A radar layer position tracking method fusing confidence clustering and wavelet energy discrimination

Abstract

The present application relates to radar data processing technical field, specifically to a kind of radar layer position tracking method of fusing confidence clustering and wavelet energy discrimination, comprising the following steps: S1, the gray processing is carried out to each frame radar image and the candidate layer position point of the edge gradient greater than threshold is extracted using Sobel operator;S2, detect and eliminate outlier, false edge and low confidence point;S3, according to confidence weighted clustering to generate structure point;S4, optimal trajectory is spliced using DTW algorithm to form layer position path;S5, calculate wavelet energy variation rate to correct deviated path;S6, curve fitting and smoothing to generate final layer position trajectory.The present application, by outlier elimination, confidence weighted clustering, dynamic path planning and wavelet energy auxiliary correction mechanism, improves the accuracy, continuity and robustness under complex structure region of radar layer position tracking.

Description

Technical Field

[0001] This invention relates to the field of radar data processing technology, and in particular to a radar layer tracking method that integrates confidence clustering and wavelet energy discrimination. Background Technology

[0002] Ground-penetrating radar (GPR), as a non-destructive electromagnetic wave detection technology, has been widely used in fields such as road structure layer identification, underground structure analysis, and defect location. By analyzing B-scan images or 3D radar data acquired by radar, it can achieve accurate tracking of structural layers. This is the foundation for supporting structural thickness estimation, crack identification, and defect assessment. Among these, automatic layer tracking algorithms based on image edge extraction or signal similarity matching are the current focus of research and engineering practice.

[0003] Most existing layer-based tracking methods are based on image edge extraction combined with clustering and path planning, or directly construct paths based on the similarity of A-scan signals. Although they have achieved certain results in some scenarios, they still have many problems when facing complex structural areas such as bridge deck paving and local repair sections. For example, existing methods generally lack a mechanism to remove abnormal points (such as artifacts, local noise, and abrupt edges) in radar images, which can easily lead to misjudgments. Path construction often relies on global matching, which has poor adaptability to local abrupt change areas. The use of signal features is also relatively simple, and it fails to integrate frequency domain information such as wavelet energy to effectively evaluate and correct the path results. These problems result in certain limitations in the physical continuity, structural consistency, and engineering adaptability of the tracking path. Summary of the Invention

[0004] This invention provides a radar stratigraphic tracking method that integrates confidence clustering and wavelet energy discrimination. Without relying on structural region prediction, it achieves automatic path optimization and adjustment and enhanced physical structure consistency. By constructing a confidence model with multi-feature fusion and an anomaly identification mechanism, it effectively improves the adaptive ability of the tracking path to complex structural regions. Furthermore, by triggering a path correction strategy through wavelet energy change rate, it ultimately generates a high-quality structural stratigraphic trajectory with strong continuity and high stability, which can be used for subsequent thickness analysis and defect identification.

[0005] A radar stratigraphic tracking method integrating confidence clustering and wavelet energy discrimination includes the following steps: S1, perform grayscale image processing on each frame of radar image, and use the Sobel edge detection operator to extract edge change points in the radar image whose edge gradient value is greater than the edge gradient threshold as candidate layer sites; S2, anomaly detection is performed on candidate layer sites, and the identified local outliers, false edges or false detections and low confidence points are removed; S3, perform weighted clustering of the candidate layer sites after elimination based on confidence level; S4. Based on the clustering results, a path is constructed for the intra-frame structure point sequence of each layer. The Dynamic Time Warping (DTW) algorithm is used to achieve optimal trajectory stitching and generate a layer-by-layer tracking path. S5, perform wavelet packet decomposition on the layer tracking path, calculate local energy and energy change rate, and identify and correct path segments that deviate from the energy backbone; S6 performs curve fitting and smoothing on the corrected stratigraphic tracking path to generate the final continuous structural stratigraphic trajectory.

[0006] Optionally, S1 includes: S11, convert the radar image into a grayscale image and calculate its grayscale value. ; S12, the Sobel operator is used to perform edge detection on the grayscale image, and the gradient value in the horizontal direction is calculated. gradient value in the vertical direction ; S13, combined with the gradient value in the horizontal direction gradient value in the vertical direction Calculate the gradient magnitude of each pixel. ; S14, Set edge gradient threshold ,filter The pixels are used as candidate layer sites.

[0007] Optionally, S2 includes: S21, for candidate layer sites Calculate the deviation value between its depth and the mean depth of points in the neighborhood. ,when Exceeding the set depth tolerance threshold When this occurs, it is identified as a local outlier. S22, with candidate layer sites Set a neighborhood window with a fixed shape as the center. Count the number of candidate layer sites. And based on the actual area of ​​the neighboring window Calculate point density ,like Below the density threshold Then candidate layer sites Isolated distribution within a local area is identified as a false edge or a false detection point; S23, Calculate the coefficient of variation of edge intensity within the neighborhood of candidate layer sites. ,like Exceeding the neighborhood edge strength threshold If it is, then it is judged as a low confidence point.

[0008] Optionally, S3 includes: S31, extract multidimensional features from the eliminated candidate layer sites, including edge gradient values. Local texture clarity and the wavelet energy density of the A-scan signal ; S32 uses the Min-Max normalization method to standardize the multidimensional features; S33, average the standardized multidimensional features and calculate the confidence level of candidate layer sites; S34. Based on the confidence level of candidate layer sites as clustering weights, the weighted KMeans clustering method is used to cluster candidate layer sites and output cluster centers.

[0009] Optionally, the confidence level of the candidate layer sites is expressed as: ; in, For confidence level, , , These are the standardized multidimensional features of teammates.

[0010] Optionally, the cluster centers are represented as: ; in, For the first Cluster centers of candidate site in class (structural layer) For the first The set of candidate layer sites for a class.

[0011] Optionally, S4 includes: S41, Based on the clustering results, extract the sequence of structure points in each frame of radar image; S42 uses the Dynamic Time Warping (DTW) algorithm to recursively calculate the minimum path matching distance between the current frame point and its adjacent frame points, expressed as: ; S43, introduce dynamic window constraints during the matching process. ; S44 uses the Dynamic Time Warping (DTW) algorithm to accumulate the cost matrix and backtrack the optimal matching path to achieve continuous splicing of structural points between multiple frames, generating a layer tracking path.

[0012] Optionally, S5 includes: S51, extract the A-scan signal corresponding to each structural point along the stratigraphic tracing path. Then, perform wavelet packet decomposition on it to obtain the corresponding sub-band signal after decomposition. ; S52, calculate the local energy for the decomposition results of each structural point. , is represented as: ; S53, calculate the rate of energy change for adjacent points. , is represented as: ; S54, if the energy change rate of the layer tracing path exceeds the mutation threshold If the path deviates from the main energy structure region, a path correction operation is performed to reselect the local optimal structure point.

[0013] Optionally, S6 includes: S61, polynomial fitting is used to modify the structure point sequence. Curve fitting was performed to construct a continuous trajectory of the hierarchical structure. S62, the sliding window averaging method is used to denoise and smooth the fitted trajectory; S63 generates the final continuous structural layer trajectory based on the smoothed layer tracking path, including the three structural layer depth values ​​corresponding to each frame of A-scan, three complete structural path lines, a visualization map, and an exportable structural layer depth data file.

[0014] The beneficial effects of this invention are: This invention introduces a multi-index driven anomaly identification mechanism, constructs an anomaly detection model based on the depth deviation, neighborhood density, and edge strength variability of candidate points, and achieves accurate removal of isolated points, outliers, and low-confidence points. This significantly reduces the interference of noise points on subsequent hierarchical clustering and path splicing, and effectively improves the accuracy and robustness of the initial structure point set of the path.

[0015] This invention constructs a confidence system that integrates edge gradient, texture clarity, and A-scan wavelet energy density, and introduces the confidence level as a weight into the weighted KMeans clustering process, making the cluster centers closer to the real layer structure. Furthermore, by combining the Dynamic Time Warping (DTW) algorithm and the sliding window constraint strategy, it achieves stable stitching of structural layers in multi-frame radar images, improving path continuity and robustness in matching complex regions.

[0016] This invention designs a path correction mechanism based on wavelet packet energy change rate, triggering local path backtracking and structural point re-estimation in energy mutation regions. Combined with polynomial fitting and sliding window smoothing, it further optimizes the physical consistency and global stability of the path. The final output layer trajectory has good continuity, visibility and derivability, and is suitable for high-precision structural layer identification, thickness analysis and engineering map display and other application scenarios. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only for this invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the tracking method according to an embodiment of the present invention. Detailed Implementation

[0019] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. Those skilled in the art may employ other alternative methods to implement some well-known technologies; moreover, the accompanying drawings are only for more specific description of the embodiments and are not intended to specifically limit the present invention.

[0020] like Figure 1 As shown, a radar stratigraphic tracking method integrating confidence clustering and wavelet energy discrimination includes the following steps: S1, perform grayscale image processing on each frame of radar image, and use the Sobel edge detection operator to extract edge change points in the radar image whose edge gradient value is greater than the edge gradient threshold as candidate layer sites; S11, convert the radar image into a grayscale image and calculate its grayscale value. , is represented as: ; in, , , These are the red, green, and blue channel values ​​of the original image at that pixel, respectively. S12, the Sobel operator is used to perform edge detection on the grayscale image, and the gradient value in the horizontal direction is calculated. gradient value in the vertical direction , is represented as: ; ; in, , These are the horizontal and vertical convolution kernels of Sobel, respectively; ; ; S13, combined with the gradient value in the horizontal direction gradient value in the vertical direction Calculate the gradient magnitude of each pixel. , is represented as: ; S14, Set edge gradient threshold ,filter The pixels are used as candidate layer sites, represented as follows: ; in, This is the edge gradient threshold.

[0021] S2, anomaly detection is performed on candidate layer sites, and the identified local outliers, false edges or false detections and low confidence points are removed; S21, for candidate layer sites Calculate the deviation value between its depth and the mean depth of points in the neighborhood. ,when Exceeding the set depth tolerance threshold ( When set to 12), it is determined to be a local outlier, represented as: ; in, Candidate layer sites depth, Candidate layer sites Number of candidate layer sites in the neighborhood; S22, with candidate layer sites Set a neighborhood window with a fixed shape as the center. Count the number of candidate layer sites. And based on the actual area of ​​the neighboring window Calculate point density ,like Below the density threshold ( If set to 0.08), then the candidate layer site Isolated distributions within a local area are identified as false edges or false detection points, and are represented as follows: ; S23, Calculate the coefficient of variation of edge intensity within the neighborhood of candidate layer sites. ,like Exceeding the neighborhood edge strength threshold ( If the value is set to 0.5, it is considered a low-confidence point, as shown below: ; in, , These are the standard deviation and mean of the neighborhood edge strength, respectively.

[0022] S3, perform weighted clustering of the candidate layer sites after elimination based on confidence level; S31, extract multidimensional features from the eliminated candidate layer sites, including edge gradient values. Local texture clarity and the wavelet energy density of the A-scan signal ; S32 uses the Min-Max normalization method to standardize the multidimensional features; S33, the standardized multidimensional features are averaged to calculate the confidence score of the candidate layer sites, expressed as: ; in, For confidence level, , , These are the standardized multidimensional features of teammates; S34, using the confidence score of candidate layer sites as clustering weights, the weighted KMeans clustering method is employed to cluster the candidate layer sites, outputting the cluster centers, as follows: ; in, For the first Cluster centers of candidate site in class (structural layer) For the first The set of candidate layer sites for a class.

[0023] S4. Based on the clustering results, a path is constructed for the intra-frame structure point sequence of each layer. The Dynamic Time Warping (DTW) algorithm is used to achieve optimal trajectory stitching and generate a layer-by-layer tracking path. S41, Based on the clustering results, extract the sequence of structure points in each frame of radar image; S42 uses the Dynamic Time Warping (DTW) algorithm to recursively calculate the minimum path matching distance between the current frame point and its adjacent frame points, expressed as: ; S43, introduce dynamic window constraints during the matching process. Limit the path offset range to avoid crossing structural regions or mismatching. S44 uses the Dynamic Time Warping (DTW) algorithm to accumulate the cost matrix and backtrack the optimal matching path to achieve continuous splicing of structural points between multiple frames, generating a layer tracking path.

[0024] S5, perform wavelet packet decomposition on the layer tracking path, calculate local energy and energy change rate, and identify and correct path segments that deviate from the energy backbone; S51, extract the A-scan signal corresponding to each structural point along the stratigraphic tracing path. Then, perform wavelet packet decomposition on it to obtain the corresponding sub-band signal after decomposition. ; S52, calculate the local energy for the decomposition results of each structural point. , is represented as: ; S53, calculate the rate of energy change for adjacent points. , is represented as: ; S54, if the energy change rate of the layer tracing path exceeds the mutation threshold If the stratum tracking path deviates from the main energy structure region, a path correction operation is performed to reselect the local optimal structure point in order to achieve path stability and structural consistency. Mutation threshold Represented as: ; in, , , respectively, are the mean and standard deviation of the energy change rate for all structural point pairs. This is the adjustment coefficient.

[0025] S6, Perform curve fitting and smoothing on the corrected stratigraphic tracking path to generate the final continuous structural stratigraphic trajectory; S61, polynomial fitting is used to modify the structure point sequence. Curve fitting is performed to construct a continuous trajectory of the hierarchical structure, represented as follows: ; in, The horizontal position This represents the depth coordinates corresponding to that location. For the fitting function, The fitting order is [number]. S62, the sliding window averaging method is used to denoise and smooth the fitted trajectory, as shown below: ; in, For the smoothed first A depth value, The width of the sliding window. For the sliding window The fitting depth of each adjacent point; S63 generates the final continuous structural layer trajectory based on the smoothed layer tracking path, including the three structural layer depth values ​​corresponding to each frame of A-scan, three complete structural path lines, a visualization map, and an exportable structural layer depth data file.

[0026] This invention encompasses any substitutions, modifications, equivalent methods, and solutions made within the spirit and scope of this invention. To provide the public with a thorough understanding of this invention, specific details are described in detail in the following preferred embodiments; however, those skilled in the art will fully understand the invention even without these details. Furthermore, to avoid unnecessary misunderstanding of the essence of this invention, well-known methods, processes, procedures, components, and circuits are not described in detail.

[0027] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A radar stratigraphic tracking method integrating confidence clustering and wavelet energy discrimination, characterized in that, Includes the following steps: S1, perform grayscale image processing on each frame of radar image, and use the Sobel edge detection operator to extract edge change points in the radar image whose edge gradient value is greater than the edge gradient threshold as candidate layer sites; S2, anomaly detection is performed on candidate layer sites, and the identified local outliers, false edges or false detections and low confidence points are removed; S3, perform weighted clustering of the candidate layer sites after elimination based on confidence level; S4. Based on the clustering results, the path is constructed for the intra-frame structural point sequence of each layer. The dynamic time warping algorithm is used to achieve optimal trajectory stitching and generate the layer tracking path. S5, perform wavelet packet decomposition on the layer tracking path, calculate local energy and energy change rate, and identify and correct path segments that deviate from the energy backbone; S6 performs curve fitting and smoothing on the corrected stratigraphic tracking path to generate the final continuous structural stratigraphic trajectory.

2. The radar stratum tracking method integrating confidence clustering and wavelet energy discrimination according to claim 1, characterized in that, S1 includes: S11, convert the radar image into a grayscale image and calculate its grayscale value. ; S12, the Sobel operator is used to perform edge detection on the grayscale image, and the gradient value in the horizontal direction is calculated. gradient value in the vertical direction ; S13, combined with the gradient value in the horizontal direction gradient value in the vertical direction Calculate the gradient magnitude of each pixel. ; S14, Set edge gradient threshold ,filter The pixels are used as candidate layer sites.

3. The radar stratigraphic tracking method integrating confidence clustering and wavelet energy discrimination according to claim 2, characterized in that, S2 includes: S21, for candidate layer sites Calculate the deviation value between its depth and the mean depth of points in the neighborhood. ,when Exceeding the set depth tolerance threshold When this occurs, it is identified as a local outlier. S22, with candidate layer sites Set a neighborhood window with a fixed shape as the center. Count the number of candidate layer sites. And based on the actual area of ​​the neighboring window Calculate point density ,like Below the density threshold Then candidate layer sites Isolated distribution within a local area is identified as a false edge or a false detection point; S23, Calculate the coefficient of variation of edge intensity within the neighborhood of candidate layer sites. ,like Exceeding the neighborhood edge strength threshold If it is, then it is judged as a low confidence point.

4. The radar stratum tracking method integrating confidence clustering and wavelet energy discrimination according to claim 3, characterized in that, S3 includes: S31, extract multidimensional features from the eliminated candidate layer sites, including edge gradient values. Local texture clarity and the wavelet energy density of the A-scan signal ; S32 uses the Min-Max normalization method to standardize the multidimensional features; S33, average the standardized multidimensional features and calculate the confidence level of candidate layer sites; S34. Based on the confidence level of candidate layer sites as clustering weights, the weighted KMeans clustering method is used to cluster candidate layer sites and output cluster centers.

5. The radar stratigraphic tracking method integrating confidence clustering and wavelet energy discrimination according to claim 4, characterized in that, The confidence level of the candidate layer sites is expressed as follows: ； in, For confidence level, , , These are the standardized multidimensional features of teammates.

6. The radar stratigraphic tracking method integrating confidence clustering and wavelet energy discrimination according to claim 5, characterized in that, The cluster centers are represented as: ； in, For the first Cluster centers of candidate layer sites, For the first The set of candidate layer sites for a class.

7. The radar stratum tracking method integrating confidence clustering and wavelet energy discrimination according to claim 6, characterized in that, S4 includes: S41, Based on the clustering results, extract the sequence of structure points in each frame of radar image; S42 uses a dynamic time warping algorithm to recursively calculate the minimum path matching distance between the current frame point and its adjacent frame points, expressed as: ； S43, introduce dynamic window constraints during the matching process. ; S44: Based on the cumulative cost matrix of the dynamic time warping algorithm, backtrack the optimal matching path to realize the continuous splicing of structural points between multiple frames and generate a layer tracking path.

8. The radar stratum tracking method integrating confidence clustering and wavelet energy discrimination according to claim 7, characterized in that, S5 includes: S51, extract the A-scan signal corresponding to each structural point along the stratigraphic tracing path. Then, perform wavelet packet decomposition on it to obtain the corresponding sub-band signal after decomposition. ; S52, calculate the local energy for the decomposition results of each structural point. , represented as: ； S53, calculate the rate of energy change for adjacent points. , represented as: ； S54, if the energy change rate of the layer tracing path exceeds the mutation threshold If the path deviates from the main energy structure region, a path correction operation is performed to reselect the local optimal structure point.

9. The radar stratum tracking method integrating confidence clustering and wavelet energy discrimination according to claim 8, characterized in that, S6 includes: S61, polynomial fitting is used to modify the structure point sequence. Curve fitting was performed to construct a continuous trajectory of the hierarchical structure. S62, the sliding window averaging method is used to denoise and smooth the fitted trajectory; S63 generates the final continuous structural layer trajectory based on the smoothed layer tracking path, including the three structural layer depth values ​​corresponding to each frame of A-scan, three complete structural path lines, a visualization map, and an exportable structural layer depth data file.

Images