Bridge differential settlement assessment method based on SAR image

Abstract

The invention discloses a bridge differential settlement assessment method based on an SAR image. The method comprises the following steps: step 1, carrying out Lee filtering processing on a high-resolution No.3 data original image; 2, geocoding the SAR data by using the SRTM DEM data as reference data, and converting the SAR data from a radar coordinate system to a geographic coordinate system; 3, bridge extraction is carried out according to data processing results in the step 1 and the step 2; 4, performing differential settlement monitoring according to a bridge extraction result; and step 5, performing differential settlement evaluation. According to the invention, comprehensive and efficient monitoring of the bridge is realized, limitations of a traditional monitoring method and other image monitoring technologies are overcome, and tiny changes and potential safety hazards of the bridge structure can be found in time. Scientific basis is provided for maintenance and reinforcement of the bridge, decision makers are helped to take effective measures in time, safe operation of the bridge is guaranteed, the risk of safety accidents is reduced, and remarkable economic and social benefits are achieved.

Description

Technical Field

[0001] The present invention relates to a bridge settlement assessment method, in particular to a bridge uneven settlement assessment method based on SAR images. Background Art

[0002] As a critical component of transportation infrastructure, bridge safety is directly linked to smooth transportation and the safety of public life and property. Over the course of long-term use, bridges are subject to the influence of various natural factors (such as geological fluctuations, climate change, and river erosion) and human factors (such as traffic loads and overloading), leading to defects such as uneven settlement. If these defects are not discovered and addressed promptly, they can lead to serious safety accidents, resulting in enormous economic losses and social impacts.

[0003] Traditional bridge monitoring relies primarily on ground-based surveying methods such as leveling and total stations. These methods require the meticulous placement of numerous measurement points at key locations across the bridge. This process is not only labor-intensive, material-intensive, and time-consuming, but also requires extremely high precision and density in the placement of measurement points. Due to the limited distribution of measurement points, traditional measurement methods can only capture data at local locations, making it difficult to achieve comprehensive, systematic monitoring of the entire bridge structure. Hidden or difficult-to-reach areas often become monitoring blind spots, preventing the timely detection of potential safety hazards.

[0004] Although optical images have a relatively high resolution and can reveal some detailed features of bridges, they are significantly restricted by weather conditions. In severe weather conditions such as fog, rain, and dust, light propagation is obstructed, making it impossible to obtain clear and effective image data, resulting in interruptions in monitoring work and making it difficult to meet the needs of all-weather monitoring. Optical images mainly reflect two-dimensional information on the surface of bridges, and their ability to detect internal features of bridge structures (such as internal cracks, steel corrosion, etc.) and minor deformations (such as millimeter-level uneven settlement) is limited. These potential safety hazards are difficult to show in optical images, which limits the assessment of bridge safety conditions.

[0005] Traditional SAR imagery-based methods for bridge extraction have limited ability to identify bridge structures in complex environments, making it difficult to accurately extract key information such as the bridge's main line, geometric features, and azimuth angles. When assessing uneven settlement, traditional methods lack high accuracy in calculating settlement gradients and curvatures, and are inaccurate in identifying sudden change areas. The lack of scientifically sound criteria makes them incapable of providing a reliable basis for bridge safety assessments. Summary of the Invention

[0006] Purpose of the invention: In response to the above problems, the purpose of the present invention is to provide a method for evaluating uneven settlement of bridges based on SAR images, which can overcome the problems of cumbersome arrangement of field measurement points, limited monitoring range, and susceptibility to interference from environmental factors in traditional measurement methods, reduce manpower, material and time costs, achieve all-round and systematic monitoring of the overall structure of the bridge, eliminate monitoring blind spots, and ensure that bridge monitoring data can be stably obtained in various complex environments. It can solve the problems of severe weather dependence of optical images and insufficient detection capabilities of internal features and micro-deformations, achieve all-weather monitoring of bridges, effectively detect micro-deformations and internal feature changes such as millimeter-level uneven settlement of bridges, and improve the comprehensiveness and accuracy of bridge safety assessments. It can improve the accuracy of data preprocessing, eliminate the influence of noise and terrain, and accurately extract key information such as the main line, geometric features and direction angles of the bridge; in the uneven settlement assessment, accurately calculate the settlement gradient and curvature, scientifically set the mutation area judgment standard, and provide a reliable basis for bridge safety assessment.

[0007] Technical solution: A method for evaluating uneven bridge settlement based on SAR images, including the following steps:

[0008] Step 1: Process the original image of GF-3 data using Lee filtering;

[0009] Step 2: Use SRTM DEM data as reference data to geocode SAR data and convert SAR data from radar coordinate system to geographic coordinate system;

[0010] Step 3: Based on the data processing results of steps 1 and 2, bridge extraction is performed, including bridge main line extraction, bridge geometric feature extraction, and bridge direction angle extraction. The bridge main line, geometric feature, and direction angle information are integrated to form a complete bridge extraction result. The bridge result is processed, including smoothing and removing isolated points, and the bridge extraction result is output in vector data format.

[0011] Step 4: Based on the bridge extraction results, perform uneven settlement monitoring. Geocode the multi-temporal SAR images of the same bridge area to generate an interferogram. Unwrap the phase and subtract the terrain phase to extract the deformation phase of the bridge area. Convert the phase into a deformation variable to obtain a time series settlement map. This forms a deformation distribution map of the bridge area, reflecting the settlement amount d at different locations.

[0012] Step 5: Conduct uneven settlement assessment. Based on the deformation distribution map, calculate the settlement gradient along the bridge, identify and determine the sudden change area, and perform settlement curvature analysis. Compare the calculated settlement gradient and settlement curvature K with the required safety threshold of the bridge to evaluate the structural safety of the bridge.

[0013] Furthermore, in step 3, the extraction of the bridge trunk line includes the following steps:

[0014] S311: Use edge detection algorithms to process SAR images to highlight the edge features of the bridge's main line, and extract the bridge's main line through Hough transform;

[0015] S312: Use any point (ρ, θ) in the polar coordinate system to represent any corresponding straight line in the image space, where ρ represents the vertical distance from the coordinate origin to the main line, and θ represents the angle between the vertical line and the positive direction of the x-axis; the equation of the bridge main line can be expressed as:

[0016] L = xcosθ + ysinθ;

[0017] S313: For each edge point in the image space, traverse all possible values of θ and calculate the corresponding ρ value:

[0018] ρ i =x i cosθ+y i sinθ;

[0019] Then, in the accumulator space A(ρ, θ) of the parameter space, vote for each (ρ, θ) position, and add 1 to the value of the corresponding position in the accumulator:

[0020] A(ρ,θ)=A(ρ,θ)+1;

[0021] S314: After the voting is completed, the local maximum values in the accumulator space are checked. The (ρ, θ) values corresponding to these peak points represent the straight lines in the image space. The threshold T obtained adaptively is used to filter out points with fewer votes, and only the straight lines with votes exceeding the threshold are retained. All votes in the accumulator space are statistically analyzed, and their mean μ and standard deviation σ are calculated. The main bridge line is extracted based on the detected (ρ, θ) values.

[0022] For the best case, let the number of votes in the accumulator space be v1, v2, ..., v n , where n is the total number of votes, the calculation formula for the mean μ is:

[0023]

[0024] The standard deviation reflects the degree of dispersion of the number of votes relative to the mean, and its calculation formula is:

[0025]

[0026] Based on the calculated mean μ and standard deviation σ, an adjustable coefficient k is introduced to calculate the adaptive threshold T. The formula is as follows:

[0027] T=μ+kσ;

[0028] The value of k is determined empirically, typically between 1, 2, and 3. A larger value of k results in a higher threshold, which results in more lines being filtered out. A smaller value of k results in a lower threshold, which results in more lines being retained. The value of k can be adjusted based on actual conditions to achieve optimal line extraction results.

[0029] When a certain position (ρ j ,θ j )'s vote count A(ρ j ,θ j ) satisfies A(ρ j ,θ j )≥T, it is considered that (ρ j ,θ j ) value is the real straight line in the image space and is retained; and when A(ρ j ,θ j )<T, the straight line corresponding to the position is eliminated.

[0030] Furthermore, in step 3, the bridge geometric feature extraction includes the following steps:

[0031] S321: Apply morphological operations to adjust the contour, remove stray points, connect broken parts, and calculate the geometric parameters of the bridge, including length and width, to provide a basis for subsequent bridge positioning;

[0032] S322: Obtain the bridge length by calculating the longest distance between all points on the outline, obtain the outline of the bridge using the minimum enclosing rectangle, and calculate the size of these enclosing shapes as the bridge length;

[0033] S323: Select multiple cross-section locations on the bridge profile, calculate the width of the profile at each cross-section, and average the widths of all cross-sections to obtain the average width of the bridge.

[0034] Furthermore, in step 3, the direction of the bridge is obtained by calculating the main axis direction of the bridge profile. The bridge direction angle extraction includes the following steps:

[0035] S331: Calculation of main axis direction;

[0036] Calculate the distances and directions between all pairs of points on the contour, perform weighted average on these directions, and obtain the main axis direction of the contour, which is consistent with the extension direction of the bridge;

[0037] S332: calculation of tilt angle and azimuth angle;

[0038] Tilt angle: α = θ main -θ horizontal , where θ horizontal is the angle in the horizontal direction, θmain is the angle of the principal axis;

[0039] Azimuth: β = θ main -θ north , where θ north It is the angle from north, usually 0° or 360°.

[0040] Furthermore, in step 5, the calculation of the sedimentation gradient includes the following steps:

[0041] S511: Determine a calculation path based on the main axis direction of the bridge. For each pair of adjacent pixels on the path, obtain the corresponding settlements d1 and d2, and calculate the deformation difference Δd = |d2-d1|.

[0042] S512: Assume that the image resolution is r. In the Cartesian coordinate system, the coordinates of horizontally adjacent pixels are (x1, y1) and (x1+1, y1), and the coordinates of vertically adjacent pixels are (x1, y1) and (x1, y1+1). Then the distances between horizontally and vertically adjacent pixels are both Δs=r. Assume that the coordinates of diagonally adjacent pixels are (x1, y1) and (x1+1, y1+1). According to the Pythagorean theorem,

[0043] S513: Calculate the sedimentation gradient by the ratio of the deformation difference and distance between adjacent pixels:

[0044] Preferably, in step 5, the identification and determination of the mutation region comprises the following steps:

[0045] S521: Setting a sliding window, sliding the window along the main axis direction on the bridge deformation diagram;

[0046] S522: When the window is located at a certain position in the deformation map, calculate the sedimentation gradient value G of all pixel pairs in the window ij , where i and j are the row and column indices of the pixels in the window. For each pair of adjacent pixels, the sedimentation gradient is calculated:

[0047]

[0048] S523: Calculate the average value of the sedimentation gradient within the window and standard deviation σ G :

[0049]

[0050] Where N is the total number of pixel pairs in the window;

[0051]

[0052] S524: When the average sedimentation gradient in the window And set the threshold G T , there are S in the window, and S ≥ 0.3N, G S >G T , then the window area is determined to be a mutation area.

[0053] Optimally, in step 5, the settlement curvature of the bridge area is calculated by taking the second-order derivative of the bridge deformation curve, reflecting the degree of settlement curvature. The second-order derivative is calculated using the numerical difference method. For discrete settlement data points i, where i = 1, 2, …, n, the second-order derivative is expressed as:

[0054]

[0055] Optimally, in step 5, for large highway bridges, when 0.001 / m 2 <K≤0.003 / m 2 When K>0.003 / m 2 When the bridge structure is marked as being in a dangerous state, reinforcement or repair measures must be taken immediately.

[0056] Optimally, in step 5, for railway bridges, when 0.0005 / m 2 <K≤0.002 / m 2 When K>0.002 / m 2 Emergency measures are required.

[0057] Beneficial effects: Compared with the prior art, the advantages of the present invention are:

[0058] Based on the advantages of SAR images such as all-day, all-weather, high resolution, and large-area coverage, the present invention realizes comprehensive and efficient monitoring of bridges, overcomes the limitations of traditional monitoring methods and other imaging monitoring technologies, and can promptly detect subtle changes in bridge structures and potential safety hazards.

[0059] Through precise bridge extraction and scientific uneven settlement assessment methods, the geometric parameters, settlement gradient and settlement curvature of the bridge can be accurately calculated. Combined with the engineering safety threshold, it provides a scientific basis for bridge maintenance and reinforcement, helping decision makers to take effective measures in a timely manner to ensure the safe operation of the bridge and reduce the risk of safety accidents, with significant economic and social benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 Flow chart of the method of the present invention. DETAILED DESCRIPTION

[0061] The present invention will be further described below with reference to the accompanying drawings and specific examples. It should be understood that these examples are only used to illustrate the present invention and are not used to limit the scope of the present invention.

[0062] A bridge uneven settlement assessment method based on SAR images, such as Figure 1 As shown, the following steps are included:

[0063] Step 1: The original image of the GF-3 data is processed using Lee filtering to maintain good image edges and reduce the adverse effects of noise on image quality.

[0064] Step 2: Use SRTM DEM data as reference data to geocode SAR data, convert SAR data from radar coordinate system to geographic coordinate system, and eliminate the effects of perspective contraction, top and bottom displacement, etc.

[0065] Step 3: Bridge extraction;

[0066] Based on the above data processing results, bridge extraction is performed and the following design is made to ensure accurate identification of bridge structures in complex environments.

[0067] (1) Bridge main line extraction

[0068] Edge detection algorithms (such as Canny edge detection) are used to process SAR images to highlight edge features such as the main line of the bridge. Hough transform is introduced to extract the main line of the bridge. Polar coordinates (ρ, θ) are used as the parameter space, where ρ represents the vertical distance from the coordinate origin to the main line, and θ represents the angle between the vertical line and the positive direction of the x-axis. Any straight line in the image space can be represented as a point (ρ, θ) in the polar coordinate system, and a point in the image space corresponds to a sine curve in the parameter space. The equation of the main line of the bridge can be expressed as:

[0069] L = xcosθ + ysinθ;

[0070] For each edge point in the image space, we traverse all possible values of θ (usually within a preset range, such as 0° to 180°, with a set step size) and calculate the corresponding ρ value. Then, we vote (i.e., accumulate the count) for each (ρ,θ) position in the two-dimensional matrix of parameter space (or accumulator space).

[0071] After voting is complete, the local maxima (i.e., peaks) in the accumulator space are checked. The (ρ,θ) values corresponding to these peaks represent straight lines in the image space. By setting a threshold T, points with fewer votes can be filtered out, retaining only lines with votes exceeding the threshold. Based on the detected (ρ,θ) values, the main bridge line can be extracted.

[0072] Assume that the total number of votes in the accumulator space is v1,v2,…,v n , where n is the total number of votes, the calculation formula for the mean μ is:

[0073]

[0074] The standard deviation reflects the degree of dispersion of the number of votes relative to the mean, and its calculation formula is:

[0075]

[0076] Based on the calculated mean μ and standard deviation σ, an adjustable coefficient k is introduced to calculate the adaptive threshold T. The formula is as follows:

[0077] T=μ+kσ;

[0078] The value of k is determined empirically, typically between 1, 2, and 3. A larger value of k results in a higher threshold, which results in more lines being filtered out. A smaller value of k results in a lower threshold, which results in more lines being retained. The value of k can be adjusted based on actual conditions to achieve optimal line extraction results.

[0079] When a certain position (ρ j ,θ j )'s vote count A(ρ j ,θ j ) satisfies A(ρ j ,θ j )≥T, it is considered that (ρ j ,θ j ) value is the real straight line in the image space and is retained; and when A(ρ j ,θ j )<T, the straight line corresponding to the position is eliminated.

[0080] (2) Bridge geometric feature extraction

[0081] Morphological operations (such as dilation, erosion, and opening and closing) are applied to fine-tune the outline, remove stray points, and connect broken parts. Geometric parameters such as the length and width of the bridge are calculated to provide a basis for subsequent precise positioning of the bridge.

[0082] Length Calculation: The length of a bridge can be approximated by calculating the longest distance between all points on the outline. The outline is approximated using minimum bounding rectangles (MBRs), and the dimensions of these bounding shapes are then calculated as an estimate of the bridge length.

[0083] Width Calculation: The width of a bridge is not typically a single value, as it can vary at different locations. A simplified approach is to calculate the average width of the bridge profile at multiple cross-sections. Select multiple cross-sectional locations along the bridge profile and, for each cross-section, calculate the width of the profile at that location. The average of all cross-sectional widths is then taken to determine the average width of the bridge.

[0084]

[0085] (3) Bridge direction angle extraction

[0086] The orientation of a bridge can be determined by calculating the principal axis of the profile. This principal axis is typically the weighted average direction of the distances between all pairs of points on the profile. Once the principal axis is determined, parameters such as the bridge's inclination or azimuth can be calculated.

[0087] ① Calculation of main axis direction:

[0088] Calculate the distances and directions between all pairs of points on the contour. Take a weighted average of these directions to get the principal axis of the contour. The principal axis usually coincides with the extension direction of the bridge.

[0089] ② Calculation of tilt angle or azimuth angle:

[0090] After obtaining the main axis direction, the bridge's inclination angle (angle with the horizontal direction) or azimuth (angle relative to the north direction) can be further calculated. Formula: Inclination angle α = θ main -θ horizontal , where θ horizontal is the angle in the horizontal direction; azimuth angle β = θ main -θ north , where θ north is the angle from north (usually 0° or 360°).

[0091] (4) Output the water bridge extraction results

[0092] Integrate the main line, geometric features and direction angle information of the bridge to form a complete bridge extraction result. Post-process the bridge result by smoothing, removing isolated points, etc., and output the bridge extraction result in vector data format.

[0093] Step 4: Uneven settlement monitoring;

[0094] (1) Multi-temporal data registration: Geocoding multi-temporal SAR images of the same bridge area to ensure spatial consistency.

[0095] (2) Differential Interferometry (DInSAR): Generate interferograms, unwrap the phase and subtract the terrain phase (based on SRTM DEM), and extract the deformation phase of the bridge area.

[0096] (3) Deformation map generation: The phase is converted into deformation variables to obtain a time series settlement map. This generates a deformation distribution map of the bridge area, reflecting the settlement amount d at different locations.

[0097] Step 5: Differential settlement assessment;

[0098] (1) Settlement gradient calculation: Based on the deformation map, the settlement gradient along the bridge is calculated to assess the degree of uneven settlement. The settlement gradient can be calculated by the ratio of the deformation difference between adjacent pixels to the distance. The calculation path is determined based on the direction of the bridge's main axis. Then, for each pair of adjacent pixels on the path, the corresponding settlement amounts d1 and d2 are obtained, and the deformation difference Δd = |d2-d1| is calculated.

[0099] The distance Δs between adjacent pixels can be calculated based on the image resolution and coordinate system: if the image resolution is r (unit: meters / pixel), in the Cartesian coordinate system, the coordinates of horizontally adjacent pixels change to (x1, y1) and (x1+1, y1), and the coordinates of vertically adjacent pixels change to (x1, y1) and (x1, y1+1), then the distance Δs between them = r.

[0100] Obliquely adjacent pixels: For obliquely adjacent pixels, for example, from coordinates (x1, y1) and (x1+1, y1+1), according to the Pythagorean theorem,

[0101] The sedimentation gradient is:

[0102]

[0103] (2) Identification and determination of mutation areas: By setting a sliding window (for example, the window size is 5×5 pixels), the window is slid along the main axis direction on the bridge deformation map.

[0104] ① Calculation of data within the window: When the window is located at a certain position in the deformation map, calculate the sedimentation gradient value G of all pixel pairs within the window ij (i, j are the row and column indices of the pixels in the window). For each pair of adjacent pixels, (Horizontal direction) Calculate the sedimentation gradient.

[0105] ②Statistical calculation: calculate the average value of the sedimentation gradient within the window and standard deviation σ G .

[0106] average value Where N is the total number of pixel pairs in the window;

[0107] Standard deviation

[0108] ③ Mutation region determination condition: when the average sedimentation gradient within the window And set the threshold G T , there are S in the window, and S ≥ 0.3N, G S >G T , then the window area is determined to be a possible mutation area.

[0109] (3) Sedimentation curvature analysis;

[0110] By calculating the settlement curvature of the bridge area, the impact of uneven settlement on the safety of the bridge structure is further analyzed. The second-order derivative of the bridge deformation curve is calculated:

[0111] Reflects the degree of settlement curvature. Areas with high curvature may face structural stress concentration.

[0112] In specific calculations, the numerical difference method can be used to approximate the second-order derivative. For example, for discrete settlement data points (i = 1, 2, ..., n), the distance between two adjacent points is Δs, then the second-order derivative can be approximately expressed as:

[0113]

[0114] (4) Bridge structure safety assessment standards and thresholds;

[0115] The safety of bridge structures is assessed by combining engineering safety thresholds. According to bridge design specifications and safety standards, different types and grades of bridges have different safety thresholds.

[0116] ① For large highway bridges, when 0.001 / m 2 <K≤0.003 / m 2 When the bridge is exposed to high pressure, it may cause significant stress concentration in the bridge structure, affecting the structural safety;

[0117] When the settlement curvature exceeds K>0.003 / m 2 When the bridge structure is in a dangerous state, it is necessary to take immediate reinforcement or repair measures.

[0118] ② For railway bridges, since the requirements for deformation are more stringent, the corresponding safety threshold is lower, such as when the settlement curvature exceeds 0.0005 / m 2 <K≤0.002 / m 2 Close attention is required.

[0119] K>0.002 / m 2 By comparing the calculated settlement gradient and settlement curvature with these safety thresholds, the structural safety of the bridge can be accurately assessed and potential safety hazards can be discovered in a timely manner.

[0120] The present invention uses edge detection algorithms (such as Canny edge detection) to enhance the edge features of bridges in SAR images, and combines Hough transform to accurately extract the main line of the bridge through traversal calculation and voting mechanism of edge points in polar coordinate parameter space. Morphological operations (dilation, erosion, opening and closing operations) are used to refine the bridge contour, remove stray points, and connect broken parts. By calculating the longest distance of all points on the contour and approximating the contour with the help of the minimum enclosing rectangle, the length of the bridge is accurately calculated; by calculating the contour width at multiple cross-sectional positions and averaging the values, the average width of the bridge is obtained. Based on the weighted average of the distance and direction between all pairs of points on the contour, the main axis direction of the bridge contour is calculated, and then the inclination angle and azimuth of the bridge are accurately calculated based on the angular relationship between the horizontal direction and the north direction. Multi-phase SAR images of the same bridge area are geocoded to ensure the spatial consistency of the images, providing a reliable basis for differential interferometry processing. Using differential interferometry (DInSAR) technology, an interferogram is generated and the phase is unwrapped. Simultaneously, the terrain phase is subtracted from the SRTM DEM to accurately extract the deformation phase of the bridge area. A time-series settlement map is generated through the conversion relationship between phase and deformation. Based on the deformation map, a calculation path is determined according to the direction of the bridge's principal axis, and the settlement gradients of adjacent pixels are precisely calculated. The distance between adjacent pixels is accurately calculated based on the image resolution and coordinate system. A sliding window is set and slid along the principal axis of the bridge deformation map. The mean and standard deviation of the settlement gradient within the window are calculated. Areas of sudden change are identified based on strict criteria (such as the multiple relationship between the mean and standard deviation, a set threshold, and the ratio of pixel pairs meeting the threshold). A numerical difference method is used to calculate the second-order derivative of the bridge deformation curve to obtain the settlement curvature. This, combined with engineering safety thresholds for different bridge types, accurately assesses the safety of the bridge structure.

Claims

1. A bridge uneven settlement assessment method based on SAR images, characterized by The following steps are involved: Step 1: Process the original image of GF-3 data using Lee filtering; Step 2: Use SRTM DEM data as reference data to geocode SAR data and convert SAR data from radar coordinate system to geographic coordinate system; Step 3: Based on the data processing results of steps 1 and 2, bridge extraction is performed, including bridge main line extraction, bridge geometric feature extraction, and bridge direction angle extraction. The bridge main line, geometric feature, and direction angle information are integrated to form a complete bridge extraction result. The bridge result is processed, including smoothing and removing isolated points, and the bridge extraction result is output in vector data format. Step 4: Based on the bridge extraction results, perform uneven settlement monitoring. Geocode the multi-temporal SAR images of the same bridge area to generate an interferogram. Unwrap the phase and subtract the terrain phase to extract the deformation phase of the bridge area. Convert the phase into a deformation variable to obtain a time series settlement map. This forms a deformation distribution map of the bridge area, reflecting the settlement amount d at different locations. Step 5: Conduct uneven settlement assessment. Based on the deformation distribution map, calculate the settlement gradient along the bridge, identify and determine the sudden change area, and perform settlement curvature analysis. Compare the calculated settlement gradient and settlement curvature K with the required safety threshold of the bridge to evaluate the structural safety of the bridge.

2. The method for evaluating uneven bridge settlement based on SAR images according to claim 1, characterized in that: In step 3, the extraction of bridge trunk lines includes the following steps: S311: Use edge detection algorithms to process SAR images to highlight the edge features of the bridge's main line, and extract the bridge's main line through Hough transform; S312: Use any point (ρ, θ) in the polar coordinate system to represent any corresponding straight line in the image space, where ρ represents the vertical distance from the coordinate origin to the main line, and θ represents the angle between the vertical line and the positive direction of the x-axis; the equation of the bridge main line can be expressed as: L = xcosθ + ysinθ; S313: For each edge point in the image space, traverse all possible values of θ and calculate the corresponding ρ value: r i =x i cosθ+y i sinθ; Then, in the accumulator space A(ρ, θ) of the parameter space, vote for each (ρ, θ) position, and add 1 to the value of the corresponding position in the accumulator: A(ρ, θ)=A(ρ, θ)+1; S314: After the voting is completed, the local maximum values in the accumulator space are checked. The (ρ, θ) values corresponding to these peak points represent the straight lines in the image space. The threshold T obtained adaptively is used to filter out points with fewer votes, and only the straight lines with votes exceeding the threshold are retained. A statistical analysis is performed on all votes in the accumulator space, and their mean μ and standard deviation σ are calculated. The main bridge line is extracted based on the detected (ρ, θ) values.

3. The method for evaluating uneven bridge settlement based on SAR images according to claim 2, characterized in that: Assume that the number of votes in the accumulator space is v1, v2, ..., v n , where n is the total number of votes, the calculation formula for the mean μ is: The standard deviation reflects the degree of dispersion of the number of votes relative to the mean, and its calculation formula is: According to the calculated mean μ and standard deviation σ, an adjustable coefficient k is introduced to calculate the adaptive threshold T. The formula is as follows: T=μ+kσ; When a certain position (ρ j ,θ j )’s vote count A(ρ j ,θ j ) satisfies A(ρ j ,θ j )≥T, it is considered that the position (ρ j ,θ j ) value is the real straight line in the image space and is retained; and when A(ρ j ,θ j )<T, the straight line corresponding to the position is eliminated.

4. The method for evaluating uneven bridge settlement based on SAR images according to claim 1, characterized in that: In step 3, bridge geometric feature extraction includes the following steps: S321: Apply morphological operations to adjust the contour, remove stray points, connect broken parts, and calculate the geometric parameters of the bridge, including length and width, to provide a basis for subsequent bridge positioning; S322: Obtain the bridge length by calculating the longest distance between all points on the outline, obtain the outline of the bridge using the minimum enclosing rectangle, and calculate the size of these enclosing shapes as the bridge length; S323: Select multiple cross-section locations on the bridge profile, calculate the width of the profile at each cross-section, and average the widths of all cross-sections to obtain the average width of the bridge.

5. The method for evaluating uneven bridge settlement based on SAR images according to claim 1, characterized in that: In step 3, the direction of the bridge is obtained by calculating the main axis direction of the bridge profile. The bridge direction angle extraction includes the following steps: S331: Calculation of main axis direction; Calculate the distances and directions between all pairs of points on the contour, perform weighted average on these directions, and obtain the main axis direction of the contour, which is consistent with the extension direction of the bridge; S332: calculation of tilt angle and azimuth angle; Tilt angle: α = θ main -θ horizontal , where θ horizontal is the angle in the horizontal direction, θ main is the angle of the principal axis; Azimuth: β = θ main -θ north , where θ north It is the angle from north, usually 0° or 360°.

6. The method for evaluating uneven bridge settlement based on SAR images according to claim 1, characterized in that: In step 5, the calculation of the sedimentation gradient includes the following steps: S511: Determine a calculation path based on the main axis direction of the bridge. For each pair of adjacent pixels on the path, obtain the corresponding settlements d1 and d2, and calculate the deformation difference Δd = |d2-d1|. S512: Assume that the image resolution is r. In the Cartesian coordinate system, the coordinates of horizontally adjacent pixels are (x1, y1) and (x1+1, y1), and the coordinates of vertically adjacent pixels are (x1, y1) and (x1, y1+1). Then the distances between horizontally and vertically adjacent pixels are both Δs=r. Assume that the coordinates of diagonally adjacent pixels are (x1, y1) and (x1+1, y1+1). According to the Pythagorean theorem, S513: Calculate the sedimentation gradient by the ratio of the deformation difference and distance between adjacent pixels:

7. The method for evaluating uneven bridge settlement based on SAR images according to claim 6, characterized in that: In step 5, the identification and determination of the mutation region includes the following steps: S521: Setting a sliding window, sliding the window along the main axis direction on the bridge deformation map; S522: When the window is located at a certain position in the deformation map, calculate the sedimentation gradient value G of all pixel pairs in the window ij , where i and j are the row and column indices of the pixels in the window. For each pair of adjacent pixels, the sedimentation gradient is calculated: S523: Calculate the average value of the sedimentation gradient within the window and standard deviation σ G : Where N is the total number of pixel pairs in the window; S524: When the average sedimentation gradient in the window And set the threshold G T , there are S in the window, and S ≥ 0.3N, G S >G T , then the window area is determined to be a mutation area.

8. The method for evaluating uneven bridge settlement based on SAR images according to claim 6, characterized in that: In step 5, the settlement curvature of the bridge area is obtained by calculating the second-order derivative of the bridge deformation curve, which reflects the degree of settlement curvature. The second-order derivative is calculated using the numerical difference method. For discrete settlement data points i, where i = 1, 2, ..., n, the second-order derivative is expressed as:

9. The method for evaluating uneven bridge settlement based on SAR images according to claim 1 or 8, characterized in that: In step 5, for large highway bridges, when 0.001 / m 2 <K≤0.003 / m 2 When K>0.003 / m 2 When the bridge structure is marked as being in a dangerous state, reinforcement or repair measures must be taken immediately.

10. The method for evaluating uneven bridge settlement based on SAR images according to claim 1 or 8, characterized in that: In step 5, for railway bridges, when 0.0005 / m 2 <K≤0.002 / m 2 When K>0.002 / m 2 Emergency measures are required.

Images