Bridge pier underwater structure image intelligent analysis method and system

Abstract

The invention relates to the technical field of image recognition, in particular to a bridge pier underwater structure image intelligent analysis method and system, and the method comprises the following steps: obtaining RGB channel values of all pixels of a bridge underwater structure image region, converting a main color temperature, generating a color temperature map, dividing the region in a vertical direction, extracting a main color temperature track, and carrying out the linear fitting. And comparing the residual error with a continuity threshold, and marking a color temperature continuous section. According to the method, a two-dimensional map is constructed through fusion of main color temperature and space coordinates, spatial distribution expression of a structural region is enhanced, a stable region is extracted by combining residual and continuity discrimination, interval judgment and RGB finishing are introduced into gray scale lifting, image brightness balance and color consistency are enhanced, a gray scale trend model is constructed through radial chain grouping, and light attenuation interference stripping is achieved. Feature point continuous frame tracking is combined with disturbance vector mapping, structure dynamic changes are captured, abnormal behaviors are recognized through cross analysis of tension differences and response terms, and efficient recognition and interference isolation under multi-dimensional feature fusion are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition, and particularly to an intelligent analysis method and system for the underwater structure image of a bridge pier. Background Art

[0002] The technical field of image recognition includes operations such as image content acquisition, processing, feature extraction, matching, and recognition, which are used to achieve automatic recognition and information parsing of specific objects or structures. The core content of this technical field lies in the quantitative analysis of information such as shapes, edges, and textures existing in the image through image processing algorithms and feature matching methods, thereby realizing the recognition of physical structures in the image. Image recognition technology has been widely applied in fields such as industrial inspection, medical image analysis, traffic monitoring, and environmental perception. In bridge engineering, image recognition technology is used for the recognition and analysis of underwater structures. By obtaining image data of the underwater part of the bridge and extracting its key structural features, it provides image-based data support for subsequent structural condition assessment, and is one of the key technical means for intelligent bridge inspection and maintenance.

[0003] Among them, the intelligent analysis method for the underwater structure image of a bridge pier refers to the acquisition and processing of image data of the underwater structure of a bridge, and adopts methods such as image enhancement, feature region extraction, and structural form recognition to identify and classify the underwater structure of the bridge pier in the image. This method covers operations such as resolution improvement, brightness correction, and noise suppression in the image preprocessing step to improve the quality of underwater images; in the process of structural recognition, image contour tracking and feature point matching methods are used to analyze the structural boundaries and their spatial distributions of the bridge pier; and the target area calibration is completed by combining image segmentation and comparative analysis methods to extract information on surface deformation, defects, and accessory component structures of the bridge pier. This method is applied to the underwater structure recognition task in bridge engineering, and realizes the image-level analysis of the bridge pier structure through multiple steps of image processing operations.

[0004] In the prior art, image recognition of underwater structures mainly focuses on image contour and static feature extraction, insufficiently expressing the hierarchical relationship of the spatial distribution of each region in the image, lacking a hierarchical recognition mechanism for the inherent continuity features of structural regions, resulting in insignificant distinction between structural stable regions and transition regions. Image brightness adjustment is often based on a unified enhancement strategy, ignoring the gray-scale distribution differences in different regions of the image, easily causing color distortion in high-contrast regions and affecting the restoration degree of feature regions. In a complex background, when there are multi-directional gray-scale perturbations in the image, traditional image segmentation methods are difficult to effectively strip the illumination interference blocks, resulting in unclear structural boundaries and interference from accessory information. Feature recognition focuses on static frame or single-frame feature point matching, lacking continuous analysis of the structural motion state, making it difficult to capture potential perturbation trends. Without dynamic behavior quantification, the trajectory features of structural state changes are not easily distinguishable, resulting in missed or misrecognized abnormal behaviors in multi-interference scenarios and affecting the stability and timeliness of recognition results. Summary of the Invention

[0005] The object of the present invention is to solve the shortcomings existing in the prior art and propose an intelligent analysis method for the underwater structure image of a bridge pier.

[0006] To achieve the above object, the present invention adopts the following technical solutions: An intelligent analysis method for the underwater structure image of a bridge pier, including the following steps:

[0007] S1: Obtain the RGB channel values of all pixels in the underwater structure image area of the bridge, convert the main color temperature, generate a color temperature map, divide the area in the vertical direction, extract the main color temperature trajectory and linearly fit it, compare the residual and the continuity threshold, and mark the color temperature continuous paragraphs;

[0008] S2: Extract the gray-scale values based on the continuous paragraphs, set an enhancement coefficient to linearly enhance the gray scale, calculate the RGB trimming amount, and generate a brightness balance layer;

[0009] S3: Based on the brightness balance layer, set equidistant radial chains at the outer edge of the bridge pier foundation, extract the gray-scale sequence, calculate the gray-scale change rate, screen and group the chains with an included angle less than the threshold, and generate a stripping boundary map according to the ratio of the texture boundary number to the gray-scale gradient;

[0010] S4: According to the stripping boundary map, extract the continuous frame coordinates of the structural feature points at the component gaps, calculate the velocity vector and acceleration change, combine with the perturbation vector field of the non-structural area, extract the direction offset angle and velocity amplitude difference, and generate a perturbation comparison parameter group;

[0011] S5: According to the perturbation comparison parameters, extract the pairs of structural point direction offset and velocity change trajectories, analyze the coincidence degree of the tension difference and the perturbation response, screen the abnormal segments with synchronous perturbations and trajectory mutations, and output the analysis result of the underwater trajectory image of the bridge.

[0012] As a further solution of the present invention, the color temperature continuous paragraph includes a regional main color temperature value, a fitting trajectory parameter, and a continuity discrimination label. The brightness balance layer includes a grayscale enhancement pixel value, an RGB channel correction ratio, and a brightness uniform distribution feature. The peeled boundary map includes a light decay block range, a chain grouping result, and an interference recognition ratio distribution. The perturbation contrast parameter group includes a structural offset direction, a speed change amplitude, and a perturbation response difference. The analysis result of the underwater trajectory image of the bridge includes the position of an abnormal behavior segment, a trajectory direction mutation point, and a synchronous perturbation feature.

[0013] As a further solution of the present invention, the steps for obtaining the color temperature continuous paragraph are as follows:

[0014] S101: Based on obtaining the RGB channel response values of all pixel points in the area of the underwater structure image of the bridge, call the RGB channel values to perform a main color temperature conversion operation on each pixel, calculate the transformation relationship between the values of the pixel points in the R, G, and B channels and the color temperature, extract the transformed color temperature values in combination with the spatial coordinates where the pixels are located, obtain the corresponding combination of the color temperature and the image spatial coordinates, and generate a two-dimensional color temperature map;

[0015] S102: Call the two-dimensional color temperature map, divide it into multiple hierarchical areas in the vertical direction of the image, extract the color temperature values of the corresponding pixels in each divided area, calculate the average change trend of the color temperature values of the pixel points in the area, and use the formula:

[0016]

[0017] Perform an operation to obtain the change trend of the color temperature difference between adjacent areas, use the color temperature trend of the area sequence as a trajectory data set, perform a linear fitting process, obtain the fitting residual data, and obtain the distribution of the color temperature trajectory residual values;

[0018] Among them, T i,j represents the color temperature value of the jth pixel point in the ith vertical area, and y i,j is the vertical coordinate value of the corresponding pixel point, and n i represents the number of pixels in the ith area, and ΔT i is the color temperature change amount between the ith area and the (i - 1)th area;

[0019] S103: According to the distribution of the color temperature trajectory residual values, call the regional color temperature trend sequence, compare the residual values of adjacent areas according to the set color temperature continuity judgment threshold. If the residual values between adjacent areas are all lower than the judgment threshold, mark them as continuous area paragraphs, integrate the area sequences that meet the conditions, and obtain the color temperature continuous paragraph.

[0020] As a further solution of the present invention, the steps for obtaining the brightness balance layer are as follows:

[0021] S201: Based on the continuous color temperature paragraphs, extract the RGB response values of all pixel points within the paragraph area. For each pixel, respectively call the original response values of the red, green, and blue channels, perform weighted combination to calculate the grayscale value, traverse all pixels within the area, perform processing operations point by point, record the grayscale value results of all pixel points, sum up all the grayscale values in the set and calculate the average value to obtain the overall brightness reference value within the area, and obtain the grayscale mean value;

[0022] S202: According to the grayscale mean value, call the set grayscale enhancement judgment interval to determine whether the grayscale mean value is outside the interval range. If it is less than the lower limit of the interval, it is determined that the overall brightness needs to be enhanced. For each pixel value in the grayscale set, call the preset grayscale enhancement coefficient to perform multiplication processing, proportionally enhance the original grayscale value and replace the original value, while retaining the original RGB structure, remap the updated grayscale value to all pixels in the image area, synchronously collect the updated pixel set, and generate a grayscale enhancement value sequence;

[0023] S203: Call the grayscale enhancement value sequence, perform color temperature recalculation operations on each pixel point, recalculate the color temperature value using RGB values, set the main color temperature reference value, calculate the difference between the enhanced color temperature and the main color temperature for each pixel point, and perform proportional operations in combination with the change amplitude of the RGB channels. Use the formula:

[0024]

[0025] Perform operations to obtain the RGB ratio trimming amount sequence, apply the ratios to the original RGB response values of the pixels respectively, perform brightness adjustment operations on the overall image channels, reconstruct the image data structure with the trimmed channel response values, and obtain the brightness balance layer;

[0026] Among them, represents the RGB trimming ratio of the kth pixel on channel γ, is the RGB response value of the kth pixel on channel γ after enhancement, β is the main color temperature conversion threshold parameter, and σ k is the color temperature deviation normalization value corresponding to the kth pixel.

[0027] As a further solution of the present invention, the steps for obtaining the peeled boundary map are as follows:

[0028] S301: Based on the brightness balance layer, construct multiple radially distributed chains at equal intervals in the outer edge area of the pier foundation. Select the center of the pier contour as the center of the circle, set multiple angular directions at a fixed pixel interval. Each angle corresponds to a pixel path extending outward from the center. Collect the grayscale values of all pixels on the chain path along each chain direction, construct a grayscale set arranged in pixel sequence, record the grayscale change data corresponding to all chains, and obtain the grayscale sequence set;

[0029] S302: Call the grayscale sequence set, perform differential calculation on the grayscale values of adjacent pixels on each chain respectively, calculate the difference between adjacent grayscale values as the grayscale change amount, accumulate all the grayscale change amounts and divide by the chain length to obtain the average grayscale change rate of the chain. After processing all the chains in sequence, form a grayscale rate data set and obtain a grayscale change rate sequence;

[0030] S303: According to the grayscale change rate sequence, select each pair of adjacent chains, judge the included angle formed by the grayscale change directions as vectors. If the included angle is lower than the set threshold, classify them into one group. Set the included angle judgment threshold to 15 degrees, and perform the included angle screening and grouping classification operation on each group of chains. The quantization process uses the formula:

[0031]

[0032] Perform operations to obtain a sequence of chain grayscale included angle comparison values. According to the judgment of whether it is less than the critical value η = 0.26, if it holds, mark it as the same group and establish a light decay block division structure;

[0033] Among them, Υ ab represents the comparison value of the grayscale direction vector included angle between the a-th and b-th chains, Θ a and Θ b respectively represent the main grayscale direction vector values of the a-th and b-th chains, Ψ a and Ψ b are respectively the standard deviations of the grayscale change amplitudes of the a-th and b-th chains, ∈ is a very small positive value to avoid the denominator being zero, and κ is the direction difference offset constant;

[0034] S304: According to the light decay block division structure, extract the number of texture boundary pixels and the total grayscale gradient value of the pixels included in each block, calculate the ratio of the number of texture boundaries to the grayscale gradient, call the ratio as the evaluation index for interference recognition, traverse the light decay blocks in sequence and record the corresponding ratios to generate a stripped boundary map.

[0035] As a further solution of the present invention, the steps for obtaining the disturbance comparison parameter group are as follows:

[0036] S401: Based on the stripped boundary map, extract the continuous frame coordinate sequence of the structural feature points at the connecting gaps of the pier components of the underwater structure of the bridge, obtain the coordinate change values of the same structural points at different time positions between consecutive frames, calculate the velocity vector and acceleration change amount of the structural points between adjacent frames, and obtain a structural dynamic change parameter group;

[0037] S402: Call the structural dynamic change parameter group, extract the disturbance direction change value and the corresponding frame position information of the continuous pixels in the interference area, construct a background disturbance direction and displacement data pair, calculate the formed vector set, and use the formula:

[0038]

[0039] Calculate to obtain the comparison value of the perturbation amplitude and the direction offset, combine with the structural dynamic change parameter group, map the structural trajectory to the background vector field coordinates, calculate the corresponding perturbation comparison difference, and establish the perturbation direction offset difference quantity;

[0040] Among them, Θ s represents the perturbation comparison difference of the s-th region, A e represents the velocity amplitude value of the e-th point in the structural region, φ e is the direction angle value, B r is the perturbation amplitude of the r-th point in the background region, ψ r is the perturbation direction angle, κ is the logarithm of the data in the structural region, and ζ is the logarithm of the data in the background region;

[0041] S403: According to the perturbation direction offset difference quantity, compare and analyze the combined values of the corresponding direction angles and velocity amplitudes of the structural region and the background region, identify the difference mutation points between the two, integrate the corresponding coordinate positions and parameter values, and obtain the perturbation comparison parameter group.

[0042] As a further solution of the present invention, the steps for obtaining the analysis result of the underwater trajectory image of the bridge are:

[0043] S501: Based on the perturbation comparison parameter group, extract the direction offset and velocity change data of the structural point set in consecutive frames, construct the direction change sequence and velocity change sequence of each structural point, calculate the comparison difference at the same time step of the two sequences, and combine them to form the direction-velocity joint trajectory data pair of each structural point, and generate the joint trajectory comparison unit set;

[0044] S502: Call the joint trajectory comparison unit set, extract the tension vector difference matrix between each pair of structural points, perform cross-analysis in combination with the perturbation response term, and use the formula:

[0045]

[0046] Calculate to obtain the perturbation coupling comparison value sequence, combine the sequence with the structural point region division index, mark the distribution characteristics of the structural trajectory, and obtain the tension perturbation response quantity group;

[0047] Among them, Ψ b represents the perturbation coupling comparison value of the b-th group of structural points, φ b is the number of structural point pairs in the b-th group, J b,c is the tension vector difference of the c-th pair of structural points in the b-th group, Υ b,c is the perturbation response value of the c-th pair of structural points in the b-th group, Θ b,cis the spatial projection difference of the structural point pairs within the same section, |J b,c | and |Υ b,c | are respectively the absolute value moduli of the tension vectors and the disturbance responses of the structural point pairs;

[0048] S503: According to the tension disturbance response quantity group, judge the coincidence degree of the velocity fluctuation area and the direction change area within the structural section, extract the abnormal point segments that meet the conditions of the structural disturbance synchronization characteristics and exceed the set trajectory direction mutation threshold, integrate the regional identification information, and obtain the analysis result of the underwater trajectory image of the bridge.

[0049] An intelligent analysis system for the underwater structure of a bridge pier, comprising:

[0050] The color temperature calculation module obtains the RGB channel response values of all pixels in the underwater structure image, performs the main color temperature conversion and generates a two-dimensional map in combination with the spatial coordinates, divides the vertical area, fits the main color temperature trajectory of the section, compares the fitting residual with the continuity threshold, and generates a color temperature continuous paragraph;

[0051] The grayscale adjustment module extracts the corresponding grayscale value set according to the color temperature continuous paragraph, sets the grayscale enhancement coefficient and performs linear adjustment, calculates the RGB channel ratio trimming amount in combination with the main color temperature difference value, and updates the RGB response value to obtain a brightness balance layer;

[0052] The chain grayscale analysis module sets radial chains in the outer edge area of the bridge pier based on the brightness balance layer, extracts the chain grayscale sequence, calculates the grayscale change rate and judges the chain angle, groups the chains and then delimits the light decay area, calculates the ratio of the texture boundary to the gradient, and generates a peeling boundary map;

[0053] The structural trajectory extraction module extracts the continuous frame coordinates of the structural feature points in the connection gap area according to the peeling boundary map, calculates the frame-to-frame velocity and acceleration change amounts, maps the trajectory coordinates to the disturbance vector field, extracts the direction offset angle and the velocity amplitude difference, and generates a disturbance comparison parameter group;

[0054] The disturbance feature recognition module extracts the structural trajectory pairs with direction offset and velocity change according to the disturbance comparison parameter group, calculates the difference value of the tension vectors between the trajectory points, calls the difference value and the disturbance response item to perform cross comparison, judges the coincidence degree of the velocity fluctuation area and the direction change area, and performs joint screening in combination with the synchronous disturbance characteristics and the trajectory direction mutation threshold to generate the analysis result of the underwater trajectory image of the bridge.

[0055] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0056] In the present invention, a two-dimensional map is constructed by fusing the main color temperature and spatial coordinates to enhance the expression of the spatial distribution of the structural region. The stable region is extracted by combining residual and continuity discrimination. The interval judgment and RGB trimming are introduced for gray-scale enhancement to strengthen the image brightness balance and color consistency. The gray-scale trend model is constructed by grouping radial chains to achieve the stripping of light decay interference. The continuous frame tracking of feature points is combined with the perturbation vector mapping to capture the dynamic changes of the structure. The abnormal behavior is identified by the cross-analysis of the tension difference and the response term, realizing the efficient recognition and interference isolation under multi-dimensional feature fusion. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 It is the main process flow chart of the present invention;

[0058] Figure 2 It is the flow chart of S1 of the present invention;

[0059] Figure 3 It is the flow chart of S2 of the present invention;

[0060] Figure 4 It is the flow chart of S3 of the present invention;

[0061] Figure 5 It is the flow chart of S4 of the present invention;

[0062] Figure 6 It is the flow chart of S5 of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0063] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0064] In the description of the present invention, it should be understood that the orientation or positional relationship indicated by the terms "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. is based on the orientation or positional relationship shown in the accompanying drawings, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as limiting the present invention. In addition, in the description of the present invention, "a plurality of" means two or more unless otherwise specifically defined.

[0065] Please refer to Figure 1 , an intelligent analysis method for the underwater structure image of a bridge pier, including the following steps:

[0066] S1: Obtain the RGB channel response values of all pixel points in the image area of the underwater structure of the bridge. Call the RGB channel values to perform the main color temperature conversion for each pixel, extract the combination of color temperature and spatial coordinates to generate a two-dimensional color temperature map. Divide the multi-level area in the vertical direction of the image, extract the main color temperature trajectories in each area and perform linear fitting. Compare the residual value with the color temperature continuity judgment threshold to generate color temperature continuous paragraphs;

[0067] S2: According to the color temperature continuous paragraphs, extract the set of gray values of all pixels. Call the gray mean value and compare it with the gray enhancement judgment interval, set the gray enhancement coefficient, perform linear enhancement on the gray values according to the set coefficient. Call the difference value between the color temperature of the enhanced pixels and the main color temperature to calculate the RGB ratio trimming amount, and adjust the RGB response values proportionally to obtain the brightness balance layer;

[0068] S3: Based on the brightness balance layer, set equidistant radial chains in the outer edge area of the pier foundation in the underwater structure of the bridge. Extract the gray value sequences of each chain and calculate the average gray change rate. Perform the judgment of the gray gradient angle between the chains and screen the chains with an included angle less than the threshold for grouping. Delimit the light decay blocks according to the grouping results. Call the ratio of the number of texture boundaries in the block to the gray gradient to calculate the interference recognition ratio, and generate the peeling boundary map;

[0069] S4: According to the peeling boundary map, extract the continuous frame coordinate sequence of the structural feature points at the connection gaps of the pier components in the underwater structure of the bridge. Calculate the velocity vector and the acceleration change amount for the position difference between consecutive frames. Call the disturbance direction field in the non-structural area of the image to construct the background disturbance vector set. Map the structural trajectory to the vector field, and extract the difference between the direction offset angle and the velocity amplitude to generate the disturbance comparison parameter group;

[0070] S5: According to the disturbance comparison parameter group, extract the trajectory pairs of direction offset and velocity change in the structural point set as comparison units. Call the cross-analysis of the difference matrix of the tension vectors between two points and the disturbance response terms to judge the coincidence degree of the velocity fluctuation area and the direction change area in the same section. Screen the abnormal behavior segments according to the structural disturbance synchronization characteristics and the trajectory direction mutation threshold to generate the analysis result of the underwater trajectory image of the bridge.

[0071] The color temperature continuous paragraphs include the regional main color temperature values, the fitting trajectory parameters, and the continuity discriminant labels. The brightness balance layer includes the gray enhanced pixel values, the RGB channel correction ratio, and the brightness uniform distribution characteristics. The peeling boundary map includes the light decay block range, the chain grouping results, and the distribution of the interference recognition ratio. The disturbance comparison parameter group includes the structural offset direction, the velocity change amplitude, and the disturbance response difference. The analysis result of the underwater trajectory image of the bridge includes the positions of the abnormal behavior segments, the trajectory direction mutation points, and the synchronous disturbance characteristics.

[0072] Please refer to Figure 2, the steps for obtaining the continuous color temperature segment are as follows:

[0073] S101: Based on obtaining the RGB channel response values of all pixel points in the image area of the underwater structure of the bridge, call the RGB channel values to perform the main color temperature conversion operation on each pixel, calculate the transformation relationship between the values of the pixel points in the R, G, and B channels and the color temperature, extract the transformed color temperature values in combination with the spatial coordinates where the pixels are located, obtain the corresponding combination of the color temperature and the image spatial coordinates, and generate a two-dimensional color temperature map;

[0074] A single-frame image with a resolution of 1920×1080 can be used as the input data. Map the image matrix to each pixel point according to the row and column indexes. Use an image reading tool to extract the R, G, and B channel values of each pixel point respectively. For example, the R channel value of the pixel point in the first row and the first column is 124, the G channel value is 131, and the B channel value is 136. Read the RGB values of all pixel points in the entire image in sequence to form a three-dimensional array. Subsequently, call the color temperature conversion operation on the RGB values of each pixel point. This operation is based on the conversion relationship between RGB and the color temperature. The empirical transformation formula T = 0.292R + 0.625G + 0.083G can be used. Substitute the example pixel point values to get T = 0.292×124 + 0.625×131 + 0.083×136 = 36.208 + 81.875 + 11.288 = 129.371, and the corresponding color temperature value is 129.37K. During the processing, it is necessary to traverse all pixels in the image and calculate the color temperature pixel by pixel. Combine its row and column indexes to determine the coordinate position of the pixel in the image space, and represent its spatial coordinate point with the index value. For example, if the color temperature of the pixel (1,1) is 129.37K, then it corresponds to the spatial coordinate (1,1) and the color temperature value 129.37K. Finally, combine the spatial coordinates and color temperature values of all pixel points to generate a two-dimensional data matrix with complete pixel coordinates and corresponding color temperature distributions as the two-dimensional color temperature map. Each element in this map consists of the spatial index position and the color temperature value of the pixel point in the image, with the complete color temperature response ability of the image area, which can be further used for subsequent vertical direction layering processing.

[0075] S102: Call the two-dimensional color temperature map, divide multiple hierarchical areas in the vertical direction of the image, extract the color temperature values of the corresponding pixels in each divided area, calculate the average change trend of the color temperature values of the pixel points in the area, and use the formula:

[0076]

[0077] Perform operations to obtain the change trend of the color temperature difference values between adjacent areas, use the color temperature trend of the area sequence as the trajectory data set, perform linear fitting processing, obtain the fitting residual data, and get the distribution of the color temperature trajectory residual values;

[0078] Among them, T i,jrepresents the color temperature value of the j-th pixel in the i-th vertical region, y i,j is the vertical coordinate value of the corresponding pixel, n i represents the number of pixels in the i-th region, ΔT i is the color temperature change amount between the i-th region and the (i - 1)-th region;

[0079] Call the two-dimensional color temperature map, divide multiple regions in the vertical direction of the image. The height of the image is 1080 pixels. Set every 120 pixels as a region unit, forming 2 regions numbered R1 to R2. Each region contains multiple pixels. The color temperature value corresponding to each pixel is denoted as Q r,u , and the vertical coordinate is denoted as Z r,u , during the execution, first select regions R1 and R2 for color temperature trend analysis. Extract the color temperature values of the pixels in region R1 as 5100, 5200, 5300, and the corresponding vertical coordinates are 10, 20, 30. The color temperature values of the pixels in region R2 are 5400, 5350, 5500, and the corresponding vertical coordinates are 10, 20, 30. According to the formula

[0080]

[0081] Perform the difference calculation of the weighted average color temperature value. First, calculate the weighted average color temperature value of region R1, Q1 = (5100×10 + 5200×20 + 5300×30)÷(10 + 20 + 30) = (51000 + 104000 + 159000)÷60 = 314000÷

[0082] 60 = 5233.33;

[0083] Calculate the weighted average color temperature value of region R2, Q2 = (5400×10 + 5350×20 + 5500×30)÷

[0084] (10 + 20 + 30) = (54000 + 107000 + 165000)÷60 = 326000÷60 = 5433.33. Substitute it into the formula to get Λ r = |5433.33 - 5233.33| = 200.00. Finally, obtain the inter-region color temperature weighted difference value sequence between region R1 and region R2 as 200.00. This value is used to evaluate the intensity of the color temperature change in the regions of the image. Slide the region window successively within the overall range of the image, and perform the above weighted average difference calculation process for each region one by one to obtain the continuous color temperature change trend curve between regions. This trend sequence is subsequently input into the linear fitting analysis module to extract the residual value between the fitting curve and the original sequence, forming a color temperature trajectory residual difference sequence for subsequent analysis.

[0085] S103: According to the distribution of color temperature trajectory residual values, call the regional color temperature trend sequence, compare the residual values of adjacent regions based on the set color temperature continuity judgment threshold. If the residual values between adjacent regions are all lower than the judgment threshold, mark them as continuous region paragraphs, integrate the region sequences that meet the conditions, and obtain the color temperature continuous paragraphs;

[0086] It is necessary to call the residual values between each pair of adjacent regions item by item. Set the color temperature continuity judgment threshold to 100K. This threshold is set based on the average amplitude of color temperature fluctuations between regions in the actual structure monitoring image, and is generally obtained through manual testing or historical calibration. If all the residual values within a certain region sequence are less than this threshold, it is considered that the color temperature change within this region sequence is stable, that is, it meets the color temperature continuous judgment condition. During the execution process, first traverse the residual sequence. For example, the residual sequence is [80, 90, 110, 95]. The differences between the first two regions are 80K and 90K respectively, both of which are lower than 100K. Therefore, they are marked as continuous region segments. The third difference is 110K, which is greater than the threshold of 100K and does not meet the condition, so the counting restarts. The fourth item is 95K, which meets the condition again, and continues to compare with subsequent regions. During the execution process, record the set of region numbers that meet the continuous condition to form a paragraph index sequence, and further generate the boundary positions of the continuous color temperature paragraphs through the combination of the start index and the end index. For example, if regions R1 to R2 meet the condition and R4 to R5 meet the condition, then the final generated paragraph is [(R1, R2), (R4, R5)]. This set is the color temperature continuous paragraph, which is used for subsequent clustering recognition or anomaly comparison in the structural image region.

[0087] Please refer to Figure 3 , and the steps for obtaining the brightness balance layer are as follows:

[0088] S201: Based on the color temperature continuous paragraphs, extract the RGB response values of all pixel points within the paragraph regions. For each pixel, call the original response values of the red, green, and blue channels respectively, perform weighted combination to calculate the gray value, traverse all pixels within the region, perform the processing operation point by point, record the gray value results of all pixel points, sum up all the gray values in the set and calculate the average value to obtain the overall brightness reference value within the region, and obtain the gray mean value;

[0089] Extract the set of RGB response values of all pixel points in the corresponding area of the paragraph in the image, perform a weighted calculation operation on each pixel point by point to obtain the grayscale value. The RGB response values respectively extract the numerical values of the red, green, and blue channels. In the actual underwater structure image of the bridge, assume that six pixel points are selected as the sampling objects, numbered P1 to P6 respectively. Their RGB channel values are set as P1 with R = 120, G = 140, B = 160, P2 with R = 130, G = 135, B = 150, P3 with R = 110, G = 125, B = 140, P4 with R = 100, G = 115, B = 130, P5 with R = 90, G = 105, B = 120, P6 with R = 80, G = 95, B = 110. Perform grayscale conversion operations on each pixel respectively, and obtain the grayscale values in the RGB weighted calculation method, which are P1: 0.299×120 + 0.587×140 + 0.114×160 = 136.3, P2 is 134.45, P3 is 123.85, P4 is 113.7, P5 is 103.55, P6 is 93.4. Form a set of the grayscale values of all pixels, and perform addition summation and averaging based on this set to obtain the grayscale mean value of (136.3 + 134.45 + 123.85 + 113.7 + 103.55 + 93.4) / 6 = 117.21. This value reflects the overall brightness level in the current continuous paragraph of color temperature and is used as a reference value for subsequent grayscale enhancement judgment to obtain the grayscale mean value.

[0090] S202: According to the grayscale mean value, call the set grayscale enhancement judgment interval to determine whether the grayscale mean value is outside the interval range. If it is less than the lower limit of the interval, it is determined that the overall brightness needs to be enhanced. Perform multiplication processing on each pixel value in the grayscale set by calling the preset grayscale enhancement coefficient, replace the original grayscale value after scaling it proportionally, and at the same time retain the original RGB structure. Remap the updated grayscale value to all pixels in the image area, and synchronously collect the updated pixel set to generate a grayscale enhancement value sequence;

[0091] According to the calculated grayscale mean value of 117.21, the preset grayscale enhancement judgment interval [100,180] is called for judgment. This interval is a standard reference interval set according to the common grayscale distribution in underwater images. If the grayscale mean value is lower than the lower limit of 100, the image brightness needs to be significantly enhanced. If it is higher than the upper limit of 180, it is determined that the brightness does not need to be enhanced. The current grayscale value of 117.21 is within the interval but biased towards the lower limit. Therefore, it is determined that the image can be appropriately enhanced. The grayscale enhancement coefficient is set to 1.2. This coefficient is derived from the actual experience of color level compensation. It means Grayscale enhancement should be magnified by 20% based on the original value, and the grayscale values ​​of P1 to P6 are enhanced in sequence. The results are P1: 136.3×1.2=163.56, P2 is 161.34, P3 is 148.62, P4 is 136.44, P5 is 124.26, and P6 is 112.08. All the enhanced grayscale values ​​are reconstructed into a pixel mapping table and mapped to the corresponding image pixel position information in sequence to form a complete image brightness enhancement structure, which is used for subsequent RGB ratio adjustment calculations to obtain a grayscale enhancement value sequence.

[0092] S203: Call the grayscale boost value sequence, perform a color temperature recalculation operation on each pixel, use the RGB value to recalculate the color temperature value, set the main color temperature reference value, calculate the difference between the boosted color temperature and the main color temperature for each pixel, and perform a proportional operation in combination with the change range of the RGB channel, using the formula:

[0093]

[0094] Obtain the RGB ratio trimming amount sequence by operation, apply the ratio to the original RGB response value of the pixel, perform brightness adjustment operation on the entire channel of the image, reconstruct the image data structure with the trimmed channel response value, and obtain the brightness balance layer;

[0095] in, represents the RGB trimming ratio of the k-th pixel on channel γ, is the RGB response value of the kth pixel on channel γ after enhancement, β is the main color temperature conversion threshold parameter, and σ k is the normalized value of the color temperature deviation corresponding to the kth pixel;

[0096] The RGB value of each pixel is recalculated to determine the degree of deviation between the color temperature value after the enhancement and the main color temperature. Assuming the main color temperature is 5200K, after the RGB channel response value of each pixel is updated, the new color temperature is calculated and the difference analysis is performed with 5200K. Taking the typical RGB enhanced value as an example, if the RGB value of a pixel after enhancement is R=180, G=195, B=210, the corresponding color temperature calculation formula is:

[0097] T' = 0.299×R + 0.587×G + 0.114×B;

[0098] Substitute specific values for calculation:

[0099] T' = 0.299×180 + 0.587×195 + 0.114×210 = 53.82 + 114.465 + 23.94 = 192.225;

[0100] Next, calculate the deviation by comparing the color temperature difference value with the main color temperature of 5200K:

[0101] ΔT = |T' - 5200| = |192.225 - 5200| = 5007.775;

[0102] Use the color temperature difference value to calculate the RGB ratio trimming amount, and the trimming formula is:

[0103]

[0104] Among them, the parameter represents the RGB trimming ratio on channel γ, represents the enhanced RGB response value, β is the main color temperature conversion threshold parameter, set to 3200, and σ k is the normalized color temperature deviation value, set to 0.12.

[0105] For the R channel in the example:

[0106]

[0107] This trimming ratio is used to adjust the RGB values of pixels. Apply this ratio to adjust the response values of each channel to generate new RGB response values with a color temperature closer to the main color temperature. After this process is completed, remap these trimmed pixel values into the image to construct a luminance balance layer with a more reasonable luminance and color temperature balance.

[0108] Please refer to Figure 4 , the steps to obtain the stripped boundary map are as follows:

[0109] S301: Based on the luminance balance layer, construct multiple equidistantly distributed radial chains in the outer edge area of the pier foundation. Select the center of the pier contour as the center of the circle, set multiple angular directions at a fixed pixel interval. Each angle corresponds to a pixel path extending outward from the center. Collect all pixel gray values on the chain path along each chain direction, construct a gray value set arranged in pixel sequence, record the gray value change data corresponding to all chains, and obtain the gray value sequence set;

[0110] Equidistant chains are set along the radial direction in the outer edge area of the pier foundation in the image. The center point of the image is used as the central coordinate of the pier structure, and the angular step is set to 15 degrees. 24 radial paths are generated between 0 and 360 degrees. Each path samples pixel points along the radius direction. In a scenario with an image resolution of 1920×1080, the chain length is set to 50 pixels, and sampling is performed at a 1-pixel step. After reading the gray values of the corresponding pixel points on each chain, a gray value sequence set is constructed. For example, the pixel gray value sequence contained in chain L1 is [122, 125, 127, 129, 130, 132, 134, 135, 136, 137]. The gray data of all chains are extracted in turn to generate a gray sequence set.

[0111] S302: Call the gray sequence set, and perform differential calculations on the adjacent pixel gray values on each chain respectively. Calculate the difference between adjacent gray values as the gray change amount. After accumulating all the gray change amounts and dividing by the chain length, it is used as the gray average change rate of the chain. After processing all the chains in turn, a gray rate data set is formed, and a gray change rate sequence is obtained;

[0112] Perform gray change rate calculation on each chain. Obtain the average rate by summing the gray differences between adjacent pixels within the chain and dividing by the number of chain pixels. For example, the gray sequence of chain L1 is [122, 125, 127, 129, 130, 132, 134, 135, 136, 137]. The calculated difference sequence is [3, 2, 2, 1, 2, 2, 1, 1, 1]. The sum of the differences is 15, the number of chain pixels is 10, and the gray change rate is 15÷10 = 1.5. After performing the same calculation on all chains, a rate sequence is formed, and a gray change rate sequence is obtained.

[0113] S303: According to the gray change rate sequence, select each pair of adjacent chains, and judge the included angle of the vector formed by the gray change directions. If the included angle is lower than the set threshold, they are grouped into one group. The included angle judgment threshold is set to 15 degrees. Perform included angle screening and grouping classification operations on each group of chains. The quantization process uses the formula:

[0114]

[0115] Calculate to obtain the chain gray included angle comparison value sequence, and judge whether it is less than the critical value η = 0.26. If it holds, mark it as the same group and establish a light decay block division structure;

[0116] Among them, Υ ab represents the gray direction vector included angle comparison value between the a-th and b-th chains, Θ a and Θ b represent the main gray direction vector values of the a-th and b-th chains respectively, Ψ a and Ψ bThey are the standard deviations of the gray - scale amplitude variations of the a - th and b - th chains respectively. ∈ is a very small positive value to avoid a zero denominator, and κ is the direction - difference offset constant;

[0117] Extract the gray - scale direction vectors between any two adjacent chains for angle judgment. Set the main gray - scale direction value and the amplitude - variation standard deviation of each chain as the vector direction and the modulus - length basis respectively. Suppose the main direction values of chain a and chain b are 0.85 and 0.90 respectively, and the standard deviations are 0.13 and 0.15 respectively. Introduce a minimum value ∈ = 0.0001 to prevent a zero denominator, and an offset coefficient κ = 0.5, and substitute them into the formula

[0118]

[0119] Perform the calculation to obtain;

[0120]

[0121] The obtained angle - comparison value is 2.225, which is less than the grouping - determination critical value η = 2.5. Therefore, chain a and chain b are determined to be in the same group, and an optical - decay block - division structure is generated.

[0122] S304: According to the optical - decay block - division structure, extract the number of texture - boundary pixels and the total gray - scale gradient of the pixels included in each block, calculate the ratio of the texture - boundary number to the gray - scale gradient, call the ratio as the evaluation index for interference recognition, traverse the optical - decay blocks in sequence and record the corresponding ratios, and generate a stripped - boundary map;

[0123] Perform the extraction of the number of texture - boundary pixels and the summation of the gray - scale gradient for each pixel point in each block. Use an image - gradient operator to calculate the difference values in the horizontal and vertical directions for each pixel and then take the average to obtain the gray - scale gradient value. Judge whether there is a situation where the gray - scale mutation of adjacent pixels is greater than 20 for each pixel. If it exists, record it as a boundary pixel, count the number of boundary pixels in the entire block, and record the total gray - scale gradient value. For example, an optical - decay block contains 27 boundary pixels and the total gray - scale gradient is 3510. Calculate its interference - recognition ratio as 27÷3510≈0.0077. Repeat the calculation process for all blocks, form a set of interference - index corresponding to all blocks in the pixel space, and establish a stripped - boundary map.

[0124] Please refer to Figure 5 , and the steps for obtaining the perturbation - comparison parameter group are as follows:

[0125] S401: Based on the stripped - boundary map, extract the continuous - frame coordinate sequence of structural feature points at the connecting seams of pier components of the underwater structure of the bridge, obtain the coordinate change values of the same structural points at different time positions between consecutive frames, calculate the velocity vector and the acceleration change amount of the structural points between adjacent frames, and obtain the structural dynamic - change parameter group;

[0126] Extract the continuous frame coordinate sequence of structural feature points at the connecting gap of pier components in the underwater structure of the bridge. In actual operation, first, it is necessary to clarify the recognizable area of pier components in the underwater environment, especially the location of the connecting gap. Obtain a continuous image frame sequence through an underwater imaging device, and use boundary extraction means to locate the edges of components in the image. For example, for a video image sequence of 20 seconds with a sampling frequency of 5 frames per second, 100 frames of images can be obtained. 4 feature points are extracted from each frame of image, and a total of 8 tracking points are set on both sides of the gap corresponding to the pier. Establish a numbering system and label them as P1 to P8. Then record the spatial coordinates of each feature point in the continuous frames. Set the coordinates of each point in the t-th frame as (X t , Y t ). By calculating the coordinate changes of the point positions between adjacent frames, use ΔX = X t+1 - X t , ΔY = Y t+1 - Y t . Further derive the velocity vector V t = √(ΔX 2 + ΔY 2 ) / Δt. Calculate the average velocity of all tracking points to obtain the velocity change range. For example, the velocity sequence of P1 within 5 frames is [0.023, 0.028, 0.031, 0.027] m / s. Then calculate the acceleration change amount A t = (V t+1 - V t ) / Δt. If Δt = 0.2 seconds, the acceleration sequence of P1 is [0.025, 0.015, -0.02] m / s 2 . Process all feature points according to the above steps to obtain the motion trend of each point in the continuous frames. Finally, merge the velocity vectors and acceleration amounts of all points to obtain the structural dynamic change parameter group.

[0127] S402: Call the perturbation direction field in the unstructured area of the image, extract the perturbation direction change values of continuous pixels and the corresponding frame position information in the interference area, construct the background perturbation direction and displacement data pairs, calculate the formed vector set, and use the formula:

[0128]

[0129] Operate to obtain the perturbation amplitude and direction offset comparison value. Combine the structural dynamic change parameter group, map the structural trajectory to the background vector field coordinates, calculate the corresponding perturbation comparison difference, and establish the perturbation direction offset difference quantity;

[0130] Among them, Θ s represents the perturbation comparison difference of the s-th area, A eRepresents the velocity amplitude value of the e-th point in the structural region, φ e Is the direction angle value, B r Is the perturbation amplitude of the r-th point in the background region, ψ r Is the perturbation direction angle, κ is the logarithm of the data in the structural region, and ζ is the logarithm of the data in the background region;

[0131] Call the perturbation direction field in the unstructured region of the image, extract the perturbation direction change value and the corresponding frame position information of the continuous pixels in the interference region. In the specific implementation process, first, a masking operation needs to be performed on the unstructured part of the image to exclude the pixels corresponding to the pier components and the connection gap region. Then, a filter is used to extract the direction gradient of the rest of the image. The perturbation direction is defined as the pixel gray gradient direction, and the unit angle is represented by the counterclockwise 0-360° notation. For example, in the 10th frame, the direction angles of 5 pixels in a certain interference block in the unstructured region are 31°, 34°, 32°, 29°, and 30° respectively. The weighted average method is used to calculate the regional direction, and the average perturbation direction is 31.2°. Map this direction data to the pixel coordinates and extract the perturbation amplitude of the interference region. The perturbation amplitude is calculated by the absolute value of the gray difference between the current pixel value and the pixel value of the previous frame. For example, the gray difference between a certain pixel in the 11th frame and the 10th frame is 12, and the normalized perturbation value is 12 / 255≈0.047. Combining the direction angle of this perturbation point, the background perturbation vector B can be formed r , and then match the velocity amplitude value and the direction angle in the structural region, using the formula:

[0132]

[0133] Among them, the structural velocity amplitude A e And the direction angle φ e Come from the structural dynamic change parameter group extracted in Paragraph 1. The background perturbation velocity amplitude B r And the direction ψ r Are generated from the perturbation information in the unstructured region. Set 5 groups of structural perturbation data as A e =[0.18, 0.22, 0.19, 0.21, 0.20], φ e =[35, 40, 38, 42, 37], corresponding to the background perturbation B r =[0.09, 0.12, 0.08, 0.11, 0.10], ψ r =[33, 39, 36, 43, 35], then:

[0134] ∑(A e ·φ e ) = 38.54, ∑(B r ·ψ r ) = 18.76;

[0135]

[0136] This value is the contrast difference between the structural perturbation and the background perturbation, which is used to quantify the offset degree of the structural perturbation relative to the background perturbation and establish the perturbation direction offset difference quantity.

[0137] S403: According to the perturbation direction offset difference quantity, compare and analyze the combined values of the corresponding direction angles and velocity amplitudes of the structural region and the background region, identify the mutation points of the difference between the two, integrate the corresponding coordinate positions and parameter values, and obtain the perturbation comparison parameter group;

[0138] Compare and analyze the combined values of the corresponding direction angles and velocity amplitudes of the structural region and the background region respectively. First, obtain the combined terms of the perturbation vectors in the structural region and the background region in the same time frame, and construct the difference sequence Δθ k =|φ k -ψ k |, ΔV k =|A k -B k |. For example, at the 12th frame, the structural perturbation φ k is 40°, and the background perturbation ψ k is 36°, then Δθ k =4°. The structural velocity amplitude is 0.22 m / s, and the background perturbation amplitude is 0.12 m / s, then ΔV k =0.10 m / s. Repeat the above process for all matching pairs within the region, and set the judgment reference value as the perturbation direction difference threshold θ t =6° and the velocity difference threshold V t =0.08 m / s. Any item where Δθ k >θ t and ΔV k >V t is marked as an offset perturbation point. Continue to extract the spatial coordinate values corresponding to these points, and form a data tuple with the coordinates, their direction offset amounts, and velocity differences. For example, the coordinates of a certain point are (x = 123, y = 56), its Δθ = 7.3°, and ΔV = 0.095 m / s, then the perturbation parameter of this point is (123, 56, 7.3, 0.095). Record all abnormal perturbation points in sequence, and finally obtain the perturbation comparison parameter group.

[0139] Please refer to Figure 6 , and the steps to obtain the analysis result of the underwater trajectory image of the bridge are as follows:

[0140] S501: Based on the disturbance comparison parameter group, extract the data of the direction offset and velocity change of the points in the structural point set in consecutive frames, construct the direction change sequence and velocity change sequence for each structural point, calculate the comparison difference at the same time step of the two sequences, and combine them to form the direction-velocity joint trajectory data pair for each structural point, and generate the joint trajectory comparison unit set;

[0141] First, arrange the image sequence in frame numbers, and sequentially select the structural point sets in consecutive frames. For each structural point, by extracting the displacement vector between the current frame and the previous frame, the direction offset is obtained, which is represented by the coordinate difference of the structural point in the coordinate system. The direction calculation is quantified by the angle between the motion vector of the point and the reference axis, and the direction unit is in degrees. Further, through the horizontal and vertical coordinate differences and time interval of the structural point in consecutive frames, the velocity change value is calculated, with the velocity unit of mm / s and the direction unit of degrees. The velocity change amount is calculated by the instantaneous velocity difference between the latter frame and the previous frame. Each structural point forms a set of direction sequences and velocity sequences within multiple frames. Arrange the two sequences in chronological order and store them as a trajectory pair. After obtaining the direction-velocity trajectory combination, select the structural point group in a specific area of the lower bridge structure, such as 8 points numbered S101 to S108. The structural points at the corresponding positions construct direction sequences such as 15.3, 14.8, 13.7, 17.2, etc., with the unit of degrees, and velocity sequences such as 12.4, 10.6, 11.2, 13.5. Pair the two sequences in chronological order as a trajectory combination and uniformly label it as a trajectory comparison unit. The sampling time interval of the structural point set is 0.5 seconds, and a total of 10 frames of data are selected for sequence combination. After trajectory generation, index and label according to the structural point number and its spatial coordinates to complete the construction of the direction-velocity joint trajectory of different structural points and obtain the joint trajectory comparison unit set.

[0142] S502: Call the joint trajectory comparison unit set, extract the tension vector difference matrix between each pair of structural points, perform cross-analysis in combination with the disturbance response term, and use the formula:

[0143]

[0144] Perform operations to obtain the disturbance coupling comparison value sequence, combine the sequence with the structural point area division index, and label the structural trajectory distribution characteristics to obtain the tension disturbance response quantity group;

[0145] Among them, Ψ b represents the disturbance coupling comparison value of the b-th group of structural points, φ b is the number of structural point pairs in the b-th group, J b,c is the tension vector difference of the c-th pair of structural points in the b-th group, Υ b,c is the disturbance response value of the c-th pair of structural points in the b-th group, Θ b,cis the spatial projection difference of the structural point pairs within the same section, |J b,c | and |Υ b,c | are respectively the absolute value moduli of the tension vector and the perturbation response of the structural point pairs;

[0146] First, determine the structural point grouping sequence according to the comparison of the structural point numbers, and perform paired mapping on the trajectory paths within each structural point grouping. The tension vector difference sequence J b,c 、the perturbation response value sequence Υ b,c , and the spatial projection difference Θ b,c between the structural points are recorded in each paired mapping. Here, the tension vector difference sequence J b,c is the vector offset difference of the corresponding positions of the structural point pairs in consecutive frame images. The specific calculation method is to calculate the Euclidean distance after differentiating the coordinate values of the c-th pair of points in the b-th group in the front and back frames, and then perform difference processing. The perturbation response value Υ b,c is derived from the velocity fluctuation response term of the corresponding structural points under the perturbation comparison parameter group. The average velocity variance is obtained by setting the instantaneous velocity fluctuation interval within the previous and subsequent 5 frames. The spatial projection difference Θ b,c represents the normal distribution difference of the structural point pairs in the local coordinate system. The calculation formula is the absolute difference of the two points in the normal component. In the example, 4 groups of structural point data are selected, namely the point pairs (P1, P2), (P3, P4), (P5, P6), (P7, P8), and their corresponding parameters are J =

[0147] 2.5, 3.1, 2.8, 3.5, Υ = 1.2, 1.7, 1.4, 2.0, Θ = 0.8, 1.1, 0.9, 1.3. Substitute into the formula:

[0148]

[0149] The judgment reference value Ψ th is set through statistics. Based on the historical interference sample data, a 95% confidence interval is constructed, with a mean of 0.59 and a standard deviation of 0.05. The calculated upper confidence limit is 0.687. Therefore, the threshold Ψ th = 0.65 is set for classification judgment.

[0150] 0.673 > 0.65, so the current structural point group outputs a high interference feature, and a tension perturbation response quantity group is established.

[0151] S503: According to the tension perturbation response quantity group, judge the coincidence degree of the velocity fluctuation area and the direction change area within the structural section, extract the abnormal point segments that meet the conditions of the structural perturbation synchronization characteristics and exceed the set trajectory direction mutation threshold, integrate the regional identification information, and obtain the analysis result of the underwater trajectory image of the bridge;

[0152] First, divide the structural section into a sequence of fixed-length sections, for example, each section is 10m for spatial index coding, and they are respectively labeled as A1, A2, A3, etc. Each section corresponds to a set of structural point group numbers. For the structural points within each structural section, count the speed fluctuation areas and direction change areas to which they belong. The speed fluctuation areas are marked as disturbance points by setting a speed variance threshold. Set the speed fluctuation variance threshold to 20. The direction change areas are judged by the angle offset mutation threshold. Set the direction mutation angle to 10 degrees. Using whether it exceeds the two thresholds as the judgment basis, extract the point sets that meet the conditions and compare the coincidence rate. The coincidence rate is calculated as the proportion of the intersection of the two sets in the total number of structural points. If the number of intersection points is 6 and the total number of points is 10, the coincidence rate is 0.6. Further, call the trajectory direction mutation threshold setting value of 12 degrees, and perform a sliding window detection on the point trajectory direction sequence within the determined coincidence section. Use the average value of the angles between the front and rear two frames for jump identification. If it exceeds the threshold continuously twice, this point is marked as a mutation abnormal point. Integrate the section number and the structural point number to output an abnormal behavior section list. Finally, complete the coincidence determination and mutation screening of all structural sections to obtain the analysis result of the underwater trajectory image of the bridge.

[0153] An intelligent analysis system for the underwater structure of a bridge pier, comprising:

[0154] The color temperature calculation module obtains the RGB channel response values of all pixels in the underwater structure image, performs the main color temperature conversion and generates a two-dimensional map in combination with the spatial coordinates, divides the vertical area, fits the main color temperature trajectory of the section, compares the fitting residual with the continuity threshold, and generates a color temperature continuous paragraph;

[0155] The gray scale adjustment module extracts the corresponding gray scale value set according to the color temperature continuous paragraph, sets the gray scale enhancement coefficient and performs linear adjustment, calculates the RGB channel ratio trimming amount in combination with the main color temperature difference value, and updates the RGB response value to obtain a brightness balance layer;

[0156] The chain gray scale analysis module, based on the brightness balance layer, sets a radial chain in the outer edge area of the bridge pier, extracts the chain gray scale sequence, calculates the gray scale change rate and judges the chain angle, groups the chains and then delimits the light decay area, calculates the ratio of the texture boundary to the gradient, and generates a peeling boundary map;

[0157] The structural trajectory extraction module, according to the peeling boundary map, extracts the continuous frame coordinates of the structural feature points in the connection gap area, calculates the frame-to-frame speed and acceleration change amounts, maps the trajectory coordinates to the disturbance vector field, extracts the direction offset angle and the speed amplitude difference value, and generates a disturbance comparison parameter group;

[0158] The disturbance feature recognition module extracts structural trajectory pairs with direction offsets and speed changes according to the disturbance comparison parameter group, calculates the difference in tension vectors between trajectory points, calls for cross-comparison between the difference and the disturbance response term, determines the degree of overlap between the speed fluctuation area and the direction change area, and performs joint screening in combination with the synchronous disturbance characteristics and the trajectory direction mutation threshold to generate the analysis result of the underwater trajectory image of the bridge.

[0159] The above are only the preferred embodiments of the present invention, and are not intended to limit the present invention in other forms. Any person skilled in the art may use the technical content disclosed above to make changes or modifications into equivalent embodiments with equivalent changes and apply them to other fields. However, as long as the technical solution content of the present invention is not departed from, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention still fall within the protection scope of the technical solution of the present invention.

Claims

1. An intelligent analysis method for the underwater structure image of a bridge pier, characterized in that, It includes the following steps: S1: Obtain the RGB channel values of all pixels in the image area of the underwater structure of the bridge, convert the main color temperature, generate a color temperature map, divide the area vertically, extract the main color temperature trajectory and perform linear fitting, compare the residual and the continuity threshold, and mark the continuous color temperature paragraphs; S2: Extract the gray values based on the continuous paragraphs, set a boosting coefficient to linearly enhance the gray scale, calculate the RGB trimming amount, and generate a brightness balance layer; S3: Based on the brightness balance layer, set equidistant radial chains at the outer edge of the pier foundation, extract the gray scale sequence, calculate the gray scale change rate, screen and group the chains with an included angle less than the threshold, and generate a peeling boundary map according to the ratio of the number of texture boundaries to the gray scale gradient; S4: According to the peeling boundary map, extract the continuous frame coordinates of the structural feature points at the component gaps, calculate the velocity vector and the acceleration change, combine the disturbance vector field in the non-structural area, extract the direction offset angle and the velocity amplitude difference, and generate a disturbance comparison parameter group; S5: According to the disturbance comparison parameters, extract the pairs of the direction offset and the velocity change trajectory of the structural points, analyze the coincidence degree of the tension difference and the disturbance response, screen the abnormal segments with synchronous disturbance and trajectory mutation, and output the analysis result of the underwater trajectory image of the bridge.

2. The intelligent analysis method for the underwater structure image of a pier according to claim 1, wherein The continuous color temperature paragraphs include the regional main color temperature value, the fitting trajectory parameters, and the continuity discrimination label. The brightness balance layer includes the gray scale enhanced pixel values, the RGB channel correction ratio, and the brightness uniform distribution characteristics. The peeling boundary map includes the light decay block range, the chain grouping result, and the interference recognition ratio distribution. The disturbance comparison parameter group includes the structural offset direction, the velocity change amplitude, and the disturbance response difference. The analysis result of the underwater trajectory image of the bridge includes the position of the abnormal behavior segment, the trajectory direction mutation point, and the synchronous disturbance characteristics.

3. The intelligent analysis method for the underwater structure image of a pier according to claim 2, wherein The steps for obtaining the continuous color temperature paragraphs are as follows: S101: Based on the RGB channel response values of all pixel points in the obtained image area of the underwater structure of the bridge, call the RGB channel values to perform the main color temperature conversion operation on each pixel, calculate the transformation relationship between the values of the pixel points in the R, G, and B channels and the color temperature, extract the transformed color temperature value in combination with the spatial coordinates where the pixel is located, obtain the corresponding combination of the color temperature and the image spatial coordinates, and generate a two-dimensional color temperature map; S102: Call the two-dimensional color temperature map, divide multiple hierarchical areas in the vertical direction of the image, extract the color temperature values of the corresponding pixels in each divided area, calculate the average change trend of the color temperature values of the pixel points in the area, and use the formula: Perform the operation to obtain the change trend of the color temperature difference between adjacent areas, use the color temperature trend of the area sequence as the trajectory data set, perform linear fitting processing, obtain the fitting residual data, and obtain the distribution of the color temperature trajectory residual values; Among them, T i,j represents the color temperature value of the j-th pixel point in the i-th vertical area, y i,j is the vertical coordinate value of the corresponding pixel point, n i represents the number of pixels in the i-th area, ΔT i is the color temperature change amount between the i-th area and the (i - 1)-th area; S103: According to the distribution of the color temperature trajectory residual values, call the regional color temperature trend sequence, compare the residual values of adjacent areas according to the set color temperature continuity judgment threshold. If the residual values between adjacent areas are all lower than the judgment threshold, mark them as continuous regional paragraphs, integrate the area sequences that meet the conditions, and obtain the continuous color temperature paragraphs.

4. The intelligent analysis method for the underwater structure image of a pier according to claim 3, wherein, The steps for obtaining the brightness balance layer are as follows: S201: Based on the continuous color temperature paragraphs, extract the RGB response values of all pixel points within the paragraph area. For each pixel, separately call the original response values of the red, green, and blue channels, perform weighted combination to calculate the grayscale value, traverse all pixels within the area, perform the processing operation point by point, record the grayscale value results of all pixel points, sum up all the grayscale values in the set and calculate the average value to obtain the overall brightness reference value within the area, and obtain the grayscale mean value; S202: According to the grayscale mean value, call the set grayscale enhancement judgment interval to determine whether the grayscale mean value is outside the interval range. If it is less than the lower limit of the interval, it is determined that the overall brightness needs to be enhanced. For each pixel value in the grayscale set, call the preset grayscale enhancement coefficient to perform multiplication processing, proportionally enhance the original grayscale value and replace the original value, while retaining the original RGB structure, remap the updated grayscale value to all pixels in the image area, synchronously collect the updated pixel set, and generate a grayscale enhancement value sequence; S203: Call the grayscale enhancement value sequence, perform color temperature recalculation operation on each pixel point, recalculate the color temperature value using the RGB values, set the main color temperature reference value, calculate the difference between the enhanced color temperature and the main color temperature for each pixel point, and perform proportional operation in combination with the change amplitude of the RGB channels. Use the formula: Perform the operation to obtain the RGB ratio trimming amount sequence, apply the ratios to the original RGB response values of the pixels respectively, perform brightness adjustment operation on the overall image channels, reconstruct the image data structure with the trimmed channel response values, and obtain the brightness balance layer; Among them, represents the RGB trimming ratio of the k-th pixel on channel γ, is the RGB response value of the k-th pixel on channel γ after enhancement, β is the main color temperature conversion threshold parameter, and σ k is the color temperature deviation normalization value corresponding to the k-th pixel.

5. The intelligent analysis method for the underwater structure image of a pier according to claim 4, characterized in that The steps for obtaining the peeled boundary map are as follows: S301: Based on the brightness balance layer, construct multiple equally spaced radial chains in the outer edge area of the pier foundation. Select the center of the pier contour as the center of the circle, set multiple angular directions at a fixed pixel interval. Each angle corresponds to a pixel path extending outward from the center. Along each chain direction, collect the grayscale values of all pixels on the chain path, construct a grayscale set arranged in pixel sequence, record the grayscale change data corresponding to all chains, and obtain the grayscale sequence set; S302: Call the grayscale sequence set, perform differential calculation on the adjacent pixel grayscale values of each chain respectively, calculate the difference between the adjacent grayscale values as the grayscale change amount, sum up all the grayscale change amounts and divide by the length of the chain as the grayscale average change rate of the chain. After processing all chains in turn, form a grayscale rate data set and obtain the grayscale change rate sequence; S303: According to the grayscale change rate sequence, select each pair of adjacent chains, judge the vector included angle formed by the grayscale change directions. If the included angle is lower than the set threshold, classify them into one group. Set the included angle judgment threshold as 15 degrees, perform included angle screening and grouping classification operations on each group of chains. The quantization process uses the formula: Perform the operation to obtain the chain grayscale included angle comparison value sequence, and judge whether it is less than the critical value η = 0.

26. If it holds, mark it as the same group and establish a light decay block division structure; Among them, Υ ab represents the comparison value of the gray - scale direction vector angle between the a - th and b - th chains, Θ a , Θ b respectively represent the principal gray - scale direction vector values of the a - th and b - th chains, Ψ a , Ψ b are respectively the standard deviations of the gray - scale amplitude variations of the a - th and b - th chains, ∈ is a very small positive value to avoid the denominator being zero, and κ is the direction - difference offset constant; S304: According to the light attenuation block division structure, for each block, extract the number of texture boundary pixels and the total gray - scale gradient value in the pixels, calculate the ratio of the texture boundary number to the gray - scale gradient, use the ratio as the evaluation index for interference recognition, traverse the light attenuation blocks in sequence and record the corresponding ratios, and generate a peeled boundary map.

6. The intelligent analysis method for the underwater structure image of a bridge pier according to claim 5, characterized in that The steps for obtaining the perturbation comparison parameter group are as follows: S401: Based on the peeled boundary map, extract the continuous - frame coordinate sequence of structural feature points at the connecting seams of pier components of the underwater structure of the bridge, obtain the coordinate change values of the same structural points at different time positions between consecutive frames, calculate the velocity vector and acceleration change amount of the structural points between adjacent frames, and obtain the structural dynamic change parameter group; S402: Call the structural dynamic change parameter group, extract the perturbation direction change value and the corresponding frame position information of consecutive pixels in the interference area, construct the background perturbation direction and displacement data pair, calculate the formed vector set, and use the formula: Perform operations to obtain the perturbation amplitude and direction offset comparison value, combine with the structural dynamic change parameter group, map the structural trajectory to the background vector field coordinates, calculate the perturbation comparison difference correspondingly, and establish the perturbation direction offset difference quantity; Among them, Θ s represents the disturbance contrast difference of the sth region, A e represents the velocity amplitude value of the eth point in the structural region, φ e is the direction angle value, B r is the disturbance amplitude of the rth point in the background region, ψ r is the disturbance direction angle, κ is the logarithm of the data in the structural region, and ζ is the logarithm of the data in the background region; S403: According to the perturbation direction offset difference quantity, perform a comparative analysis on the combined values of the self - corresponding direction angle and velocity amplitude of the structural area and the background area, identify the mutation points of the difference between the two, integrate the corresponding coordinate positions and parameter values, and obtain the perturbation comparison parameter group.

7. The intelligent analysis method for the underwater structure image of a pier according to claim 6, characterized in that, The steps for obtaining the analysis result of the underwater trajectory image of the bridge are as follows: S501: Based on the perturbation comparison parameter group, extract the direction offset and velocity change data of the structural point set in consecutive frames, construct the direction change sequence and velocity change sequence of each structural point, calculate the comparison difference of the two sequences at the same time step, combine to form the direction - velocity joint trajectory data pair of each structural point, and generate the joint trajectory comparison unit set; S502: Call the joint trajectory comparison unit set, extract the difference matrix of the tension vectors between each pair of structural points, perform cross - analysis in combination with the perturbation response term, and use the formula: Perform operations to obtain the perturbation coupling comparison value sequence, combine the sequence with the structural point area division index, mark the distribution characteristics of the structural trajectory, and obtain the tension perturbation response quantity group; Among them, Ψ b represents the perturbation coupling contrast value of the b-th group of structural points, φ b is the number of structural point pairs in the b-th group, J b,c is the difference in the tension vectors of the c-th pair of structural points in the b-th group, Υ b,c is the perturbation response value of the c-th pair of structural points in the b-th group, Θ b,c is the spatial projection difference of the structural point pair within the same section, |J b,c | and |Υ b,c | are respectively the absolute value moduli of the tension vector and the perturbation response of the structural point pair; S503: According to the tension perturbation response quantity group, judge the coincidence degree of the velocity fluctuation area and the direction change area in the structural section, extract the abnormal point segments that meet the conditions of the structural perturbation synchronization characteristic and exceed the set trajectory direction mutation threshold, integrate the regional identification information, and obtain the analysis result of the underwater trajectory image of the bridge.

8. An intelligent image analysis system for underwater structure of bridge pier, characterized in that, The system is used to execute an intelligent analysis method for the underwater structure image of a pier according to any one of claims 1 - 7, including: The color temperature calculation module obtains the RGB channel response values of all pixels in the underwater structure image, performs the main color temperature conversion and generates a two - dimensional map in combination with the spatial coordinates, divides the vertical area, fits the main color temperature trajectory of the section, compares the fitting residual with the continuity threshold, and generates the color temperature continuous paragraph; The grayscale adjustment module extracts the corresponding grayscale value set according to the continuous paragraph of color temperature, sets the grayscale enhancement coefficient and performs linear adjustment, calculates the RGB channel ratio trimming amount in combination with the main color temperature difference value, updates the RGB response value, and obtains the brightness balance layer; The chain grayscale analysis module, based on the brightness balance layer, sets radial chains in the outer edge area of the bridge pier, extracts the chain grayscale sequence, calculates the grayscale change rate and judges the chain angle, groups the chains and then delimits the light decay area, calculates the ratio of the texture boundary to the gradient, and generates the peeling boundary map; The structure trajectory extraction module, according to the peeling boundary map, extracts the continuous frame coordinates of the structure feature points in the connection gap area, calculates the change amount of the inter-frame speed and acceleration, maps the trajectory coordinates to the perturbation vector field, extracts the direction offset angle and the speed amplitude difference, and generates the perturbation comparison parameter group; The perturbation feature recognition module, according to the perturbation comparison parameter group, extracts the structure trajectory pairs with direction offset and speed change, calculates the difference of the tension vectors between the trajectory points, calls the difference and the perturbation response term to perform cross comparison, judges the coincidence degree of the speed fluctuation area and the direction change area, and performs joint screening in combination with the synchronous perturbation characteristics and the trajectory direction mutation threshold to generate the analysis result of the underwater trajectory image of the bridge.

Images