Improved ground-based SAR (Synthetic Aperture Radar) high-coherence point selection method

Abstract

The invention relates to the technical field of image processing, in particular to an improved ground-based SAR (synthetic aperture radar) high-coherence point selection method, which comprises the following steps of: extracting amplitude and phase frames, calculating statistical characteristics, fusing the statistical characteristics to generate an index diagram, dividing regions, constructing a gray histogram to generate a Boolean mask, establishing a co-occurrence matrix, eliminating abnormal regions, and selecting high-coherence points. According to the method, a multi-source fusion layer is constructed in combination with amplitude and phase features to enhance the image discrimination dimension, the region recognition precision is improved through gray frequency interval extraction and Boolean mask generation, and the region recognition precision is improved. Local structure features are extracted by using a gray level co-occurrence matrix, abnormal pixels are removed in combination with boundary coherence and density features, an adjacent path-based graph structure relationship is established to form a pixel connection network, and a pixel set with aggregation and connectivity is screened from the network to realize high-coherence point clustering output; and the integrity of structure expression in the complex region and the aggregation rationality of target selection points are enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to an improved method for selecting high-coherence points of ground-based SAR. Background Art

[0002] The field of image processing technology primarily studies how to acquire, enhance, analyze, and understand image information. This involves processing image data acquired from image sensors through algorithms such as noise reduction, enhancement, feature extraction, image segmentation, pattern recognition, and image understanding. It is widely used in fields such as remote sensing monitoring, medical imaging, industrial inspection, and intelligent transportation. Remote sensing image processing, as a key branch, is dedicated to improving the accuracy of interpreting Earth observation images. In particular, image processing technology plays a key role in surface deformation monitoring, target recognition, and information extraction in synthetic aperture radar (SAR) imagery. Traditional ground-based SAR high-coherence point selection methods involve performing amplitude and phase stability analysis on time-series SAR images and using a fixed threshold to select a set of highly stable points for interferometric processing. These methods typically use the amplitude deviation method, the phase standard deviation method, or the amplitude consistency index from stable scatterer interferometry as a basis to first extract candidate pixels and then select them based on their stability in the time series. Some methods also incorporate local neighborhood judgment to eliminate noise points.

[0003] When relying on a single statistical indicator to evaluate pixel stability, it is difficult to identify subtle structural differences in mixed areas of ground objects, and it is easy to cause misidentification when there are areas with similar grayscale but significant differences in spatial characteristics. The lack of joint consideration of the internal grayscale relationship and boundary structure of the region makes it difficult to eliminate misjudged areas. In scenes with discontinuous pixel boundaries or blurred grayscale textures, edge diffusion and false point selection are prone to occur, resulting in unclear identification of the internal structure of the region and insufficient accuracy in point set selection, affecting the graph structure construction and connectivity analysis in the subsequent interference modeling process. Summary of the Invention

[0004] In order to solve the technical problems existing in the prior art, an embodiment of the present invention provides an improved ground-based SAR high coherence point selection method, comprising the following steps: To achieve the above object, the present invention adopts the following technical solution: an improved ground-based SAR high coherence point selection method, comprising the following steps: S1: Extract the original image frames from the ground-based SAR amplitude and phase sequences, calculate the mean and deviation of the amplitude image according to the image size, perform difference expansion between phase frames and calculate the variance map, and after normalization, overlay them according to the regional label fusion logic to output the joint indicator layer map; S2: Divide the urban area and the vegetation mixed area into blocks in the joint indicator layer map, count the grayscale distribution frequency of pixels in the blocks to construct a grayscale histogram, select the frequency grayscale interval as the recognition range, logically judge the pixel value of each block to generate a Boolean mask map, and output the regional distribution mask map; S3: Based on the selected blocks in the regional distribution mask map, extract the original image to construct a grayscale co-occurrence matrix, obtain the grayscale correlation features of the blocks, detect the boundary closure and density ratio, eliminate the discontinuous boundary and sparse areas, and output the structure texture screening map; S4: extracting pixel coordinates from the structural texture screening graph, establishing distance associations between adjacent pixels, comparing path grayscale differences to determine neighboring rationality, constructing relationships between structural graph nodes and edges, and outputting a spatial connection graph structure object.

[0005] As a further solution of the present invention, the joint indicator layer map includes an amplitude statistical feature map, a phase difference feature map, a normalization processing map, and a regional fusion map; the regional distribution mask map includes an urban area identification mask, a vegetation mixed area mask, a grayscale frequency screening map, and a logical judgment result map; the structural texture screening map includes a grayscale correlation map, a boundary integrity map, a density distribution map, and a screening area map; the spatial connection map structure object includes a pixel node set, an adjacency relationship set, a path grayscale difference map, and a structural edge map.

[0006] As a further solution of the present invention, the definition of the frequency grayscale interval refers to the frequency of occurrence in the block grayscale histogram, representing the continuous grayscale value interval of the pixel grayscale distribution, and the judgment standard of the pixel value of the identification area; The definition of the boundary discontinuity and sparse area refers to the part in the structure texture map where the edge contour is discontinuous and the pixel density is lower than the set threshold, and the verification area is not complete and representative; The definition of establishing the distance association between adjacent pixels refers to determining the rationality of the connection by calculating the spatial distance between pixels and the grayscale change of the path, and constructing a node-edge network that reflects the spatial topological relationship.

[0007] As a further solution of the present invention, the specific steps of S1 are: S101: extracting original image frames from the ground-based SAR amplitude sequence and phase sequence, superimposing multiple frames of the grayscale value of each pixel in the amplitude image according to the image width, calculating the mean and deviation value based on the total number, cumulative value and square sum, and generating an amplitude grayscale statistics layer; S102: calling the amplitude grayscale statistics layer, calculating the inter-frame phase difference value for the phase image frames in time sequence, performing linear unwrapping processing, and calculating the variance based on the unwrapped value set of each pixel to generate a phase unwrapping variance layer; S103: performing normalization processing on the amplitude grayscale statistics layer and the phase unwrapped variance layer, setting a superposition rule according to the region label, performing weighted superposition on the normalized value of each pixel, and obtaining a joint indicator layer map.

[0008] As a further solution of the present invention, the specific steps of S2 are: S201: Based on the block division result of the urban area and the mixed vegetation area in the joint indicator layer image, the grayscale values ​​of all pixels in the block are counted, and a corresponding relationship between the grayscale value and the number of occurrences is established according to the grayscale level of the pixel, and the grayscale frequency distribution of each block is obtained to generate the grayscale frequency statistics of the block; S202: Based on the grayscale frequency statistics of the image block, extract the grayscale value range with the cumulative frequency in the middle section of the grayscale frequency distribution, determine whether the frequency span of continuous grayscale levels meets the threshold condition based on the frequency, obtain the upper and lower limits of the grayscale values ​​that meet the condition, and generate the grayscale recognition interval of the image block; S203: Call the grayscale identification interval of the block, perform logical judgment based on whether the grayscale value of each pixel in the block falls into the corresponding identification interval, mark the judgment result as a binary Boolean value, and construct a binary logic map according to the map size to obtain a regional distribution mask map.

[0009] As a further solution of the present invention, the specific steps of S3 are: S301: Calling the selected pixel block in the regional distribution mask map, obtaining the pixel gray value combination for the corresponding block in the original image, performing joint frequency statistics on all adjacent pixel gray pairs in each block, and constructing a co-occurrence matrix based on the gray pairs, statistically analyzing the distribution relationship of the gray pairs, and obtaining the gray correlation distribution characteristics of the blocks; S302: Based on the grayscale correlation distribution characteristics of the image block, perform boundary closure judgment on the grayscale adjacency relationship of the edge pixels in the image block, detect whether each edge point has a continuous grayscale connection, filter out the boundary pixel point set with missing connections, and generate a boundary closed pixel set; S303: Call the boundary closed pixel set, calculate the ratio of the total number of pixels to the number of candidate pixels for the remaining pixels in the block, determine whether the block meets the density threshold condition, eliminate the area blocks where the pixel distribution is lower than the density ratio requirement, and obtain the structural texture screening map.

[0010] As a further solution of the present invention, the specific steps of S4 are: S401: extracting the coordinate positions of the remaining pixels in the structural texture screening image, dividing the image into blocks according to the image size, extracting the spatial positions of the pixels in each block one by one, and calculating the spatial-grayscale composite distance value, establishing a pairing relationship between the pixels according to the distance threshold, and obtaining pixel distance association pairs; S402: calling the pixel distance association pair, comparing the grayscale value difference of the paired pixels, making a judgment based on the grayscale difference and a set grayscale difference threshold, filtering out pixel pairs with grayscale mutations, and retaining only pixel combinations that meet the adjacency condition, to obtain grayscale adjacent valid pairs; S403: Based on the grayscale adjacent valid pairs, the pixels are used as nodes of the structure graph, the adjacency relationship between the pixels is constructed as edges in the graph, all nodes and edges in the entire image block are uniformly organized, a connectivity network structure is established, and a spatial connection graph structure object is obtained.

[0011] As a further embodiment of the present invention, the method further comprises: S5: Based on the node positions and the number of connections in the spatial connection graph structure object, a starting pixel point set is set in the vegetation mixed boundary block, and full-image aggregation path selection and distribution balance judgment are performed. The path results are mapped to the original image to obtain a ground-based SAR high-coherence point selection scheme.

[0012] As a further solution of the present invention, the ground-based SAR high coherence point selection solution includes a clustering path graph, a node starting set, a distribution balance graph, and a clustering result graph.

[0013] As a further solution of the present invention, the specific steps of S5 are: S501: Calling all graph node positions and connection numbers in the spatial connection graph structure object, setting a starting pixel point set in the vegetation mixed boundary block, sorting the nodes from small to large by the number of connections, and selecting the starting point set as candidate nodes, tracing adjacent nodes in the graph structure in order by connection path length, and generating a candidate cluster path set; S502: Based on the candidate clustered path set, calculate the spatial distribution variance of all nodes in the path, and count the spatial average distances between nodes on the path, determine whether the distribution density balance meets the set spatial distribution balance threshold condition, and obtain a spatially balanced path set; S503: Call the spatially balanced path set, establish cluster numbers according to the pixel positions in the original image corresponding to the path nodes, divide the categories according to the paths and map the pixels in each path back to the corresponding positions in the original image, output the cluster number layer, and obtain the ground-based SAR high coherence point selection scheme.

[0014] Compared with the prior art, the advantages and positive effects of the present invention are: In the present invention, a multi-source fusion layer is constructed by combining amplitude and phase features to enhance the image discrimination dimension, the regional recognition accuracy is improved by grayscale frequency interval extraction and Boolean mask generation, the local structural features are extracted using the grayscale co-occurrence matrix and the abnormal pixels are eliminated by combining the boundary coherence and density features, and a pixel connection network is constructed based on the graph structure relationship based on the adjacency path. From this network, a set of pixels with aggregation and connectivity is screened to achieve high coherence point clustering output, thereby enhancing the integrity of the structural expression in complex areas and the aggregation rationality of target selection. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0016] Figure 1 Schematic diagram of the steps of the present invention; Figure 2 This is a schematic diagram of the refinement of S1 of the present invention; Figure 3 This is a schematic diagram of the refinement of S2 of the present invention; Figure 4 This is a schematic diagram of the refinement of S3 of the present invention; Figure 5 This is a schematic diagram of the refinement of S4 of the present invention; Figure 6 This is a schematic diagram of the refinement of S5 of the present invention. DETAILED DESCRIPTION

[0017] The technical solution of the present invention is described below in conjunction with the accompanying drawings.

[0018] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to indicate examples, illustrations, or explanations. Any embodiment or design described as an "exemplary" in the present invention should not be interpreted as being preferred or advantageous over other embodiments or designs. Rather, the use of the word "exemplary" is intended to present concepts in a concrete manner. Furthermore, in the embodiments of the present invention, "and / or" can mean both or either of the two.

[0019] In the embodiments of the present invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, when the distinction is not emphasized, the meanings they convey are the same. The terms "of," "corresponding," and "corresponding" may sometimes be used interchangeably. It should be noted that, when the distinction is not emphasized, the meanings they convey are the same.

[0020] In the embodiments of the present invention, sometimes a subscript such as W1 may be written as a non-subscript such as W1. When the difference is not emphasized, the meanings to be expressed are the same.

[0021] In order to make the technical problems, technical solutions and advantages to be solved by the present invention clearer, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0022] See also Figure 1 The embodiment of the present invention provides an improved method for selecting high coherence points of ground-based SAR, comprising the following steps: S1: Extract the original image frames from the ground-based SAR amplitude sequence and phase sequence, calculate the mean and deviation of the amplitude image according to the image size, perform difference expansion between the phase frames and calculate the variance map, normalize the amplitude and phase layers, perform overlay operation according to the fusion logic rules set by the regional label, and output the joint index layer map; S2: Based on the block division of urban and vegetation mixed areas in the joint indicator layer, the grayscale distribution frequency of pixels in the regional blocks is counted and a grayscale histogram is constructed. The frequency grayscale interval is selected as the recognition range, and pixel value logical judgment is performed on each block to generate a Boolean mask map, and the regional distribution mask map is output; S3: Call the selected pixel block in the regional distribution mask map, construct the grayscale co-occurrence matrix for the selected area in the original image, obtain the grayscale correlation distribution characteristics of the block, perform boundary closure detection and density ratio judgment on the blocks in the same area, eliminate the discontinuous boundaries and sparse candidate point areas, and output the structure texture screening map; S4: Extract the coordinate positions of the remaining pixels in the structural texture screening graph, establish distance association relationships between the pixels and their neighboring pixels, compare the grayscale differences of the pixels in the connection path and determine the adjacency rationality, combine the adjacency results into structural graph node and edge relationships, and output the spatial connection graph structure object; S5: Call the positions and number of connections of all graph nodes in the spatial connection graph structure object, set the starting pixel point set in the vegetation mixed boundary block, perform clustering path selection and distribution balance judgment on the entire graph structure, map the path results to the original image and output the cluster identification to obtain the ground-based SAR high coherence point selection scheme.

[0023] The joint indicator layer map includes the amplitude statistical feature map, the phase difference feature map, the normalization processing map, and the regional fusion map. The regional distribution mask map includes the urban area identification mask, the vegetation mixed area mask, the grayscale frequency screening map, and the logical judgment result map. The structural texture screening map includes the grayscale correlation map, the boundary integrity map, the density distribution map, and the screening area map. The spatial connection map structural objects include the pixel node set, the adjacency relationship set, the path grayscale difference map, and the structural edge map. The ground-based SAR high coherence point selection scheme includes the aggregation path map, the node starting set, the distribution balance map, and the clustering result map.

[0024] See also Figure 2 , the specific steps of S1 are: S101: extracting original image frames from the ground-based SAR amplitude sequence and phase sequence, superimposing multiple frames of the grayscale value of each pixel in the amplitude image according to the image width, calculating the mean and deviation value based on the total number, cumulative value and square sum, and generating an amplitude grayscale statistics layer; To extract the original image frames from the ground-based SAR amplitude sequence and phase sequence, it is necessary to load the SAR image data frame by frame, confirm the acquisition time and area range corresponding to each frame, and manage the maps covering the same geographical area with a unified number. For example, the map numbered A001 contains 15 frames of images. Then, a multi-frame overlay operation is performed on the grayscale value of each pixel in the amplitude image. Before the operation, all image frames in the map should be aligned at the pixel level to ensure that the (i, j) pixel position of each frame is consistent in all frames. For a pixel position (200, 300) in the map numbered A001, its grayscale values ​​in the 15 frames are 65, 68, 70, 71, 72, 75, 73, 69, 68, 66, 67, 70, 71, 74, and 72 respectively. The number of frames involved in the pixel point is 15, the total grayscale value is 1051, and the square grayscale value is 0. The total value is 73753. The mean grayscale value of the pixel is 1051 divided by 15, which is 70.07. Then its grayscale deviation is calculated. First, the difference between each grayscale value and the mean is calculated and squared to obtain a set of deviation square values. After summing them, the sum is divided by the number of frames to obtain the average deviation square value. The square root is then taken to obtain the deviation value. If some pixels in a certain image have missing data, for example, there is no value for the corresponding pixels in the 9th and 11th frames, the number of valid frames is corrected to 13. The total grayscale value and the sum of the square values ​​need to be recalculated after eliminating the invalid frame data. The grayscale value of each frame needs to be amplitude normalized to eliminate grayscale offset caused by non-target factors such as sensor gain. For example, the maximum grayscale value is set to 80, and all grayscale values ​​are scaled to between 0 and 1. The above superposition and deviation calculation are then performed to finally obtain an amplitude grayscale statistics layer containing the grayscale statistical characteristics of each pixel.

[0025] S102: calling the amplitude grayscale statistics layer, calculating the inter-frame phase difference value for the phase image frames in time sequence, performing linear unwrapping processing, and calculating the variance based on the unwrapped value set of each pixel to generate a phase unwrapping variance layer; Call the amplitude grayscale statistics layer, process the phase image frame frame by frame in the time series, select two consecutive frames as frame pairs for calculation, and use the 1st frame and the 2nd frame to form frame pair 1, the 2nd frame and the 3rd frame to form frame pair 2, and so on. Extract the phase value of the current frame and the next frame for each pixel position, perform subtraction to obtain the phase difference value. If the phase values ​​of a pixel in frame pair 1 are 3.1 and 3.3 respectively, the difference is 0.2. When there are periodic mutations in adjacent frames, linear expansion processing is required to remap the phase value into a continuous change curve. The conventional processing method is to accumulate the difference frame by frame from the first frame and perform trend fitting. If there is a jump exceeding π in frame pair 3, that is, about 3.14, it can be judged that there is a phase wrapping phenomenon, and the jump threshold can be set to 2.5. All inter-frame phase When the absolute value of the phase difference exceeds the threshold, the difference is adjusted to the interpolation prediction result of the phase change trend of the previous and next frames to ensure that the phase trajectory of each frame after reconstruction is continuous and stable. Taking a certain pixel as an example, the phase unwrapped values ​​obtained after processing are 2.9, 3.1, 3.2, 3.2, 3.3, 3.4, 3.3, 3.5, and 3.4 for a total of 9 frames. The average value is 3.26. The sum of the squares of the differences between each unwrapped value and the average value and divided by 9 to obtain a variance of 0.036. A small variance indicates that the pixel phase trajectory is stable. Here, the reference threshold for judging whether the variance is stable is set to 0.05. Pixels below this value are judged to be relatively stable, and pixels above this value are marked as unstable pixels. Each pixel in the variance layer corresponds to a specific value, forming a phase unwrapping variance layer.

[0026] S103: performing normalization processing on the amplitude grayscale statistics layer and the phase unwrapped variance layer, setting a superposition rule according to the region label, performing weighted superposition on the normalized value of each pixel, and obtaining a joint indicator layer map; Normalization is performed based on the amplitude grayscale statistic layer and the phase unwrapped variance layer. All pixel values ​​of the two layers are read, and their global minimum and maximum values ​​are calculated by region, which are used as the lower and upper limits of normalization respectively. For example, the maximum grayscale value of the amplitude layer is 78 and the minimum value is 45. The maximum value of the phase variance layer is 0.12 and the minimum value is 0.01. For each pixel, its grayscale value is subtracted from the minimum value and then divided by the difference between the maximum and minimum values. For example, the grayscale value of pixel A is 63, then the normalized value is (63-45) / (78-45)=0.545, and the corresponding phase variance value is 0.04, then the normalized value is (0.04-0.01) / (0.12-0.01)=0.273. Then, the overlap is set according to the regional label. Add weights. If pixel A belongs to the urban area, the amplitude normalization value weight is set to 0.7, the phase normalization value weight is set to 0.3, and the weighted superposition result is 0.7×0.545+0.3×0.273=0.4738. If pixel B belongs to the farmland area, the amplitude and phase normalization value weights are set to 0.4 and 0.6, and the weighted result under the same grayscale and phase values ​​is 0.4×0.545+0.6×0.273=0.3816. When setting the weights, the impact of regional differences on the sensitivity of the indicators is referred to. Generally, the weight range is determined to vary between 0.3 and 0.7 through field sample tests. Finally, the weighted result of each pixel is output as a joint indicator layer map. The value of each pixel reflects the degree of combination of its normalized attributes in the corresponding area.

[0027] See also Figure 3 , the specific steps of S2 are: S201: Based on the block division results of the urban area and the mixed vegetation area in the joint indicator layer image, the grayscale values ​​of all pixels in the block are counted, and a corresponding relationship between the grayscale value and the number of occurrences is established according to the grayscale level of the pixel, and the grayscale frequency distribution of each block is obtained to generate the grayscale frequency statistics of the block; Based on the block division results of urban areas and mixed vegetation areas in the joint indicator layer map, it is necessary to read the block boundary information in the layer map and establish a mapping table between the block number and the layer pixel coordinates. For example, the urban block numbers are set to U01 and U02, and the vegetation mixed area block numbers are V01 and V02. The grayscale values ​​of all pixels in the block U01 are extracted row by row to construct a grayscale value set. The grayscale value range is an integer between 0 and 255. Each pixel is counted once to form a grayscale frequency statistics table. For example, there are 3000 pixels in the U01 block, 400 pixels with a grayscale value of 85, and 430 pixels with a grayscale value of 86. In this way, the corresponding relationship between the grayscale value and the number of occurrences is constructed step by step, and the frequency statistics Each grayscale value in the statistics is used as a key, and the number of times it appears in the block is used as the value to form the grayscale frequency histogram distribution data. In the statistical process, the grayscale grading unit needs to be set. For example, each level is 1 or 2, that is, grayscale 85 and 86 are set as a group or counted separately. The method of 1 grayscale unit per level can obtain more refined distribution characteristics. If the number of pixels in block V02 is 2500, and the grayscale values ​​are mainly distributed between 110 and 140, then the corresponding grayscale frequency is dense in this interval. The frequency of each grayscale value in this interval is counted, and finally a statistical pair set with grayscale value as key and frequency as value is formed. Statistics are performed on each block independently, and data is not merged across blocks. Finally, the grayscale frequency statistical results of each block are output.

[0028] S202: Based on the grayscale frequency statistics of the image block, extract the grayscale value range with the cumulative frequency in the middle section of the grayscale frequency distribution, determine whether the frequency span of the continuous grayscale levels meets the threshold condition based on the frequency, obtain the upper and lower limits of the grayscale values ​​that meet the condition, and generate the grayscale recognition interval of the image block; According to the grayscale frequency statistics of the blocks, the grayscale value and frequency pair sets constructed in each block are processed in turn, and their frequencies are accumulated in ascending order of grayscale values ​​to obtain the cumulative frequency sequence, and the total frequency value is calculated. For example, the total number of pixels in block U01 is 3000. Sorted by cumulative frequency, the cumulative frequency of grayscale values ​​below 85 is 1000, accounting for about 33%, the cumulative frequency of grayscale values ​​between 86 and 120 is 1700, accounting for 56%, and the cumulative frequency of grayscale values ​​between 121 and 150 is 300, accounting for 10%. It is determined that the grayscale range with the cumulative frequency in the middle section is 86 to 120. The frequency span of all continuous grayscale levels in this interval is judged, that is, whether the frequency change between continuous grayscale levels meets the threshold condition. The frequency span threshold is set to 5%. If the frequency of a grayscale value is 200 and the frequency of the next grayscale value is 215, the frequency change The value of (215-200) / 3000 is approximately 0.5%, which does not meet the 5% threshold requirement. Therefore, this pair of values ​​is excluded. If the grayscale value is 110 and the corresponding frequency is 160, and the grayscale value 111 and the corresponding frequency are 220, the frequency change is 60 / 3000, or 2%, which still does not meet the threshold. Similarly, the condition is met only when the frequency change between consecutive grayscale levels is greater than or equal to 150 / 3000=5%. For example, the frequency span between grayscale values ​​95 and 96 is 180 and 330, and the corresponding change is 150 / 3000=5%, which meets the threshold condition. Then, 95 and 96 are included in the target grayscale interval, and all grayscale value segments that meet the frequency span threshold requirement in the middle segment are extracted. The minimum value is recorded as the lower limit of the grayscale value, and the maximum value is recorded as the upper limit of the grayscale value. Finally, the grayscale recognition interval corresponding to the block is obtained. For example, the recognition interval of block U01 is [95, 120].

[0029] S203: Calling the grayscale recognition interval of the image block, performing a logical judgment based on whether the grayscale value of each pixel in the image block falls into the corresponding recognition interval, marking the judgment result as a binary Boolean value, and constructing a binary logic map according to the image size to obtain a regional distribution mask map; Call the block grayscale recognition interval, perform the recognition operation on each block, read the upper and lower limit values ​​of the block recognition interval, for example, the recognition interval of block U01 is [95, 120], then traverse all the pixels in the block, read the pixel grayscale value one by one and make a judgment. The judgment method is to check whether the pixel grayscale value is within the recognition interval. If it falls into the interval, it is judged as True and marked as a logical value 1, otherwise it is False and marked as a logical value 0. For example, the grayscale value of pixel A in block U01 is 102, which falls into the recognition interval, the judgment value is 1, and the grayscale value of pixel B is 0. The value is 85, which is not in the identification interval, and the judgment value is 0. Each judgment operation involves only one numerical comparison and boundary judgment, and does not involve other statistical operations. All judgment results are recorded according to the corresponding position of the pixel in the block. The recorded result is a binary matrix with the same row and column dimensions as the original layer of the block. The judgment result matrices of all blocks are merged according to the map sheet number to generate a binary map of the entire map sheet. The corresponding identification interval is called for each block in the map sheet for logical judgment and marking. Finally, a complete binary logic map is constructed according to the map sheet number. After merging, the layer is identified as a regional distribution mask map.

[0030] See also Figure 4 , the specific steps of S3 are: S301: Calling the selected pixel block in the regional distribution mask map, obtaining the pixel gray value combination for the corresponding block in the original image, performing joint frequency statistics on all adjacent pixel gray pairs in each block, and constructing a co-occurrence matrix based on the gray pairs, statistically analyzing the distribution relationship of the gray pairs, and obtaining the gray correlation distribution characteristics of the block; Call the selected pixel block in the regional distribution mask map, read the area where all pixel values ​​are 1 in the mask map, extract the corresponding block number and pixel coordinate set, for example, the block number is T001, which contains 450 marked pixels, read the grayscale image data of block T001 from the original image, construct a grayscale value matrix for all pixels in the block, traverse each pixel in the matrix in turn to form a grayscale pair with its adjacent pixels, and the adjacency relationship is determined by the 8-neighborhood principle, that is, each pixel has at most 8 adjacent pixels. The grayscale value of pixel (100, 150) is 128, and the grayscale value of its adjacent pixel (100, 151) is 130, forming a grayscale pair (128, 130), record the frequency of joint occurrence of this grayscale pair, continue to traverse all pixels in the block downward and count the number of joint occurrences of all grayscale pairs, and accumulate the frequency of repeated grayscale pairs, for example The grayscale pair (128, 130) appeared 85 times, and the grayscale pair (130, 130) appeared 110 times. These grayscale pairs are used as row and column coordinates to construct a co-occurrence matrix. The row represents the first grayscale value and the column represents the second grayscale value. Each element of the co-occurrence matrix records the frequency of occurrence of the corresponding grayscale pair. The dimension of the co-occurrence matrix is ​​256×256. If only the actual grayscale combinations are counted, a sparse matrix representation can be established. Then, the frequency values ​​of all elements in the co-occurrence matrix are traversed and sorted from high to low according to the frequency of grayscale pairs. It can be observed that certain grayscale pairs appear frequently in the block, thereby identifying the strength of the grayscale value correlation in the block. For example, if the high-frequency grayscale pairs are mainly concentrated in the grayscale range of 120~135, it can be determined that the pixels inside the block have a strong grayscale coupling relationship within this grayscale range, and finally the grayscale correlation distribution characteristics of the block are output.

[0031] S302: Based on the grayscale correlation distribution characteristics of the block, perform boundary closure judgment on the grayscale adjacency relationship of the edge pixels in the block, detect whether each edge point has a continuous grayscale connection, filter out the boundary pixel point set with missing connections, and generate a boundary closed pixel set; According to the grayscale correlation distribution characteristics of the block, read the co-occurrence matrix counted in the previous step and screen the grayscale connection relationship of the edge pixels according to the grayscale value combination. First, all edge pixels in the block are identified. Edge pixels refer to pixels that are adjacent to the mask boundary in their 8-neighborhood. For example, the edge pixel point B1 in block T001 is located at coordinates (10, 100), and some of its adjacent pixels are located in the area where the mask value is 0. Extract the grayscale values ​​of this point and its 8-neighborhood to determine whether there is a continuous grayscale connection, that is, whether the frequency between the grayscale value of this point and the grayscale value of one of its adjacent points in the co-occurrence matrix is ​​greater than the connection threshold. The connection threshold is set to 10, that is, if the grayscale pair has a frequency of more than 10 times in the co-occurrence matrix, it is considered to be connected. For example, the grayscale value of point B1 is 122, and the grayscale of the adjacent point is 124. The frequency of the grayscale pair (122, 124) in the co-occurrence matrix is ​​15. If the condition is met, it is recorded as a connection. If the grayscale pair frequency is lower than the threshold, it is recorded as a connection loss. This judgment operation is performed on each edge pixel. If the connection frequency with all pixels in the 8-neighborhood does not exceed the threshold, that is, there is no continuous connection, then the edge point is removed from the block. The connection threshold used in the judgment can be set according to the overall grayscale distribution density of the block. If the total number of block pixels is 500 and the average frequency of co-occurrence grayscale pairs is 12, it is reasonable to set the connection threshold to 10. The connection threshold can be 8 for medium-density blocks, and the value can be increased to 15 for high-density blocks. Repeat the above operation for all edge pixels, and finally retain the edge pixels with at least one pair of continuous grayscale connections to construct the boundary closed pixel set of the block.

[0032] S303: Calling the boundary closed pixel set, calculating the ratio of the total number of pixels to the number of candidate pixels for the remaining pixels in the block, determining whether the block meets the density threshold condition, eliminating the blocks in the area where the pixel distribution is lower than the density ratio requirement, and obtaining the structure texture screening map; Call the boundary closed pixel set and perform density calculation on the remaining pixel sets in the block that have not been eliminated. First, count the total number of pixels in the block N_total and the number of candidate pixels in the boundary closed pixel set N_candidate. For example, the original total number of pixels in block T001 is 450, and the number of retained pixels in the closed set is 390. The density ratio is calculated as 390 divided by 450 to get 0.867. Then, this ratio is compared with the preset density threshold. The density threshold is set to 0.85. That is, when the proportion of candidate pixels exceeds 85%, the block is judged to meet the structural density requirements. If the ratio is lower than the threshold, it means that the internal structure of the block is too sparse. The tile needs to be marked as an area that does not meet the requirements and is removed from subsequent processing. The setting of the threshold refers to the texture concentration level of different structural areas. For urban areas with relatively concentrated structural textures, the density threshold can be set to 0.9, and for natural areas with slightly discrete textures, it can be set to 0.8. The value range is generally controlled between 0.75 and 0.95. During the judgment process, the ratio result of each tile needs to be directly compared with the corresponding threshold without the need for re-normalization or transformation. After traversing all tiles, the tiles that meet the density ratio requirements are recorded as retained areas to form a structural texture screening map. Only tile areas with complete structure and qualified density are retained in the output layer.

[0033] See also Figure 5 , the specific steps of S4 are: S401: Extract the coordinate positions of the remaining pixels in the structural texture screening image, divide the image into blocks according to the image size, extract the spatial position of each pixel in each block, calculate the spatial-grayscale composite distance value, establish a pairing relationship between the pixels according to the distance threshold, and obtain pixel distance association pairs; The calculation formula of the space-grayscale composite distance value is as follows: ; in, Indicates the first Pixels and The spatial-grayscale composite distance value between pixels, Indicates the The row coordinates of the pixels, Indicates the The row coordinates of the pixels, Indicates the The column coordinates of the pixels, Indicates the The column coordinates of the pixels, Indicates the The gray value of a pixel, Indicates the The gray value of a pixel, Represents the adjustment factor for converting grayscale difference into spatial pixel distance; Parameter description and value acquisition process: : The row and column coordinates of the pixel, which are directly read through the image index and are in pixels. For example, the position of pixel A is (42, 78) and that of pixel B is (46, 81). : is the grayscale value of the corresponding pixel, collected through the image grayscale channel, the unit is dimensionless (8-bit image, range 0–255), such as the grayscale of A is 138, and that of B is 122; : is the conversion coefficient between grayscale difference and spatial distance. Assuming the block size is 120×120 pixels, the maximum spatial distance is: ; The maximum grayscale difference is 255, so: ; Substitute into the calculation: Spatial distance term: ; Grayscale difference: ; Comprehensive distance: ; Table 1 Example of pixel composite distance calculation

[0034] As shown in Table 1, the composite distance between two pixels is 15.64 pixel units.

[0035] The results show that in block T201, although pixels A and B are only 5 pixels apart in space, the grayscale difference reaches 16, which results in an additional distance of 10.64 after conversion, causing the overall composite distance to rise to 15.64. This reflects that the pixel pair lacks grayscale continuity under the condition of spatial proximity and should not be regarded as a structural unit within the same texture area.

[0036] The benefit of the formula is that by introducing the grayscale conversion factor , achieving a unified expression of grayscale and spatial attributes under different dimensions, so that the calculation of structure-texture association relationships no longer depends on subjective weight settings, thereby enhancing the accuracy of the subsequent adjacent boundary establishment process.

[0037] S402: Calling pixel distance association pairs, comparing the grayscale value difference of paired pixels, and making a judgment based on the grayscale difference and the set grayscale difference threshold, filtering out pixel pairs with grayscale mutations, and retaining only pixel combinations that meet the adjacency condition, to obtain grayscale adjacent valid pairs; After calling the pixel distance association pair, read the grayscale value of each pair of pixels in the original image, perform a pair-by-pair comparison operation, obtain the grayscale difference of each pair of pixels, and then compare and judge with the set grayscale difference threshold. If the grayscale difference between the pixels is less than or equal to the threshold, the pixel pair is retained and considered to have grayscale continuity. Otherwise, it is eliminated. The grayscale difference threshold is set to 12. This threshold is set according to the grayscale dynamic range of the image and the local texture details. When the grayscale value range is 0 to 255, the grayscale ratio corresponding to 12 is about 4.7%. It is used to determine whether the local area maintains consistent texture. It is suitable for structural areas with medium contrast characteristics such as buildings and land features. For example For example, the grayscale of the pixel pair (105, 205)-(106, 206) is 132 and 140 respectively, and the grayscale difference is 8, which is less than the threshold of 12, so this pair is retained. The grayscale of the pixel pair (110, 210)-(111, 211) is 130 and 147 respectively, and the grayscale difference is 17, which exceeds the threshold, so it is deleted from the set. In the entire block G101, there are a total of 1180 pairs of pixel combinations. After grayscale difference judgment, the final remaining valid combinations are 980 pairs, and the remaining 200 pairs are eliminated due to drastic grayscale changes. All retained pixel pairs are marked as the grayscale adjacent valid pair set, which is used to construct the graph structure.

[0038] S403: Based on the grayscale adjacent valid pairs, the pixels are used as nodes in the structure graph, and the adjacency relationship between the pixels is constructed as edges in the graph. All nodes and edges in the entire image block are uniformly organized to establish a connectivity network structure and obtain a spatial connection graph structure object; According to the grayscale adjacency valid pairs, each pixel is set as a node element in the structure graph, and each retained grayscale adjacency pair is used as an edge in the graph structure to perform structural organization operations. When constructing the graph for block G101, a node number table is first generated. For example, the pixel (105, 205) is numbered G101_001, and (106, 206) is numbered G101_002. When there is a valid grayscale adjacency relationship, an edge connection is established between the two nodes. Each edge records the connection relationship without additional weight information. In block G101, 980 pairs of valid adjacency pairs will be It is constructed as 980 edges, and all related pixels totaling about 460 participate in at least one pair of connections. The number of connections for each pixel can be obtained by counting the number of times it appears in the adjacent pairs. For example, pixel G101_012 participates in 5 edge connections, that is, it has a spatial adjacency relationship with 5 pixels with consistent grayscale. All nodes and edges of block G101 are organized to form a local graph structure. Then, the graph structures of multiple blocks are merged according to the map sheet number to construct the spatial connection graph structure object corresponding to the entire image, which is used to express the connectivity status and spatial proximity characteristics between structural pixels.

[0039] See also Figure 6 , the specific steps of S5 are: S501: Calling all graph node positions and connection numbers in the spatial connection graph structure object, setting a starting pixel point set in the vegetation mixed boundary block, sorting the nodes from small to large by the number of connections, and selecting the starting point set as candidate nodes, tracing adjacent nodes in the graph structure in order by connection path length, and generating a candidate cluster path set; Call all the graph node positions and connection numbers in the spatial connection graph structure object, record the corresponding pixel position coordinates of each graph node and the number of connections with other nodes in the graph structure, for example, the corresponding coordinates of graph node A are (210, 310), the number of connections is 2, the coordinates of graph node B are (211, 309), the number of connections is 5, set the starting pixel point set in the vegetation mixed boundary block, select all the nodes within the boundary of the block as the initial screening range, and then sort these nodes from small to large according to the number of connections, and form the candidate starting point set of the nodes with the least number of connections. For example, set the candidate starting number to the first 5% of the nodes. If there are 200 boundary nodes in the block, then take The first 10 nodes with the least number of connections are used as the starting point set, for example, the node set is {A, B, C, D, E}. Then, in the graph structure, each starting node is used as the starting point to trace its adjacent nodes in turn. When constructing the path, a connection number sorting priority strategy is adopted, that is, priority is given to extending to adjacent nodes with a smaller number of connections. The tracking path is limited to a preset maximum path length, for example, set to 20 nodes, to prevent the path from being too long and resulting in a discrete distribution. At each tracking step, the node number and its spatial position are recorded to form a candidate path. This operation is repeated with all starting nodes as the starting point to finally generate a candidate clustered path set. For example, a total of 58 paths are generated, and each path records the node sequence and its spatial coordinate sequence.

[0040] S502: Based on the candidate clustered path set, calculate the spatial distribution variance of all nodes in the path, and count the spatial average distances between nodes on the path, determine whether the distribution density balance meets the set spatial distribution balance threshold condition, and obtain a spatially balanced path set; According to the set of candidate clustered paths, the spatial position coordinates of all nodes in each path are extracted and the spatial distribution variance of the path is calculated. During the statistical process, the standard deviations of the horizontal and vertical coordinates need to be calculated respectively to evaluate the spatial diffusion of the path. At the same time, the average distance from any node to all other nodes in the path is calculated to obtain the average distance between each node and other nodes in the path. The average of all average distance values ​​is then taken as the overall average distance index of the path. For example, a path P contains 10 nodes, the standard deviation of its X coordinate is 3.2, and the standard deviation of its Y coordinate is 2.8. The spatial distribution variance after merging is 3.0, the average spacing on the path is 4.5 pixels, the spatial distribution variance of the path is compared with the set spatial distribution balance threshold to determine whether the balance of the path meets the conditions. The threshold is set with reference to the image resolution and structural texture characteristics. For tile areas with larger spatial scales, the variance threshold can be set to no more than 5.0 and the average spacing can be set to no more than 6.0. Paths that meet both conditions are determined to be balanced paths, otherwise they are considered discrete paths. Path numbers that meet the requirements are screened from the path set. For example, paths P01, P03, and P06 meet the conditions and constitute a spatially balanced path set for subsequent cluster mapping operations.

[0041] S503: Call the spatially balanced path set, establish cluster numbers according to the pixel positions in the original image corresponding to the path nodes, divide the categories according to the paths, and map the pixels in each path back to the corresponding positions in the original image. Output the cluster number layer to obtain the ground-based SAR high coherence point selection solution; Call the spatially balanced path set, perform cluster numbering on each path in the set, assign values ​​starting from 1 according to the path number, take the nodes in each path as a cluster unit, read the original image pixel coordinate position corresponding to each node, and map the original image in the output layer in the form of cluster numbering. Each pixel is assigned a unique cluster ID value according to the path number to which it belongs. For example, the pixel coordinates corresponding to the node in path P01 are {(210, 310), (211, 309), (212, 308)}. In the output layer, the corresponding values ​​of these three positions are all set to 1, and the pixel in path P02 is set to 2. This is deduced inversely until all paths are mapped. After the numbering is completed on the entire image, a cluster number layer is formed. Each non-zero pixel in this layer corresponds to a type of clustered path area, and the pixels not participating in the path clustering are set to zero. The layer structure maintains pixel consistency with the original image, ultimately forming a ground-based SAR high coherence point selection scheme with pixel-level path aggregation identification.

[0042] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. An improved method for selecting high coherence points for ground-based SAR, characterized in that: The following steps are involved: S1: Extract the original image frames from the ground-based SAR amplitude and phase sequences, calculate the mean and deviation of the amplitude image according to the image size, perform difference expansion between phase frames and calculate the variance map, and after normalization, overlay them according to the regional label fusion logic to output the joint indicator layer map; S2: Divide the urban area and the vegetation mixed area into blocks in the joint indicator layer map, count the grayscale distribution frequency of pixels in the blocks to construct a grayscale histogram, select the frequency grayscale interval as the recognition range, logically judge the pixel value of each block to generate a Boolean mask map, and output the regional distribution mask map; S3: Based on the selected blocks in the regional distribution mask map, extract the original image to construct a grayscale co-occurrence matrix, obtain the grayscale correlation features of the blocks, detect the boundary closure and density ratio, eliminate the discontinuous boundary and sparse areas, and output the structure texture screening map; S4: extracting pixel coordinates from the structural texture screening graph, establishing distance associations between adjacent pixels, comparing path grayscale differences to determine neighboring rationality, constructing relationships between structural graph nodes and edges, and outputting a spatial connection graph structure object.

2. The improved ground-based SAR high coherence point selection method according to claim 1, characterized in that: The joint indicator layer map includes an amplitude statistical feature map, a phase difference feature map, a normalization processing map, and a regional fusion map; the regional distribution mask map includes an urban area identification mask, a vegetation mixed area mask, a grayscale frequency screening map, and a logical judgment result map; the structural texture screening map includes a grayscale correlation map, a boundary integrity map, a density distribution map, and a screening area map; the spatial connection map structure object includes a pixel node set, an adjacency relationship set, a path grayscale difference map, and a structural edge map.

3. The improved ground-based SAR high coherence point selection method according to claim 1, characterized in that: The definition of the frequency grayscale interval refers to the frequency of occurrence in the grayscale histogram of the block, the continuous grayscale value interval representing the grayscale distribution of the pixel, and the judgment standard for the pixel value of the identification area; The definition of the boundary discontinuity and sparse area refers to the part of the structure texture map where the density of the pixel with discontinuous edge contour is lower than the set threshold, and the verification area is not complete and representative; The definition of establishing the distance association between adjacent pixels refers to determining the rationality of the connection by calculating the spatial distance between pixels and the grayscale change of the path, and constructing a node-edge network that reflects the spatial topological relationship.

4. The improved ground-based SAR high coherence point selection method according to claim 1, characterized in that: The specific steps of S1 are: S101: extracting original image frames from the ground-based SAR amplitude sequence and phase sequence, superimposing multiple frames of the grayscale value of each pixel in the amplitude image according to the image width, calculating the mean and deviation value based on the total number, cumulative value and square sum, and generating an amplitude grayscale statistics layer; S102: calling the amplitude grayscale statistics layer, calculating the inter-frame phase difference value for the phase image frames in time sequence, performing linear unwrapping processing, and calculating the variance based on the unwrapped value set of each pixel to generate a phase unwrapping variance layer; S103: performing normalization processing on the amplitude grayscale statistics layer and the phase unwrapped variance layer, setting a superposition rule according to the region label, performing weighted superposition on the normalized value of each pixel, and obtaining a joint indicator layer map.

5. The improved ground-based SAR high coherence point selection method according to claim 1, characterized in that: The specific steps of S2 are: S201: Based on the block division result of the urban area and the mixed vegetation area in the joint indicator layer image, the grayscale values ​​of all pixels in the block are counted, and a corresponding relationship between the grayscale value and the number of occurrences is established according to the grayscale level of the pixel, and the grayscale frequency distribution of each block is obtained to generate the grayscale frequency statistics of the block; S202: Based on the grayscale frequency statistics of the image block, extract the grayscale value range with the cumulative frequency in the middle section of the grayscale frequency distribution, determine whether the frequency span of continuous grayscale levels meets the threshold condition based on the frequency, obtain the upper and lower limits of the grayscale values ​​that meet the condition, and generate the grayscale recognition interval of the image block; S203: Call the grayscale identification interval of the block, perform logical judgment based on whether the grayscale value of each pixel in the block falls into the corresponding identification interval, mark the judgment result as a binary Boolean value, and construct a binary logic map according to the map size to obtain a regional distribution mask map.

6. The improved ground-based SAR high coherence point selection method according to claim 1, characterized in that: The specific steps of S3 are: S301: Calling the selected pixel block in the regional distribution mask map, obtaining the pixel gray value combination for the corresponding block in the original image, performing joint frequency statistics on all adjacent pixel gray pairs in each block, and constructing a co-occurrence matrix based on the gray pairs, statistically analyzing the distribution relationship of the gray pairs, and obtaining the gray correlation distribution characteristics of the blocks; S302: Based on the grayscale correlation distribution characteristics of the image block, perform boundary closure judgment on the grayscale adjacency relationship of the edge pixels in the image block, detect whether each edge point has a continuous grayscale connection, filter out the boundary pixel point set with missing connections, and generate a boundary closed pixel set; S303: Call the boundary closed pixel set, calculate the ratio of the total number of pixels to the number of candidate pixels for the remaining pixels in the block, determine whether the block meets the density threshold condition, eliminate the area blocks where the pixel distribution is lower than the density ratio requirement, and obtain the structural texture screening map.

7. The improved ground-based SAR high coherence point selection method according to claim 1, characterized in that: The specific steps of S4 are: S401: extracting the coordinate positions of the remaining pixels in the structural texture screening image, dividing the image into blocks according to the image size, extracting the spatial positions of the pixels in each block one by one, and calculating the spatial-grayscale composite distance value, establishing a pairing relationship between the pixels according to the distance threshold, and obtaining pixel distance association pairs; S402: calling the pixel distance association pair, comparing the grayscale value difference of the paired pixels, making a judgment based on the grayscale difference and a set grayscale difference threshold, filtering out pixel pairs with grayscale mutations, and retaining only pixel combinations that meet the adjacency condition, to obtain grayscale adjacent valid pairs; S403: Based on the grayscale adjacent valid pairs, the pixels are used as nodes of the structure graph, the adjacency relationship between the pixels is constructed as edges in the graph, all nodes and edges in the entire image block are uniformly organized, a connectivity network structure is established, and a spatial connection graph structure object is obtained.

8. The improved ground-based SAR high coherence point selection method according to claim 1, characterized in that: The method further comprises: S5: Based on the node positions and the number of connections in the spatial connection graph structure object, a starting pixel point set is set in the vegetation mixed boundary block, and full-image aggregation path selection and distribution balance judgment are performed. The path results are mapped to the original image to obtain a ground-based SAR high-coherence point selection scheme.

9. The improved ground-based SAR high coherence point selection method according to claim 8, characterized in that: The ground-based SAR high coherence point selection scheme includes a clustering path graph, a node starting set, a distribution balance graph, and a clustering result graph.

10. The improved ground-based SAR high coherence point selection method according to claim 8, characterized in that: The specific steps of S5 are: S501: Calling all graph node positions and connection numbers in the spatial connection graph structure object, setting a starting pixel point set in the vegetation mixed boundary block, sorting the nodes from small to large by the number of connections, and selecting the starting point set as candidate nodes, tracing adjacent nodes in the graph structure in order by connection path length, and generating a candidate cluster path set; S502: Based on the candidate clustered path set, calculate the spatial distribution variance of all nodes in the path, and count the spatial average distances between nodes on the path, determine whether the distribution density balance meets the set spatial distribution balance threshold condition, and obtain a spatially balanced path set; S503: Call the spatially balanced path set, establish cluster numbers according to the pixel positions in the original image corresponding to the path nodes, divide the categories according to the paths and map the pixels in each path back to the corresponding positions in the original image, output the cluster number layer, and obtain the ground-based SAR high coherence point selection scheme.

Images