on continuous visual images to determine the area affected by dynamic light spots. Based on the area affected by the dynamic light spot, a spatial range analysis is performed on the corresponding area to obtain the changes in the location boundary of the area covered by the dynamic light spot. Based on the changes in the location boundaries, texture change analysis is performed on the area affected by the dynamic light spot to obtain the corresponding texture change range.

[0011] Preferably, the spatial expansion range, spatial contraction range, and spatial movement range of the dynamic light spot coverage area in the continuous visual image are determined based on the illumination range parameters, and the area affected by the dynamic light spot is determined by combining the coverage range change relationship at different acquisition times.

[0012] Preferably, the criteria for identifying stable defects are determined based on the continuity characteristics of crack contours, the differences in peeling texture, and the distribution characteristics of abnormal pixels. The specific steps are as follows: The distribution state of texture change in the sensing area is determined based on the range of texture change. The edge capture accuracy is adjusted according to the location of texture change, the boundary of texture change area, the amplitude of texture change, and the duration of texture change, thus forming edge change description information. Based on the adjusted edge capture accuracy, feature extraction is performed on the lining structure area. Crack contour continuity features are obtained according to edge connection relationship, edge direction change relationship, boundary extension relationship and spatial continuity relationship. In addition, peeling texture difference features and abnormal pixel distribution features are obtained by combining crack contour continuity features. Based on the continuity characteristics of crack contours, the differences in peeling texture, and the distribution characteristics of abnormal pixels, feature correspondences are established. The positional correspondences, boundary correspondences, and change correspondences are described. The set of disease features with continuous relationships is determined as the basis for identifying stable diseases.

[0013] Preferably, the continuity features of the crack contour are obtained based on edge change description information, specifically including: Based on the edge change description information, the positional connection relationship, edge direction change relationship, boundary extension relationship and spatial continuity relationship between adjacent edges are obtained. Edges that are continuously connected and spatially continuous are classified into the same crack contour. Based on the edge continuity degree, edge direction consistency degree, edge spatial connection state and contour extension state corresponding to the crack contour, a comprehensive description is formed to form crack contour continuity features that include positional continuity relationship, direction continuity relationship, edge connection relationship and contour extension relationship.

[0014] Preferably, false anomaly information generated by the movement of dynamic light spots is filtered out according to feature association rules, while retaining continuous visual features corresponding to the true deterioration area of ​​the lining structure. The specific steps are as follows: The continuous features of crack contours, the differences in peeling texture, and the distribution features of abnormal pixels are obtained. The disease features corresponding to different acquisition times in continuous visual images are correlated to determine the positional preservation relationship, the boundary preservation relationship, the continuous change relationship, and the temporal continuity relationship. Based on position preservation, boundary preservation, and continuous change relationships, feature association rules are established between continuous visual images. Position correspondence, boundary correspondence, change correspondence, and temporal continuity are used as association constraints to determine the continuous visual features corresponding to the actual deterioration area. By using feature association rules to filter visual change regions in continuous visual images, visual change regions that conform to the feature association rules are retained as continuous visual features corresponding to the real deterioration regions. False anomaly information generated by the movement of dynamic light spots is filtered out, and the deterioration perception information of the lining structure that is resistant to light spot interference is output.

[0015] Preferably, based on the feature association rules between continuous visual images, positional correspondence, boundary correspondence, change correspondence, and temporal continuity are obtained, and feature association rule parameters are calculated based on the positional correspondence, boundary correspondence, change correspondence, and temporal continuity. When the visual change region meets the corresponding conditions of the feature association rule parameters, it is determined that the visual change region belongs to the continuous visual feature corresponding to the real deterioration region.

[0016] Preferably, the defect localization parameters in the continuous visual image detection process are updated based on continuous visual features, and the lining structure deterioration perception information that is resistant to light spot interference is output. The specific steps are as follows: The continuous visual features retained after dynamic spot interference removal are obtained, and the defect localization parameters are updated based on the continuous position information, continuous contour change information, continuous texture change information, and continuous distribution information of abnormal pixels. The updated defect location parameters are associated with continuous visual features, and the location range and change status of the actual deteriorated area are determined based on the location location parameters, range location parameters, area change parameters, and area coverage parameters. Based on the determined location and change status of the actual deterioration area, anti-spot interference deterioration perception information of the lining structure is generated, which includes spatial location, regional range and continuous change status.

[0017] The technical effects and advantages provided by this application in the above technical solution are as follows: This application distinguishes between areas affected by dynamic light spot interference and actual deterioration areas by extracting dynamic light spot interference features and combining them with a brightness fluctuation separation process. Then, by combining texture change range and stable defect identification criteria, continuous visual feature analysis is conducted. This reduces the interference of dynamic light spots on crack outlines, spalling textures, and abnormal pixel distributions, improves the stability of identifying actual deterioration features under complex lighting conditions, and enhances the consistency between depth visual perception results and the actual state of the lining structure. Furthermore, this application constructs feature association rules between continuous visual images and updates defect location parameters based on continuous visual features. It filters out false anomalies formed by the migration of dynamic light spots and outputs light spot-resistant lining structure deterioration perception information. This maintains the spatial correspondence and temporal continuity of actual defects during continuous detection, improves the continuous tracking capability of hidden cracks and early spalling areas, and provides more stable and reliable perception results for subsequent structural safety assessments and defect development analysis. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this application. For those skilled in the art, other drawings can be obtained based on these drawings.

[0019] Figure 1 This is a flowchart illustrating the overall method of this application. Detailed Implementation

[0020] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the description of this disclosure will be more complete and fully convey the concept of the exemplary embodiments to those skilled in the art.

[0021] This application provides, as follows: Figure 1 The depth-based visual perception method for identifying deterioration features of lining structures, as shown, includes the following steps: Step 1: Collect continuous visual images during tunnel inspection. Extract the light spot movement trajectory, brightness change amplitude, and illumination range parameters based on the continuous visual images. Generate dynamic light spot interference feature information based on the light spot movement trajectory, brightness change amplitude, and illumination range parameters. Dynamic light spot interference feature information is generated based on the light spot movement trajectory, brightness change amplitude, and illumination range parameters. The specific steps are as follows: During tunnel inspection, continuous visual images of the lining surface are acquired using visual acquisition equipment. Multiple continuous visual image frames are obtained in chronological order of acquisition time. The image information in the continuous visual image frames is then processed to ensure that the visual information of the continuous visual image frames acquired at different times corresponds to the visual information of the same detection area.

[0022] Continuous visual images include image brightness information, region location information, and image change information corresponding to each acquisition time. Among them, image brightness information is used to describe the brightness distribution of different regions in the continuous visual image, region location information is used to describe the spatial positional relationship of each visual region in the continuous visual image, and image change information is used to describe the visual changes of the same detection region at different acquisition times.

[0023] Continuous tracking of brightness change regions in continuous visual image frames is performed. Based on the positional relationship of brightness change regions at different acquisition times, the direction, distance, and range of movement of brightness change regions in continuous visual images are determined. The positional information of brightness change regions at multiple continuous acquisition times is correlated in chronological order to obtain the position migration process of dynamic light spots in continuous visual images, and the position migration process is used as the movement trajectory of light spots.

[0024] By acquiring the trajectory of light spot movement, dynamic changing areas in continuous visual images can be described based on spatial positional change characteristics. This distinguishes the visual characteristics of deteriorated areas that are fixed on the lining surface from the visual characteristics of areas affected by illumination that change position during the acquisition process, providing a spatial basis for subsequently determining the dynamic light spot interference characteristics.

[0025] Based on the trajectory of the light spot movement, brightness change analysis is performed on the region corresponding to the trajectory in continuous visual images to obtain the brightness change amplitude and illumination range parameters at different acquisition times during the dynamic light spot's action. The specific acquisition process is as follows: The visual region traversed by the dynamic light spot is determined based on its movement trajectory. Corresponding analysis of consecutive visual image frames before and after the dynamic light spot's passage is performed to obtain brightness difference information for the same visual region at different acquisition times. Based on this brightness difference information, the amplitude of brightness change caused by the dynamic light spot is determined. The brightness change amplitude describes the degree of brightness increase, decrease, and duration of brightness change in the visual region during the dynamic light spot's action.

[0026] For the visual region corresponding to the movement trajectory of the light spot, the boundary changes of the dynamic light spot coverage area are further obtained. The illumination range parameter is determined according to the position change range of the dynamic light spot coverage area at different acquisition times. The illumination range parameter is used to describe the size of the region affected by the dynamic light spot on the continuous visual image, the direction of region change, and the expansion or contraction of the region.

[0027] By correlating the movement trajectory of the light spot, the amplitude of brightness change, and the illumination range parameters, the spatial change process, brightness change degree, and affected area range of the dynamic light spot in continuous visual images can be uniformly described. This avoids mistaking real deterioration feature areas or normal texture change areas as the affected areas of the dynamic light spot when judging abnormal areas based solely on single brightness change information, thereby improving the accuracy of dynamic light spot interference information extraction.

[0028] Finally, dynamic light spot interference feature information is generated based on the light spot movement trajectory, brightness change amplitude, and illumination range parameters. This dynamic light spot interference feature information is then used to describe the dynamic illumination changes in continuous visual images. Specifically: The trajectory of the light spot movement is used as the basis for the spatial change of the dynamic light spot, the amplitude of the brightness change is used as the basis for the brightness change of the dynamic light spot, and the illumination range parameter is used as the basis for the change of the area affected by the dynamic light spot. The continuous visual image change information corresponding to different acquisition times is associated and recorded to generate dynamic light spot interference feature information that includes position change features, brightness change features, and area influence features.

[0029] Dynamic light spot interference characteristics include the movement path of the dynamic light spot in continuous visual images, the degree of brightness change in the corresponding area, and the coverage area. These characteristics describe the continuous changes in the dynamic light spot's state during the inspection process. Since dynamic light spots typically exhibit time-varying positional migration characteristics, while deteriorated areas of the lining structure usually have relatively stable spatial distribution characteristics, dynamic light spot interference characteristics can be used to identify areas affected by illumination changes in continuous visual images. This provides a basis for subsequent brightness fluctuation separation and anomaly feature screening in continuous visual images.

[0030] By employing the above methods, the deep visual perception process for the deterioration characteristics of lining structures not only focuses on abnormal texture changes in visual images, but also considers the temporal variation characteristics of dynamic light spots. This reduces the impact of false abnormal information caused by dynamic light spots on the deterioration characteristic identification process, and improves the perception reliability of hidden cracks and shallow damage areas under complex lighting inspection environments.

[0031] Step 2: Based on the dynamic light spot interference feature information, perform brightness fluctuation separation on the continuous visual image to determine the area affected by the dynamic light spot in the continuous visual image and obtain the texture change range corresponding to the area affected by the dynamic light spot. The specific steps to obtain the texture change range corresponding to the area affected by dynamic light spots are as follows: Based on the dynamic light spot interference characteristic information, brightness fluctuation separation is performed on continuous visual images. According to the light spot movement trajectory, brightness change amplitude, and illumination range parameters in the dynamic light spot interference characteristic information, temporal correlation processing is performed on the brightness change regions in the continuous visual images to determine the brightness change regions caused by the dynamic light spot, and to obtain the range of regions affected by the dynamic light spot in the continuous visual images. The specific acquisition process is as follows: The system reads the light spot movement trajectory information from the dynamic light spot interference feature information, maps the positional change process corresponding to the light spot movement trajectory to the corresponding region in the continuous visual image, and determines the visual regions with continuous migration relationships between different acquisition times based on the positional changes of brightness change regions in the continuous visual image at different acquisition times. For the visual regions with continuous migration relationships, the brightness change state of the corresponding region is further obtained, and the brightness changes of the same region in the continuous visual image at different acquisition times are correlated to determine whether the brightness change regions conform to the change pattern generated during the dynamic light spot movement.

[0032] Since dynamic light spots usually change position, brightness, and range of influence when acting on continuous visual images, the spatial relationship of brightness change areas can be determined based on the movement trajectory of the light spot, the degree of brightness change can be determined based on the amplitude of brightness change, and the range of influence of brightness change areas can be determined based on the illumination range parameters.

[0033] By comprehensively analyzing the light spot movement trajectory, brightness change amplitude, and illumination range parameters, the regions in continuous visual images that migrate over time and whose brightness change relationship corresponds to the dynamic light spot characteristics are identified as the dynamic light spot influence regions. Visual regions that do not conform to the dynamic light spot change relationship are retained as subsequent degradation feature identification regions.

[0034] By using the above method, brightness fluctuations in continuous visual images can be separated, enabling the brightness change information caused by dynamic light spots to be distinguished from the visual change information generated by the deterioration area of ​​the lining structure, thereby reducing the interference of local brightness changes caused by dynamic light spots on the subsequent perception process of deterioration features.

[0035] Based on the brightness fluctuation separation results, and combined with the illumination range parameters and brightness change amplitude from the dynamic light spot interference feature information, the range of the area affected by dynamic light spots in continuous visual images is determined, thus obtaining the range of the area affected by dynamic light spots. The acquisition process is as follows: Based on the dynamic light spot-affected area determined during the brightness fluctuation separation process, spatial range analysis is performed on the corresponding area in the continuous visual image to obtain the positional boundary changes of the dynamic light spot-covered area at different acquisition times. The spatial expansion, contraction, and movement range of the dynamic light spot-covered area in the continuous visual image are determined based on the illumination range parameters, and the actual range of the area affected by the dynamic light spot is determined according to the coverage range changes at different acquisition times.

[0036] Furthermore, the degree of brightness change at different locations within the area affected by the dynamic light spot is described based on the amplitude of the brightness change. This ensures that the determined area affected by the dynamic light spot includes not only the location range information of the dynamic light spot but also the degree of its influence. Specifically, when a region in a continuous visual image has a location change relationship corresponding to the movement trajectory of the light spot, and simultaneously exhibits changes consistent with the amplitude of the brightness change and the illumination range parameters, this region is defined as the area affected by the dynamic light spot. The area affected by the dynamic light spot describes the actual area of ​​influence of the dynamic light spot in the continuous visual image, including the location distribution information, area coverage information, and range of change information of the area affected by the dynamic light spot.

[0037] By utilizing the dynamic light spot interference feature information to further define the brightness change area, the determined area affected by the dynamic light spot can accurately reflect the visual impact range caused by dynamic illumination changes, avoiding the direct identification of local brightness change areas as deterioration areas of the lining structure, and improving the accuracy of distinguishing abnormal areas in continuous visual images.

[0038] Based on the area affected by dynamic light spots, texture change analysis is performed on the corresponding areas in continuous visual images to obtain the texture change range corresponding to the area affected by dynamic light spots. This texture change range is then used as a reference for subsequent anomaly feature screening. The specific process is as follows: The target region to be analyzed in continuous visual images is determined based on the area affected by the dynamic light spot. The texture changes of the target region at multiple acquisition times are correlated to obtain the location, extent, and duration of visual texture changes during the dynamic light spot's influence. The texture change range describes the visual texture changes caused by brightness variations within the area affected by the dynamic light spot, including the location of the texture change, the boundary of the texture change region, the magnitude of the texture change, and the duration of the texture change.

[0039] For regions affected by dynamic light spots in continuous visual images, the texture variation range is correlated with the interference feature information of the dynamic light spots. This allows the texture changes caused by dynamic light spots to be distinguished from the stable texture features generated by the deterioration of the lining structure. Since the texture changes caused by dynamic light spots usually change with the movement trajectory of the light spots, while the visual features corresponding to the actual deteriorated areas have relatively stable positional relationships, the texture variation range can be used to describe the visual change patterns in the regions affected by dynamic light spots. This provides a data basis for subsequently adjusting the edge capture accuracy of the perception area of ​​continuous visual images based on the texture variation range.

[0040] Based on the dynamic light spot interference feature information, brightness fluctuation separation is performed on continuous visual images to determine the area affected by the dynamic light spot, and the corresponding texture change range is further obtained. This allows the visual change information generated by the dynamic light spot to be described separately, reducing the impact of the dynamic light spot on the extraction process of crack contours, shallow damage and other deterioration features. This improves the stability and accuracy of the deep visual perception process for lining structure deterioration features in complex lighting inspection environments.

[0041] Step 3: Adjust the edge capture accuracy of the perception area in the continuous visual image based on the range of texture changes, extract features from the lining structure area in the continuous visual image, obtain the crack outline continuity features, the peeling texture difference features and the abnormal pixel distribution features, and determine the basis for identifying stable diseases based on the crack outline continuity features, the peeling texture difference features and the abnormal pixel distribution features. The criteria for identifying stable defects are determined based on the continuity characteristics of crack outlines, the differences in exfoliation texture, and the distribution characteristics of abnormal pixels. The specific steps are as follows: The edge capture accuracy of the perceived region in continuous visual images is adjusted based on the range of texture variation, and the edge variation description of the lining structure region in continuous visual images is performed. The edge variation description process is as follows: Based on the acquired texture change range, the texture change location, texture change region boundary, texture change amplitude, and texture change duration of the corresponding region in the continuous visual image are read. The texture change range is then mapped to the lining structure region in the continuous visual image to determine the texture change distribution state corresponding to each perception region.

[0042] Furthermore, the distribution area of ​​edge changes in each perception area is determined based on the location of texture changes, the spatial coverage of edge changes is determined based on the boundary of the texture change area, the degree of edge changes is determined based on the amplitude of texture changes, and the retention state of edge changes in continuous visual images is determined based on the persistence state of texture changes, so that each perception area forms corresponding edge change description information.

[0043] Based on this, the edge capture accuracy is adjusted for perception areas with different texture change states. For perception areas where the texture change position remains stable, the texture change state is continuous, and the edge change range remains consistent, the edge capture accuracy of the corresponding area is increased, so that the edge change is described more completely. For perception areas where the texture change position moves with the dynamic light spot, the texture change state changes significantly, and the edge change range changes continuously, the edge capture accuracy of the corresponding area is reduced, so that the edge change formed by the dynamic light spot is no longer described as the real edge of the lining structure.

[0044] Subsequently, the adjusted edge change information of each perception area is re-mapped to the lining structure area in the continuous visual image, so that each area forms a unified edge description state, thereby obtaining the edge change basis suitable for feature extraction of the lining structure area, and establishing a one-to-one correspondence between the texture change range and the edge capture accuracy, providing a unified data basis for the subsequent acquisition of crack contour continuity features, peeling texture difference features, and abnormal pixel distribution features.

[0045] Based on the adjusted edge capture accuracy, feature extraction is performed on the lining structure region in continuous visual images to obtain the continuity features of the crack contour. The acquisition process is as follows: Based on the adjusted edge change description information, the edge distribution in continuous visual images is continuously described. The positional connectivity, edge direction change, boundary extension, and spatial continuity relationships between adjacent edges are obtained, and it is determined whether each edge belongs to the same continuous contour based on these relationships. When multiple edges maintain continuous connectivity, consistent edge extension directions, and continuous spatial correspondence, they are classified as the same crack contour, and the extension path of the crack contour in the continuous visual images is continuously described. When edge changes only manifest as local brightness changes or positional migrations, and cannot maintain continuous connectivity and spatial correspondence, they are not considered part of the continuous crack contour description.

[0046] Furthermore, the continuity of edges, the consistency of edge direction, the spatial connectivity of edges, and the extension state of the contour corresponding to the same crack outline are comprehensively described to form crack outline continuity features. Crack outline continuity features include the positional continuity, directional continuity, edge connectivity, and contour extension of the crack outline. These features are used to fully describe the true continuity of cracks in continuous visual images and serve as the basis for target area division in the subsequent extraction of exfoliation texture difference features. This ensures that texture analysis always revolves around the area corresponding to the true crack outline, thereby guaranteeing that all types of disease features are established on a unified spatial reference basis.

[0047] Using the continuity of crack contours as the basis for target region segmentation, feature extraction is further performed on the lining structure region in continuous visual images to obtain the spalling texture difference features and abnormal pixel distribution features. Based on the continuity of crack contours, target regions in continuous visual images that need further analysis are determined. Combined with the texture change range of the corresponding regions, the texture distribution state within the target regions is continuously described, obtaining the texture arrangement change relationship, texture density change relationship, texture boundary change relationship, texture continuous distribution relationship, and texture correspondence relationship between adjacent regions. The above texture change information is correlated and described to form the spalling texture difference features.

[0048] The spalling texture difference feature is used to describe the texture change state between the spalled area and the normal lining structure area, including differences in texture arrangement, texture continuity, texture boundary integrity, and texture spatial distribution, enabling the spalled area to form a stable texture change description. Furthermore, based on the crack contour continuity feature and the spalling texture difference feature, the pixel changes at each location within the target area are continuously described, obtaining the positional distribution relationship, spatial clustering relationship, discrete distribution relationship, boundary adjacency relationship, and correspondence with the crack contour of abnormal pixels. Abnormal pixels with continuous distribution features, spatial clustering features, and boundary correspondence features are uniformly described to form abnormal pixel distribution features.

[0049] The abnormal pixel distribution features include the distribution location, aggregation range, continuity status of abnormal pixels, and the correspondence between them and crack contours and spalling textures. This allows the crack contour continuity features, spalling texture difference features, and abnormal pixel distribution features to jointly reflect the spatial correlation between different types of deterioration features within the lining structure area, while maintaining a unified target area description method. This provides complete and continuous feature information for the formation of stable disease identification criteria.

[0050] The criteria for identifying stable defects are determined based on the continuity characteristics of crack outlines, the differences in exfoliation texture, and the distribution characteristics of abnormal pixels. The specific determination process is as follows: First, a unified feature correspondence is established based on the spatial positional relationship of the three types of features, so that the crack outline continuity feature, the spalling texture difference feature, and the abnormal pixel distribution feature correspond to the same lining structure area in continuous visual images. Then, based on the continuous maintenance state of the three types of features in continuous visual images, the positional correspondence, boundary correspondence, and change correspondence between the various types of features are continuously described. When the three types of features always correspond to the same spatial area and the correspondence between them remains consistent, the three types of features are combined to form a unified set of disease features. Further, based on the formed set of disease features, the continuous state of crack outline, the change state of spalling texture, and the abnormal pixel distribution state are comprehensively described, so that different types of disease features jointly characterize the deterioration state of the same lining structure area. Finally, the set of disease features with unified spatial correspondence, continuous maintenance, and comprehensive description is determined as the basis for stable disease identification.

[0051] The criteria for identifying stable defects include the continuity of crack outlines, the changes in spalling texture, the distribution of abnormal pixels, and their interrelationships. It also includes the positional, boundary, and continuous change relationships of various features in continuous visual images. This allows the criteria for identifying stable defects to fully describe the visual characteristics of defects actually existing in the lining structure area. Furthermore, it forms a continuous, consistent, and traceable technical chain with the aforementioned texture change range, edge capture accuracy, crack outline continuity features, spalling texture differences, and abnormal pixel distribution features. This provides a sufficient, complete, and directly implementable technical foundation for the next step of constructing feature association rules between continuous visual images, filtering out false anomalies, and retaining genuine deterioration continuous visual features.

[0052] Step 4: Using stable disease identification criteria, perform temporal feature correlation analysis on continuous visual images, construct feature correlation rules between continuous visual images, and filter out false anomaly information generated by the movement of dynamic light spots according to the feature correlation rules, while retaining continuous visual features corresponding to the real deterioration area of ​​the lining structure. Based on feature association rules, false anomalies generated by the movement of dynamic light spots are filtered out, while continuous visual features corresponding to the true deterioration areas of the lining structure are retained. The specific steps are as follows: By utilizing stable disease identification criteria, temporal feature correlation analysis is performed on continuous visual images to establish the correspondence between disease features among continuous visual images. The establishment process is as follows: Based on the established criteria for identifying stable diseases, the continuity features of crack contours, the differences in exfoliation texture, and the distribution features of abnormal pixels are obtained. The positional preservation relationships, boundary preservation relationships, and continuous change relationships corresponding to each type of feature are then read. The criteria for identifying stable diseases are mapped to the same detection area in continuous visual images, ensuring a unified basis for disease feature description in continuous visual images acquired at different times. For the same area corresponding to different acquisition times in continuous visual images, the changes in corresponding disease features are extracted. The positional correspondence, contour extension relationship, and boundary change relationship of crack contour continuity features across different acquisition times are obtained. The texture distribution correspondence, texture change preservation relationship, and texture range change relationship of exfoliation texture differences features across different acquisition times are obtained. The positional distribution correspondence, abnormal pixel continuity relationship, and abnormal region change relationship of abnormal pixel distribution features across different acquisition times are obtained.

[0053] Furthermore, the corresponding changes of the above three types of features in continuous visual images are described in association, and the spatial position preservation state, boundary change preservation state, and feature change preservation state of the same diseased area at different acquisition times are determined, so that the visual features corresponding to the real deteriorated area can form a continuous change relationship in continuous visual images.

[0054] By utilizing stable disease identification criteria to conduct temporal feature correlation analysis, disease feature judgment no longer relies on visual changes at a single acquisition time, but is described based on the continuous correspondence of disease features in continuous visual images, providing a complete temporal feature foundation for the subsequent construction of feature correlation rules between continuous visual images.

[0055] Based on the temporal feature correlation analysis results, feature association rules are constructed between consecutive visual images. Specifically, based on the changes in disease characteristics at different acquisition times in consecutive visual images, the association relationships between consecutive visual images used to describe the changing patterns of the actual deteriorated areas are determined. The specific process is as follows: First, based on the positional changes corresponding to the continuity features of the crack outline, the spatial correspondence of the same crack region in continuous visual images is determined; based on the texture changes corresponding to the differences in peeling texture, the texture correspondence of the same peeling region in continuous visual images is determined; based on the pixel changes corresponding to the abnormal pixel distribution features, the distribution correspondence of the same abnormal region in continuous visual images is determined.

[0056] Subsequently, based on the aforementioned spatial correspondence, texture correspondence, and distribution correspondence, the boundary correspondence, change correspondence, and temporal continuity relationship among various disease characteristics are further determined, so that the same disease area can maintain a consistent feature description at different collection times.

[0057] Furthermore, positional correspondence is used as a spatial constraint for feature association between consecutive visual images, boundary correspondence is used as a range constraint for feature association between consecutive visual images, change correspondence is used as a change constraint for feature association between consecutive visual images, and temporal continuity is used as a preservation constraint for feature association between consecutive visual images. Feature association rules between consecutive visual images are established based on the above multiple correspondences.

[0058] The specific process for establishing feature association rules between consecutive visual images is as follows: Based on the continuity characteristics of crack contours, the differences in peeling texture, and the distribution characteristics of abnormal pixels in the criteria for identifying stable defects, a correspondence analysis is performed on adjacent image frames in a continuous visual image sequence. Let the continuous visual image sequence be represented as: ,in, Indicates the first Continuous visual images corresponding to each acquisition time. This represents the total number of frames in a continuous visual image. For any two adjacent acquisition times, this refers to the visual images... and Each stable disease identification criterion corresponds to a set of disease features: ,in, Indicates the first The set of stable disease characteristics corresponding to each collection time; Indicates the first The location features of the diseased area at each acquisition time are used to describe the spatial location information of the diseased area in continuous visual images; Indicates the first The boundary features of the diseased area corresponding to each collection time are used to describe the outline range and boundary change information of the diseased area; Indicates the first The disease change features corresponding to each acquisition time are used to describe the relationship between the changes in crack contour continuity features, peeling texture difference features, and abnormal pixel distribution features between adjacent image frames. Indicates the first The temporal continuity features corresponding to each acquisition time are used to describe the retention status of disease features in continuous visual images. Among them, It is a normalized set of variation features formed by the continuity variation features of crack contours, the difference variation features of peeling texture, and the variation features of abnormal pixel distribution. Each change feature is numerically described based on the location range, boundary range, and distribution range of the corresponding disease feature in continuous visual images, so that the change features of different types of diseases can be compared within the same value range.

[0059] First, spatial constraints are established based on positional correspondence. For adjacent image frames... and Obtain the location characteristics of the same diseased area. and Calculate the positional changes: ,in, The positional correspondence parameter is used to indicate the degree to which the position of the disease area is maintained between adjacent consecutive visual images; This represents the location distance parameter, used to describe the range of location change of the same diseased area between two data collection times. When the preset spatial correspondence conditions are met, it is determined that the disease areas corresponding to the two acquisition times have a spatial correspondence, and this spatial correspondence is used as a spatial constraint parameter for the feature association rules between continuous visual images. Because Used to characterize the locational difference of the same diseased area between adjacent sampling times; the smaller the locational difference, the better. The closer a value is to 1, the higher the degree to which the location of the diseased area is preserved in continuous visual images; the greater the location difference, The closer a value is to 0, the less stable the corresponding region is in terms of location.

[0060] Secondly, range constraints are established based on boundary correspondences. Boundary features of corresponding disease areas in adjacent continuous visual images are obtained. and The degree of boundary change is determined based on the overlap between two boundary features: ,in, The boundary correspondence parameter describes the degree of consistency in the boundary range of disease areas between consecutive visual images; This indicates the overlapping area of ​​the diseased region at the two data collection times; This indicates the total extent of the diseased area boundary corresponding to the two data collection times. When When the boundary changes of the diseased area remain within the corresponding range, it indicates that the boundary changes satisfy a continuous maintenance relationship. This boundary correspondence is used as a range constraint parameter for the feature association rules between continuous visual images. and All are determined under the same spatial correspondence of continuous visual images. The larger the value, the higher the degree of overlap between the diseased area boundaries at adjacent data collection times. The smaller the value, the greater the degree of change in the boundary of the diseased area at adjacent sampling times.

[0061] Third, establish change constraint relationships based on the correspondence of changes. Obtain change information corresponding to the continuity features of crack contours, the difference features of peeling texture, and the distribution features of abnormal pixels in continuous visual images, and construct a set of change features: ,in, It represents the continuity and variation characteristics of the crack outline, and is used to describe the extension and change of the crack outline in continuous visual images. It represents the variation characteristics of peeling texture and is used to describe the changes in texture variation areas in continuous visual images; It represents the characteristics of abnormal pixel distribution changes, used to describe the changes in the location and range of abnormal pixel regions. , as well as All values ​​are normalized variation feature values, ranging from 0 to 1. When the corresponding type of disease feature does not exist, the variation feature value is 0. When the corresponding type of disease feature exists and forms a continuous variation description, the variation feature value is determined according to the continuous range, variation range and distribution range of the corresponding disease feature.

[0062] Calculate the consistency parameter based on the set of change features corresponding to two adjacent acquisition times: ,in, The parameter representing the change correspondence is used to describe the degree of consistency in the trend of disease characteristics change between consecutive visual images; This indicates the difference in disease characteristics between two adjacent data collection times; Indicates the first Each data collection point corresponds to a range of disease change characteristics. Indicates the first Each data collection point corresponds to a range of disease change characteristics. Among them, express , as well as The sum of the absolute values ​​of the corresponding changing eigenvalues Indicates the first The sum of the absolute values ​​of each change feature value in the set of change features corresponding to each acquisition time. It represents the sum of the absolute values ​​of the differences between the same type of feature values ​​in the feature set corresponding to two adjacent acquisition times.

[0063] This consistency parameter is obtained by the proportion of the difference in disease change characteristics between adjacent data collection times to the total amount of disease change characteristics between the two data collection times. This is first determined by... Determine the differences in change between the two frames, and then... The differences in variation are normalized, and finally, 1 is subtracted from the normalized differences to obtain the result that characterizes the consistency of the disease's trend. ; The closer a value is to 1, the more consistent the disease change trend is between adjacent data collection times. The closer the value is to 0, the greater the difference in the disease change trend between adjacent collection times.

[0064] when and When all values ​​are 0, it indicates that there are no changes in crack contour continuity, peeling texture differences, or abnormal pixel distribution in the corresponding areas at two adjacent acquisition times. These areas are not considered as objects for judging the changes in the true deteriorated area. If the value is set to 0, no change correspondence is established based on the corresponding region; when When greater than 0, according to Calculate the consistency parameter of the change.

[0065] when When the continuous change condition is met, a change correspondence is determined between two acquisition times, and the change correspondence is used as the change constraint parameter of the feature association rule between continuous visual images.

[0066] Finally, maintenance constraints are established based on temporal continuity. For multiple acquisition times within a continuous visual image sequence, the continuity of disease features is statistically analyzed: ,in, The parameter representing the temporal continuity is used to describe the degree to which disease characteristics are maintained in continuous visual images; This indicates the number of image frames that meet the condition of continuous preservation. This represents the total number of frames in the continuous visual images. Based on the temporal continuity parameter, it is determined whether the disease features remain stable in the continuous visual images, and the temporal continuity is used as a preservation constraint parameter for the feature association rules between continuous visual images. Specifically, image frames that satisfy the continuity preservation condition refer to image frames whose corresponding disease areas simultaneously satisfy positional correspondence, boundary correspondence, and change correspondence in the continuous visual images. The larger the value, the higher the degree to which the corresponding disease characteristics are maintained in continuous visual images.

[0067] Furthermore, spatial constraint parameters Range constraint parameters Variation constraint parameters and maintain constraint parameters Link: ,in, The parameter represents the feature association rule between consecutive visual images, used to describe the degree of association of disease features between two consecutive visual images; This represents the spatial constraint weight parameter, used to adjust the degree of influence of positional correspondence in feature association rules; The range constraint weight parameter is used to adjust the degree of influence of boundary correspondence in feature association rules; This represents the change constraint weight parameter, used to adjust the degree of influence of the change correspondence in the feature association rule; This represents the constraint weight parameter, used to adjust the influence of temporal continuity in feature association rules, and satisfies: ,in, , , as well as All of these are normalized constraint parameters, with values ​​ranging from 0 to 1; , , as well as respectively with , , as well as One-to-one correspondence, used to represent spatial constraint parameters, range constraint parameters, variation constraint parameters, and preservation constraint parameters in feature association rule parameters. The proportion of contribution. , , as well as All are non-negative numbers and satisfy the following conditions: , , , as well as .

[0068] The weight parameters are normalized correlations. When any weight parameter increases, at least one of the other weight parameters decreases accordingly to keep the sum of the four weight parameters equal to 1. The four weight parameters do not change the technical meaning of the corresponding constraint parameters, but are only used to determine the proportion of positional correspondence, boundary correspondence, change correspondence, and temporal continuity in the comprehensive correlation judgment.

[0069] As one specific implementation method, , , as well as All parameters are set to 0.25 to ensure that spatial constraint parameters, range constraint parameters, variation constraint parameters, and maintenance constraint parameters contribute in equal proportions. The determination.

[0070] When feature association rule parameters When the preset association conditions are met, it is determined that the visually changing regions in adjacent consecutive visual images belong to the same real deterioration region, corresponding to continuous visual features. The corresponding positional correspondence, boundary correspondence, change correspondence, and temporal continuity are stored as feature association rules between consecutive visual images. When the feature association rule parameters... If the preset association conditions are not met, the visually changed region is considered to lack stable and continuous features and will not be included in the continuous visual feature association range corresponding to the actual deteriorated region. The preset association conditions are as follows: The threshold value is not less than a preset correlation threshold, which is pre-set based on the requirement for continuous preservation of the actual degraded region in continuous visual images; as a specific implementation, the preset correlation threshold is 0.6. When the value is not less than 0.6, the visually changing regions in adjacent consecutive visual images are determined to satisfy the feature association rule. When the value is less than 0.6, it is determined that the visually changing regions in adjacent consecutive visual images do not satisfy the feature association rule.

[0071] By changing the consistency parameter in the above manner exist and When both denominators are 0, there is a clear way to handle it, avoiding the mathematically meaningless situation caused by a denominator of 0; at the same time, The calculation source, the meaning of each variable, and the technical significance of the calculation results are all clearly explained, enabling the determination of the consistency of disease change trends between adjacent consecutive visual images. Furthermore, by limiting... , , as well as Nonnegative normalization relation, one-to-one correspondence relation, specific value selection method and its relation to The contribution ratio relationship makes the correlation between weight parameters clear and explicit, thereby enabling the feature association rules between continuous visual images to be directly implemented.

[0072] Through the above process, positional correspondence is used as spatial constraint, boundary correspondence as range constraint, change correspondence as change constraint, and temporal continuity as maintenance constraint. Feature association rules between continuous visual images are established through multi-constraint association, so that the real deteriorated area can be continuously described based on spatial stability, boundary continuity, change consistency, and temporal maintenance. At the same time, it provides a clear and implementable technical basis for subsequent screening of false anomaly information generated by the movement of dynamic light spots.

[0073] Feature association rules are used to describe the stable variation patterns of real degraded regions in continuous visual images, including the spatial position preservation relationship, contour change preservation relationship, texture change preservation relationship, and abnormal pixel distribution preservation relationship corresponding to real degraded regions.

[0074] By establishing feature association rules, visual change information in continuous visual images can be distinguished according to the time change pattern, providing a clear basis for judging the elimination of false anomalies generated by the movement of dynamic light spots.

[0075] False anomalies generated by the movement of the dynamic light spot are filtered out based on feature association rules. The specific filtering process is as follows: Feature association rules are used to analyze the correlation between visually changing regions in continuous visual images. The visually changing regions at each acquisition time are compared with the continuous visual features corresponding to the actual deteriorated regions to determine whether different visually changing regions conform to the continuous change patterns of the actual deteriorated regions. When a visually changing region only exhibits changes in brightness, position migration, or range, and the positional change process of this visually changing region maintains a correspondence with the light spot movement trajectory, brightness change amplitude, and illumination range parameters in the dynamic light spot interference feature information, but fails to satisfy the spatial position preservation relationship, boundary change preservation relationship, texture change preservation relationship, and temporal continuity relationship in the feature association rules, this visually changing region is identified as pseudo-anomaly information generated by the movement of the dynamic light spot.

[0076] Specifically, pseudo-anomalies generated by dynamic light spots typically manifest as changes in the location of the anomalous area as the light spot moves, changes in the anomalous boundary as the illumination range changes, and expansion or contraction of the anomalous pixel distribution range as brightness changes. Furthermore, they cannot maintain a fixed positional correspondence or continuous change relationship in continuous visual images. For visually anomalous areas that meet these characteristics, they are filtered out from the scope of the disease feature description based on feature association rules, ensuring that visual interference information generated by dynamic light spots does not participate in the identification process of true deteriorated areas. Visually anomalous areas that can continuously maintain spatial positional correspondence, boundary correspondence, texture change correspondence, and anomalous pixel distribution correspondence are retained as continuous visual features corresponding to true deteriorated areas.

[0077] By distinguishing dynamic spot change features from true degradation features based on feature association rules, false anomaly information in continuous visual images can be effectively eliminated, reducing the impact of dynamic spot movement on the disease identification results.

[0078] Finally, continuous visual features corresponding to the actual deteriorated areas of the lining structure are retained and used for subsequent defect location parameter updates. The specific implementation process is as follows: Based on the continuous visual images after filtering out false anomalies, visual change information corresponding to the true deterioration areas that conform to feature association rules is extracted. The crack contour continuity features, peeling texture difference features, and abnormal pixel distribution features are re-associated to form continuous visual features corresponding to the true deterioration areas. The continuous visual features include the location continuity information, contour continuity change information, texture continuity change information, and abnormal pixel continuity distribution information corresponding to the true deterioration areas.

[0079] Among them, location continuity information is used to describe the positional relationship of the real deteriorated area at different acquisition times, contour continuity change information is used to describe the change process of crack contour or abnormal boundary in continuous visual images, texture continuity change information is used to describe the texture change state corresponding to the peeling area, and abnormal pixel continuity distribution information is used to describe the distribution and maintenance of abnormal areas in continuous visual images.

[0080] Furthermore, based on the location continuity information, regional change information, and time correspondence information corresponding to the continuous visual features, the persistent existence state of the real deteriorated area in the continuous visual image is determined, so that the output continuous visual features simultaneously have spatial positioning information and temporal change information.

[0081] By preserving continuous visual features, the short-term change information generated by dynamic light spots can be distinguished from the stable change information corresponding to the real deteriorated area, and the real deteriorated area can maintain a complete feature description relationship in continuous visual images.

[0082] Continuous visual features serve as the basis for subsequent updates to defect location parameters, enabling the defect location parameters to be adjusted based on the positional changes, range changes, and continuous distribution relationships of the actual deteriorated areas, thereby outputting deterioration perception information that is resistant to light spot interference.

[0083] By using the above method, the temporal feature correlation analysis of continuous visual images is performed based on stable defect identification criteria, feature correlation rules between continuous visual images are constructed, and false anomaly information generated by the movement of dynamic light spots is screened out according to the feature correlation rules. Finally, the continuous visual features corresponding to the real deterioration area of ​​the lining structure are retained, so that the entire deep visual perception process forms a complete technical chain of dynamic light spot interference feature information, texture change range, stable defect identification criteria, and continuous visual features that are correlated step by step, thereby improving the stability and reliability of lining structure deterioration feature identification under complex lighting conditions.

[0084] Step 5: Update the defect localization parameters in the continuous visual image detection process based on continuous visual features, and output the lining structure deterioration perception information that is resistant to light spot interference. The defect localization parameters in the continuous visual image detection process are updated based on continuous visual features, and the deterioration perception information of the lining structure, which is resistant to light spot interference, is output. The specific steps are as follows: The defect localization parameters in the continuous visual image detection process are updated based on continuous visual features. The specific update process is as follows: The continuous visual features retained after dynamic spot interference removal in the aforementioned steps are obtained. The continuous visual features include the continuous location information, continuous contour change information, continuous texture change information, and continuous distribution information of abnormal pixels corresponding to the real deteriorated area. The defect localization parameters in the continuous visual image detection process are updated based on the continuous visual features. Specifically, based on continuous location information, the positional changes of the actual deteriorated area at different acquisition times are determined. The positions of the actual deteriorated area at multiple acquisition times in the continuous visual images are continuously correlated to obtain the positional retention status of the actual deteriorated area. The positional positioning parameters in the defect localization parameters are then updated based on the positional retention status. Based on continuous contour change information, the change relationship of the boundary range corresponding to the actual deteriorated area in the continuous visual images is determined. The boundary extension direction, boundary change range, and boundary retention status of the actual deteriorated area are obtained. The range localization parameters in the defect localization parameters are then updated based on the boundary change relationship. Based on continuous texture change information, the visual texture change range corresponding to the actual deteriorated area is determined. The positional distribution, change direction, and continuous change status of the texture change area are obtained. The region change parameters in the defect localization parameters are then updated based on the continuous texture change relationship. Based on continuous abnormal pixel distribution information, the distribution range of abnormal pixels corresponding to the actual deteriorated area is determined. The continuous distribution status of abnormal pixels in the continuous visual images is obtained. The region coverage parameters in the defect localization parameters are then updated based on the continuous abnormal pixel distribution relationship.

[0085] The defect localization parameters include location parameters describing the spatial position of the actual deteriorated area, range parameters describing the boundary extent of the actual deteriorated area, region change parameters describing the changing state of the actual deteriorated area, and region coverage parameters describing the continuous distribution of the actual deteriorated area. Further, continuous location information, continuous contour change information, continuous texture change information, and continuous distribution information of abnormal pixels are associated with their corresponding defect localization parameters, enabling different types of features in continuous visual features to adjust their respective defect localization parameters accordingly. Specifically, the direction of position adjustment for the defect localization parameters is determined based on continuous location information; the boundary adjustment range of the defect localization parameters is determined based on continuous contour change information; the region change trend of the defect localization parameters is determined based on continuous texture change information; and the region coverage range of the defect localization parameters is determined based on continuous distribution information of abnormal pixels, ensuring that the updated defect localization parameters continuously correspond to the changes in the actual deteriorated area within the continuous visual image.

[0086] By updating the defect localization parameters using continuous visual features after dynamic spot interference filtering, the defect localization process is adjusted according to the continuous change relationship corresponding to the actual deteriorated area. This avoids the influence of local brightness changes, texture changes, and abnormal pixel changes caused by dynamic spots on the change process of defect localization parameters, thereby improving the accuracy of the correspondence between defect localization parameters and the actual deteriorated area.

[0087] Based on the updated defect location parameters, deterioration sensing information of the lining structure, resistant to light spot interference, is output. The specific process is as follows: Based on the updated defect localization parameters, the true extent of the deteriorated region in the continuous visual image is determined. The location, extent, region change, and region coverage information corresponding to the defect localization parameters are then correlated with the continuous visual features to generate anti-spot interference-resistant lining structure deterioration perception information. Specifically, the spatial distribution of the true deteriorated region in the continuous visual image is determined based on the location parameters, and this spatial distribution is correlated with the continuous location information in the continuous visual features, ensuring that the output deterioration perception information accurately describes the spatial relationship of the true deteriorated region. The boundary range corresponding to the true deteriorated region is determined based on the extent localization parameters, and this boundary range is correlated with the continuous contour change information, ensuring that the output deterioration perception information reflects the contour change state corresponding to the true deteriorated region. The visual change range corresponding to the true deteriorated region is determined based on the region change parameters, and this visual change range is correlated with the continuous texture change information, ensuring that the output deterioration perception information describes the texture change state corresponding to the true deteriorated region. The abnormal pixel distribution range corresponding to the true deteriorated region is determined based on the region coverage parameters, and this abnormal pixel distribution range is correlated with the continuous abnormal pixel distribution information, ensuring that the output deterioration perception information reflects the abnormal distribution state corresponding to the true deteriorated region.

[0088] Furthermore, location description information, area range information, contour change information, texture change information, and abnormal pixel distribution information are uniformly correlated to form lining structure degradation perception information that includes the spatial location, area range, and continuous change state of the actual deteriorated area.

[0089] Since the continuous visual features have already filtered out the false abnormal information generated by the movement of dynamic light spots based on the stable disease identification criteria and feature association rules, the output lining structure deterioration perception information only retains the continuous visual changes corresponding to the real deterioration area, avoiding the inclusion of local bright areas, texture change areas and abnormal pixel change areas caused by dynamic light spots into the final perception result.

[0090] The lining structure deterioration perception information that resists light spot interference can describe the location, extent, and change of the actual deteriorated area under complex lighting conditions, and maintain the positional correspondence and temporal continuity with continuous visual images, providing a stable visual perception basis for subsequent structural safety assessment, disease development analysis, and maintenance decisions.

[0091] By updating the defect location parameters in the continuous visual image detection process based on continuous visual features and outputting anti-spot interference lining structure deterioration perception information, the deep visual perception process for lining structure deterioration features is formed into a complete technology chain consisting of dynamic spot interference feature extraction, brightness fluctuation separation, stable defect identification, continuous visual feature association, defect location parameter update, and deterioration perception information output. This improves the identification stability and perception accuracy of hidden cracks and shallow damage areas under dynamic lighting conditions.

[0092] The foregoing has only described certain exemplary embodiments of this application by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of this application. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of this application.

Claims

1. A depth visual perception method for assessing the deterioration characteristics of lining structures, characterized in that, Includes the following steps: Continuous visual images are collected during tunnel inspection. Based on the continuous visual images, the movement trajectory of the light spot, the amplitude of brightness change, and the illumination range parameters are extracted. Dynamic light spot interference feature information is generated based on the movement trajectory of the light spot, the amplitude of brightness change, and the illumination range parameters. Based on the dynamic light spot interference feature information, brightness fluctuation separation is performed on the continuous visual image to determine the region range affected by the dynamic light spot in the continuous visual image, and the texture change range corresponding to the region range affected by the dynamic light spot is obtained. The edge capture accuracy of the perceived area in the continuous visual image is adjusted based on the texture change range. The edge capture accuracy is adjusted according to the texture change position, texture change area boundary, texture change amplitude, and texture change persistence state, and is used to determine the distribution area, spatial coverage, edge change degree, and persistence state of edge changes within the perceived area. Feature extraction is performed on the lining structure area in the continuous visual image to obtain crack contour continuity features, spalling texture difference features, and abnormal pixel distribution features. The crack contour continuity features include the positional continuity relationship, directional continuity relationship, edge connection relationship, and contour extension relationship of the crack contour. The spalling texture difference features include the texture arrangement difference, texture continuity degree difference, texture boundary integrity degree, and texture spatial distribution difference between the spalling area and the normal lining structure area. The abnormal pixel distribution features include the distribution position, aggregation range, continuity state, and correspondence between abnormal pixels and the crack contour and the spalling texture. The criteria for identifying stable defects are determined based on the crack contour continuity features, the spalling texture difference features, and the abnormal pixel distribution features. Using the stable disease identification criteria, temporal feature correlation analysis is performed on the continuous visual images to construct feature correlation rules between the continuous visual images. Based on the feature correlation rules, false anomaly information generated by the movement of the dynamic light spot is filtered out, and continuous visual features corresponding to the real deterioration area of ​​the lining structure are retained. Based on the continuous visual features, the defect localization parameters in the continuous visual image detection process are updated, and the lining structure deterioration perception information that is resistant to light spot interference is output.

2. The depth visual perception method for assessing the deterioration characteristics of lining structures according to claim 1, characterized in that, The dynamic light spot interference feature information is generated based on the light spot movement trajectory, the brightness change amplitude, and the illumination range parameters. The specific steps are as follows: The brightness change region in the continuous visual image frame is continuously tracked. Based on the positional change relationship of the brightness change region at different acquisition times, the movement direction, movement distance and movement range of the brightness change region in the continuous visual image are determined, and the movement trajectory of the dynamic light spot in the continuous visual image is obtained. The visual area traversed by the dynamic light spot is determined based on the light spot movement trajectory. The continuous visual image frames before and after the dynamic light spot pass through are analyzed to obtain the brightness difference information of the same visual area at different acquisition times. The brightness change amplitude caused by the dynamic light spot is determined based on the brightness difference information. For the visual region corresponding to the light spot movement trajectory, the boundary change of the dynamic light spot coverage area is obtained. The illumination range parameter is determined according to the position change range of the dynamic light spot coverage area at different acquisition times. The dynamic light spot interference feature information is generated by combining the light spot movement trajectory, the brightness change amplitude, and the illumination range parameter.

3. The depth visual perception method for assessing the deterioration characteristics of lining structures according to claim 2, characterized in that, The brightness change amplitude is obtained by corresponding analysis of the continuous visual image frames before and after the dynamic light spot passes through, obtaining the brightness difference information of the same visual area at different acquisition times, and determining the degree of brightness increase, the degree of brightness decrease, and the duration of brightness change during the dynamic light spot's action based on the brightness difference information.

4. The depth visual perception method for lining structure deterioration characteristics according to claim 2, characterized in that, The specific steps for obtaining the texture change range corresponding to the region affected by the dynamic light spot are as follows: Based on the light spot movement trajectory, the brightness change amplitude, and the illumination range parameter in the dynamic light spot interference feature information, the brightness fluctuation separation is performed on the continuous visual image to determine the area affected by the dynamic light spot. Based on the area affected by the dynamic light spot, a spatial range analysis is performed on the corresponding area to obtain the positional boundary changes of the area covered by the dynamic light spot. Based on the changes in the positional boundaries, a texture change analysis is performed on the region affected by the dynamic light spot to obtain the corresponding texture change range.

5. The depth visual perception method for lining structure deterioration characteristics according to claim 4, characterized in that, Based on the illumination range parameters, the spatial expansion range, spatial contraction range, and spatial movement range of the dynamic light spot coverage area in the continuous visual image are determined, and the range of the area affected by the dynamic light spot is determined by combining the coverage range change relationship corresponding to different acquisition times.

6. The depth visual perception method for lining structure deterioration characteristics according to claim 4, characterized in that, The criteria for identifying stable defects are determined based on the continuity features of the crack contour, the differences in the peeling texture, and the abnormal pixel distribution features. The specific steps are as follows: The texture change distribution state corresponding to the sensing area is determined based on the texture change range, and the edge capture accuracy is adjusted based on the texture change location, the boundary of the texture change area, the texture change amplitude, and the texture change duration. Based on the adjusted edge capture accuracy, feature extraction is performed on the lining structure area. The crack contour continuity feature is obtained according to the edge connection relationship, the edge direction change relationship, the boundary extension relationship and the spatial continuity relationship. The peeling texture difference feature and the abnormal pixel distribution feature are obtained by combining the crack contour continuity feature. Based on the continuity features of the crack outline, the difference features of the peeling texture, and the abnormal pixel distribution features, a feature correspondence relationship is established. The position correspondence relationship, boundary correspondence relationship, and change correspondence relationship are described. The set of disease features with continuous relationship is determined as the basis for identifying stable diseases.

7. The depth visual perception method for lining structure deterioration characteristics according to claim 6, characterized in that, The continuity features of the crack contour are obtained based on edge change description information, specifically including: Based on the edge change description information, the positional connection relationship, the edge direction change relationship, the boundary extension relationship, and the spatial continuity relationship between adjacent edges are obtained, and edges that maintain continuous connection and correspond to each other in spatial position are classified into the same crack contour. Based on a comprehensive description of the crack profile's edge continuity, edge direction consistency, edge spatial connection state, and profile extension state, a crack profile continuity feature is formed that includes the positional continuity relationship, the directional continuity relationship, the edge connection relationship, and the profile extension relationship.

8. The depth visual perception method for lining structure deterioration characteristics according to claim 6, characterized in that, Based on the feature association rules, the false anomaly information generated by the movement of the dynamic light spot is filtered out, and the continuous visual features corresponding to the real deterioration area of ​​the lining structure are retained. The specific steps are as follows: The crack contour continuity feature, the peeling texture difference feature, and the abnormal pixel distribution feature are obtained. The disease features corresponding to different acquisition times in the continuous visual images are correlated to determine the position preservation relationship, boundary preservation relationship, continuous change relationship, and temporal continuity relationship. Based on the position preservation relationship, the boundary preservation relationship, and the continuous change relationship, feature association rules are established between the continuous visual images. The position correspondence relationship, the boundary correspondence relationship, the change correspondence relationship, and the temporal continuity relationship are used as association constraints to determine the continuous visual features corresponding to the real deterioration area. The visual change regions in the continuous visual image are filtered using the feature association rules. The visual change regions that conform to the feature association rules are retained as the continuous visual features corresponding to the real deterioration regions. The pseudo-anomaly information generated by the movement of the dynamic light spot is filtered out, and the deterioration perception information of the lining structure that resists light spot interference is output.

9. The depth visual perception method for lining structure deterioration characteristics according to claim 8, characterized in that, Based on the feature association rules between the continuous visual images, the position correspondence, the boundary correspondence, the change correspondence, and the temporal continuity are obtained. Feature association rule parameters are calculated based on the position correspondence, the boundary correspondence, the change correspondence, and the temporal continuity. When the visual change region satisfies the corresponding condition of the feature association rule parameters, it is determined that the visual change region belongs to the continuous visual feature corresponding to the real deterioration region.

10. The depth visual perception method for lining structure deterioration characteristics according to claim 8, characterized in that, The defect localization parameters in the continuous visual image detection process are updated based on the continuous visual features, and the anti-spot interference lining structure deterioration perception information is output. The specific steps are as follows: The continuous visual features retained after the dynamic light spot interference filtering are obtained, and the defect localization parameters are updated based on the continuous position information, continuous contour change information, continuous texture change information, and continuous distribution information of abnormal pixels. The updated defect location parameters are associated with the continuous visual features, and the location range and change status of the actual deteriorated area are determined based on the location location parameters, range location parameters, area change parameters, and area coverage parameters. Based on the determined location range and change status of the actual deteriorated area, the anti-spot interference lining structure deterioration perception information is generated, which includes spatial location, regional range, and continuous change status.

Images