Intelligent analysis system of seismic data based on synthetic aperture radar technology

Abstract

The present invention relates to the field of synthetic aperture technology, and in particular to an intelligent analysis system for seismic data based on synthetic aperture radar technology, which can implement the following steps: based on synthetic aperture radar technology, obtaining a current SAR image corresponding to a to-be-seismic-detected area, and a historical clustering graph and a historical clustering cluster set corresponding to a historical SAR image set; determining a target distance feature between two pixel points; if the target distance feature between the current pixel point is less than or equal to a preset feature threshold, connecting these pixel points in a candidate clustering graph; determining the current degree of association between each two historical clustering clusters; correcting an original tolerance coefficient threshold; and splitting the current connectivity graph according to the target tolerance coefficient threshold through a connectivity graph dynamic splitting clustering algorithm to obtain a target clustering cluster. The present invention is based on synthetic aperture radar technology and improves the efficiency of clustering pixel points in multi-frame SAR images.

Description

Technical Field

[0001] The invention relates to the field of synthetic aperture technology, and in particular to a seismic data intelligent analysis system based on synthetic aperture radar technology. Background Art

[0002] The high temporal and spatial resolution of Synthetic Aperture Radar (SAR) enables it to detect tiny surface deformations, including surface subsidence or uplift caused by earthquakes, and capture potential earthquake-affected areas and disaster situations. It is of great significance for analyzing the impact of seismic activity on the surface, and can also provide useful information for post-disaster assessment and emergency rescue. Cluster analysis of SAR images collected at different times can help detect and monitor surface changes, and by identifying different patterns of surface changes, we can better understand the impact of seismic activity on the surface and provide more information for disaster prevention and response. Therefore, cluster analysis of SAR images collected at different times is a common seismic data analysis based on synthetic aperture radar technology. At present, the method commonly used for data clustering is to cluster all data that need to be clustered through the connected graph dynamic splitting clustering algorithm.

[0003] However, when clustering all pixels in all SAR images collected at different times using the connected graph dynamic splitting clustering algorithm, the following technical problems often occur:

[0004] Since the surface change monitoring process is often a continuous monitoring process, it is often necessary to continuously collect SAR images when monitoring surface changes. That is, SAR images are often updated in real time. Therefore, if each time a new SAR image is collected, the newly collected SAR image and all the pixels in the historical SAR image are clustered once through the connected graph dynamic split clustering algorithm, it will often cause the pixels in the historical SAR image to be clustered repeatedly, resulting in high computational redundancy, and further resulting in poor efficiency in clustering pixels in multiple frames of SAR images. Summary of the invention

[0005] In order to solve the technical problem of low efficiency in clustering pixel points in multi-frame SAR images, the present invention proposes a seismic data intelligent analysis system based on synthetic aperture radar technology.

[0006] The present invention provides a seismic data intelligent analysis system based on synthetic aperture radar technology, and the method executed by the system includes:

[0007] Based on synthetic aperture radar technology, the current SAR image corresponding to the area to be detected is obtained, as well as the historical clustering graph and the historical clustering cluster set corresponding to the historical SAR image set, and the pixel points in the current SAR image are added to the historical clustering graph to obtain the candidate clustering graph;

[0008] Determine the target distance feature between each current pixel point and each historical pixel point, and determine the target distance feature between the current pixel points according to the phase, radar wave wavelength and radar reflection intensity corresponding to each current pixel point and each historical pixel point, wherein the current pixel point is a pixel point in the current SAR image and the historical pixel point is a pixel point in the historical SAR image;

[0009] If the target distance feature between the pixel and the current pixel is less than or equal to the preset feature threshold, these pixels are connected in the candidate clustering graph to generate new connection edges and obtain the current clustering graph;

[0010] According to the connection between each two historical clustering clusters in the historical clustering cluster set and the current pixel point in the current clustering graph, the current correlation degree between the two historical clustering clusters is determined;

[0011] According to the number of connected edges and pixels in each current connected graph in the current clustering graph, determine the original tolerance coefficient threshold corresponding to each current connected graph;

[0012] According to the current correlation degree between all historical clusters in each current connected graph, the original tolerance coefficient threshold corresponding to each current connected graph is modified to obtain the target tolerance coefficient threshold corresponding to each current connected graph;

[0013] According to the target tolerance coefficient threshold corresponding to the current clustering graph and each current connected graph, each current connected graph is split through the connected graph dynamic split clustering algorithm to obtain the target cluster cluster.

[0014] Optionally, the obtaining of the current SAR image corresponding to the area to be detected by earthquake, and the historical clustering graph and the historical clustering cluster set corresponding to the historical SAR image set includes:

[0015] The SAR image of the area to be detected by earthquake at the current moment is obtained as the current SAR image, and the SAR image of the area to be detected by earthquake at each historical moment in the historical time period is obtained as the historical SAR image, so as to obtain a historical SAR image set;

[0016] Determine the target distance characteristics between any two historical pixel points based on the phase, radar wave wavelength and radar reflection intensity corresponding to the historical pixel points;

[0017] If the target distance feature between the historical pixels is less than or equal to the preset feature threshold, these historical pixels are connected to generate connection edges, and the graph obtained after all historical pixels are connected is used as the initial clustering graph, where the initial clustering graph includes all historical pixels;

[0018] The initial clustering graph is split by the connected graph dynamic splitting clustering algorithm, and all the pixels included in each connected graph obtained by splitting the initial clustering graph constitute a historical clustering cluster to obtain a historical clustering cluster set, and all the connected graphs obtained by splitting the initial clustering graph constitute a historical clustering graph.

[0019] Optionally, determining a target distance feature between any two pixel points includes:

[0020] Randomly determine any two pixel points as the first reference point and the second reference point. If the first reference point and the second reference point are pixel points in the same frame of SAR image, determine the target distance feature between the first reference point and the second reference point according to the phase, radar wave wavelength and radar reflection intensity corresponding to the first reference point and the second reference point, wherein the radar wave wavelength corresponding to all pixel points in all SAR images corresponding to the area to be detected by earthquake is equal;

[0021] If the first reference point and the second reference point are pixel points in different frames of SAR images, the SAR image where the first reference point is located is determined as the first SAR image, and the target distance characteristics between the first reference point and the second reference point are determined according to the phase corresponding to the pixel point in the first SAR image with the same position as the second reference point, the phase and radar wave wavelength corresponding to the first reference point, and the radar reflection intensity corresponding to the first reference point and the second reference point.

[0022] Optionally, the formula corresponding to the target distance feature between two pixels includes:

[0023] Jordi i Pixels and j pixels are pixels in the same frame of SAR image, then i Pixels and j The formula corresponding to the target distance feature between pixels is:

[0024] ;in, It is i Pixels and j Target distance feature between pixels; i and j is the sequence number of the pixel points in all SAR images corresponding to the earthquake detection area, and ; It is iThe value corresponding to the phase corresponding to each pixel point; It is j The value corresponding to the phase corresponding to each pixel point; is the arc value of 180°; It is the value corresponding to the wavelength of radar wave corresponding to the pixel point in the SAR image corresponding to the area to be detected by earthquake; It is i The value corresponding to the radar reflection intensity of each pixel point; It is j The value corresponding to the radar reflection intensity of each pixel point;

[0025] Jordi i Pixels and j pixels are pixels in different frames of SAR images, then i Pixels and j The formula corresponding to the target distance feature between pixels is:

[0026] ;in, It is i The SAR image where the pixel point is located is the same as the j The value corresponding to the phase corresponding to the pixel with the same pixel position.

[0027] Optionally, determining the current degree of association between each two historical clustering clusters in the historical clustering cluster set according to the connection between the two historical clustering clusters and the current pixel in the current clustering graph includes:

[0028] According to the connection between the pixels in the current clustering map, determine the target confidence corresponding to each current pixel;

[0029] Filter out the current pixel points that have connecting edges with the two historical clusters in the current cluster graph from all current pixel points as the intersection pixel points between the two historical clusters;

[0030] The current correlation degree between the two historical clusters is determined according to the target confidence corresponding to all the intersection pixels between the two historical clusters and the connection between the two historical clusters and all the intersection pixels between them in the current clustering graph.

[0031] Optionally, determining the target confidence corresponding to each current pixel point according to the connection status between the pixels in the current clustering map includes:

[0032] Determine any current pixel point as the current marked pixel point, and select from the historical cluster cluster set a historical cluster cluster that has a connection edge with the current marked pixel point in the current cluster graph as a reference cluster cluster of the current marked pixel point, thereby obtaining a reference cluster cluster set corresponding to the current marked pixel point;

[0033] Filtering historical pixel points connected to the current marked pixel point in the current clustering diagram from the reference cluster set corresponding to the current marked pixel point as candidate points of the current marked pixel point;

[0034] The target confidence corresponding to the currently marked pixel is determined according to the target distance characteristics between all candidate points of the currently marked pixel and the currently marked pixel in the reference cluster set corresponding to the currently marked pixel.

[0035] Optionally, the target confidence corresponding to the current pixel point and the formulas corresponding to the current correlation degree between the two historical clusters are:

[0036] ;

[0037] ;in, It is a The target confidence corresponding to the current pixel; a is the serial number of the current pixel; It is a The number of reference clusters in the reference cluster set corresponding to the current pixel point; r It is a The serial number of the reference cluster in the reference cluster set corresponding to the current pixel point; It is a The first one in the reference cluster set corresponding to the current pixel r The reference clusters and a The number of connected edges between the current pixels; It is an exponential function with a natural constant as base; It is a The first one in the reference cluster set corresponding to the current pixel r Among the reference clusters, a All candidate points of the current pixel and the a The mean of the target distance features between the current pixels; It is a In the reference cluster set corresponding to the current pixel point, a All candidate points of the current pixel and the a The mean of the target distance features between the current pixels; It isp The historical clusters and q The current degree of association between historical clusters; p and q is the sequence number of the historical cluster in the set of historical clusters, and ; It is p The historical clusters and q The number of intersection pixels between historical clusters; t It is p The historical clusters and q The sequence number of the intersection pixel between the historical clusters; is the normalization function; It is p The historical clusters and q The first t The target confidence corresponding to the intersection pixel; It is the absolute value function; It is p The historical clusters and q The first t The intersection pixel point and the p The first t The accumulated value of the target distance feature between all candidate points of the intersection pixels; It is p The historical clusters and q The first t The intersection pixel point and the q The first t The accumulated value of the target distance feature between all candidate points of the intersection pixels; It is p The historical clusters and q All the intersection pixels between the historical clusters and the q The number of connecting edges between historical clusters; It is p The historical clusters and q All the intersection pixels between the historical clusters and the p The number of connecting edges between historical clusters.

[0038] Optionally, the method of correcting the original tolerance coefficient threshold corresponding to each current connectivity graph according to the current degree of association between all historical clusters in each current connectivity graph to obtain the target tolerance coefficient threshold corresponding to each current connectivity graph includes:

[0039] According to the current correlation degree between all historical clusters in each current connected graph, the target weight corresponding to each current connected graph is determined;

[0040] The product of the target weight corresponding to each current connected graph and the original tolerance coefficient threshold is determined as the target tolerance coefficient threshold corresponding to each current connected graph.

[0041] Optionally, the formula corresponding to the target tolerance coefficient threshold corresponding to the current connectivity graph is:

[0042] ; ;in, It is the first h The target tolerance coefficient threshold corresponding to the current connected graph; h is the sequence number of the current connected graph in the current clustering graph; It is the first h The target weight corresponding to the current connected graph; It is the first h The original tolerance threshold corresponding to the current connected graph; is the normalization function; It is the first h The cumulative value of the current degree of association between all historical clusters in the current connectivity graph.

[0043] Optionally, determining the original tolerance coefficient threshold corresponding to each current connected graph according to the number of connected edges and pixels in each current connected graph in the current clustering graph includes:

[0044] The ratio of the number of connected edges to the number of pixels in each current connected graph is determined as a target ratio corresponding to each current connected graph, and the original tolerance coefficient threshold corresponding to each current connected graph is determined according to the target ratio corresponding to each current connected graph.

[0045] The present invention has the following beneficial effects:

[0046] The seismic data intelligent analysis system based on synthetic aperture radar technology of the present invention solves the technical problem of poor efficiency of clustering pixels in multi-frame SAR images based on synthetic aperture radar technology, and improves the efficiency of clustering pixels in multi-frame SAR images. First, the historical clustering graph and the historical clustering cluster set corresponding to the historical SAR image set are obtained, which can facilitate the subsequent clustering of pixels in the current SAR image and all historical SAR images on the basis of historical clustering. Secondly, the pixels in the current SAR image are added to the historical clustering graph, and only the connection between the pixels in the current SAR image and the historical pixels needs to be calculated later, without the need to calculate the connection between the historical pixels again, thereby reducing the amount of calculation to a certain extent, reducing the occupation of computing resources, and thus improving the efficiency of clustering pixels in multi-frame SAR images. Then, each pixel in the SAR image is usually composed of three main features: phase, radar wave wavelength and radar reflection intensity. Therefore, by comprehensively considering the phase, radar wave wavelength and radar reflection intensity, the target distance feature between the pixels can be quantified. The larger the value, the greater the feature difference between the pixels, and the more likely these pixels are not to belong to the same category. Next, the pixels that generate the connection edge are often the pixels with relatively small feature differences. Secondly, the current clustering graph obtained can characterize the connection between all pixels in the current SAR image and the historical SAR image set, and when determining the current clustering graph, the connection between the historical pixels is directly used without repeated calculation, thereby improving the calculation efficiency. Continuing, since the current pixel may affect the correlation between the historical clustering clusters, based on the connection between each two historical clustering clusters and the current pixel in the current clustering graph, the greater the current correlation between the two historical clustering clusters, the higher the correlation between the two historical clustering clusters under the influence of the current pixel. Afterwards, based on the current correlation between all historical clustering clusters in the current connectivity graph, the original tolerance coefficient threshold corresponding to the current connectivity graph is corrected, which can improve the accuracy of determining the target tolerance coefficient threshold corresponding to the current connectivity graph.Finally, based on the target tolerance coefficient threshold corresponding to the current clustering graph and each current connected graph, each current connected graph is split through the connected graph dynamic splitting clustering algorithm to obtain the target clustering cluster, thereby realizing the clustering of pixel points in the current SAR image and all historical SAR images. Compared with directly clustering all pixel points in the current SAR image and all historical SAR images through the connected graph dynamic splitting clustering algorithm, the present invention directly uses the connection status between historical pixel points when determining the current clustering graph, and does not perform repeated calculations again, thereby improving the clustering efficiency. Secondly, the present invention also considers the influence of the current pixel point on the correlation between the historical clustering clusters, and adaptively determines the target tolerance coefficient threshold corresponding to each current connected graph, thereby improving the accuracy of clustering all pixel points in the current SAR image and all historical SAR images. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. 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 paying creative work.

[0048] Figure 1 A flowchart of the method steps performed by the seismic data intelligent analysis system based on synthetic aperture radar technology of the present invention;

[0049] Figure 2 The diagram is a schematic diagram of the clustering process of the connected graph dynamic splitting clustering algorithm of the present invention.

[0050] in: Figure 2 (a) is the solid circle distribution diagram, Figure 2 (b) is the first clustering diagram. Figure 2 (c) in the figure is the second clustering graph. Figure 2 (d) in the figure is the third clustering diagram. Figure 2 (e) in the figure is the fourth clustering graph and Figure 2 (f) in the figure is the fifth clustering graph. DETAILED DESCRIPTION

[0051] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation methods, structures, features and effects of the technical solutions proposed by the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0052] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0053] refer to Figure 1 , shows a flow chart of the method steps performed by the seismic data intelligent analysis system based on synthetic aperture radar technology according to the present invention. The method performed by the seismic data intelligent analysis system based on synthetic aperture radar technology includes:

[0054] Step S1, based on synthetic aperture radar technology, obtain the current SAR image corresponding to the area to be earthquake detected, as well as the historical clustering map and historical cluster cluster set corresponding to the historical SAR image set, and add the pixel points in the current SAR image to the historical clustering map to obtain a candidate clustering map.

[0055] Among them, Synthetic Aperture Radar (SAR) is also called integrated aperture radar. The area to be detected by earthquakes may be an area to be detected by earthquakes. The current SAR image may be a SAR image of the area to be detected by earthquakes at the current moment. The current moment is also called the current time. The historical SAR images in the historical SAR image set may be SAR images of the area to be detected by earthquakes collected within a historical time period. The historical time period may be a time period preset before the current moment. The historical clustering graph may be a graph composed of multiple connected graphs obtained by splitting the initial clustering graph through a connected graph dynamic splitting clustering algorithm. The initial clustering graph may be a graph composed of all pixel points in the historical SAR image set for subsequent dynamic splitting clustering of connected graphs. The connected graph dynamic splitting clustering algorithm is also called the CABDDCG algorithm.

[0056] It should be noted that obtaining the historical clustering graph and the historical clustering cluster set corresponding to the historical SAR image set can facilitate the subsequent clustering of the pixels in the current SAR image and all historical SAR images based on the historical clustering. Secondly, the pixels in the current SAR image are added to the historical clustering graph. Subsequently, only the connection between the pixels in the current SAR image and the historical pixels needs to be calculated, without the need to calculate the connection between the historical pixels again. This reduces the amount of calculation to a certain extent, reduces the occupation of computing resources, and thus improves the efficiency of clustering pixels in multi-frame SAR images.

[0057] As an example, this step may include the following steps:

[0058] The first step is to obtain the SAR image of the area to be earthquake detected at the current moment through synthetic aperture radar as the current SAR image, and obtain the SAR image of the area to be earthquake detected at each historical moment in the historical time period as the historical SAR image, and obtain a historical SAR image set.

[0059] Among them, the historical moment can be a moment within a historical time period.

[0060] In the second step, the target distance characteristics between any two historical pixel points are determined based on the phase, radar wave wavelength and radar reflection intensity corresponding to the historical pixel points.

[0061] The historical pixel points may be pixel points in historical SAR images.

[0062] For example, determining the target distance feature between any two pixels may include the following sub-steps:

[0063] In the first sub-step, any two pixel points are randomly determined as the first reference point and the second reference point. If the first reference point and the second reference point are pixel points in the same frame of SAR image, the target distance characteristics between the first reference point and the second reference point are determined according to the phase, radar wave wavelength and radar reflection intensity corresponding to the first reference point and the second reference point.

[0064] It should be noted that in SAR images, each pixel is usually composed of three main features: phase, radar wave wavelength and radar reflection intensity. Phase information contains the position information of the target surface relative to the transmitting radar, and in SAR images, phase is usually represented as grayscale or color to reflect the morphology and characteristics of the target surface. Radar reflection intensity is usually used to represent the reflectivity of the target surface or the characteristics of the object surface, such as surface roughness, electromagnetic properties of the object, etc., and in SAR images, radar reflection intensity is usually represented as grayscale or color to show the brightness of the target surface. Since radar wave wavelength refers to the wavelength of the radar wave used for SAR imaging, the radar wave wavelength corresponding to all pixels in all SAR images corresponding to the area to be detected by earthquake collected by the same synthetic aperture radar can be equal.

[0065] In the second sub-step, if the first reference point and the second reference point are pixel points in different frames of SAR images, the SAR image where the first reference point is located is determined as the first SAR image, and the target distance characteristics between the first reference point and the second reference point are determined according to the phase corresponding to the pixel point at the same position as the second reference point in the first SAR image, the phase and radar wave wavelength corresponding to the first reference point, and the radar reflection intensity corresponding to the first reference point and the second reference point.

[0066] For example, the formula corresponding to the target distance feature between two pixels includes:

[0067] Jordi i Pixels and j pixels are pixels in the same frame of SAR image, then i Pixels and j The formula corresponding to the target distance feature between pixels can be:

[0068] ;in, It is i Pixels and j The target distance feature between pixels. i and j is the sequence number of the pixel points in all SAR images corresponding to the earthquake detection area, and . It is i The value corresponding to the phase of each pixel. It is j The value corresponding to the phase of each pixel. It is a radian value of 180°. It is the value corresponding to the wavelength of the radar wave corresponding to the pixel point in the SAR image corresponding to the area to be detected by the earthquake. It is i The value corresponding to the radar reflection intensity of each pixel. It is j The value corresponding to the radar reflection intensity of each pixel.

[0069] It should be noted that when The larger the value, the more likely it is that the i Pixels and j The larger the phase difference between pixels, the larger the radar wave wavelength; this often indicates that the first i Pixels and j The greater the feature difference between the pixels, the more likely it is that the i Pixels and j The more pixels there are, the more likely they are not of the same class. The larger the i Pixels and j The greater the difference in radar reflection intensity between pixels, the greater the difference in radar reflection intensity between pixels. i Pixels and j The more pixels there are, the more likely they are not in the same category. The larger the i Pixels and jThe greater the difference between pixels, the greater the i Pixels and j The more likely the pixels are not of the same class.

[0070] Jordi i Pixels and j pixels are pixels in different frames of SAR images, then i Pixels and j The formula corresponding to the target distance feature between pixels can be:

[0071] ;in, It is i The SAR image where the pixel point is located is the same as the j The value corresponding to the phase corresponding to the pixel with the same pixel position.

[0072] It should be noted that since the phases at different times cannot be directly compared, the phase difference between the two points is often calculated by comparing the phase of the pixel at the next moment to the position at the previous moment. The larger the value, the more likely it is that the i Pixels and j The larger the phase difference between pixels, the larger the radar wave wavelength; this often indicates that the i Pixels and j The greater the difference in features between pixels, the more likely it is that the pixels collected at different times i Pixels and j The more likely the pixels are not of the same class.

[0073] In the third step, if the target distance feature between the historical pixel points is less than or equal to the preset feature threshold, these historical pixel points are connected to generate connection edges, and the graph obtained after all historical pixel points are connected is used as the initial clustering graph.

[0074] The initial clustering graph may include all historical pixel points. The preset feature threshold may be a pre-set threshold. For example, the preset feature threshold may be obtained by descending the target distance features between all historical pixel points, and determining the target distance feature at the quarter position in the obtained sequence as the preset feature threshold. The connecting edge may be an edge obtained by connecting two pixel points.

[0075] For example, Figure 2 As shown, the solid dots can represent historical pixels. Figure 2 The solid circle distribution diagram (a) in the figure represents the distribution of historical pixel points. Figure 2The first cluster graph in (b) can characterize the initial cluster graph.

[0076] The fourth step is to split the initial clustering graph through the connected graph dynamic splitting clustering algorithm, and form historical clustering clusters with all the pixels included in each connected graph obtained by splitting the initial clustering graph, and obtain a set of historical clustering clusters, and form a historical clustering graph with all the connected graphs obtained by splitting the initial clustering graph. At this time, the connected graphs obtained by splitting the historical clustering graph can correspond one-to-one to the historical clustering clusters.

[0077] Among them, the connected graph dynamic splitting clustering algorithm is a clustering method based on graph theory and dynamic splitting.

[0078] For example, if the first historical connected graph is a connected graph obtained by splitting the initial clustering graph, then all the pixels included in the first historical connected graph can constitute a historical clustering cluster.

[0079] For example, Figure 2 As shown in , if the solid dots represent historical pixels, Figure 2 The solid circle distribution diagram (a) represents the distribution of historical pixel points, and Figure 2 The first cluster graph (b) in represents the initial cluster graph, then Figure 2 The second cluster graph (c) in can represent the historical cluster graph, and Figure 2 There are 5 connected graphs in the historical clustering graph represented by the second clustering graph (c) in the figure. All the pixels in each of the 5 connected graphs constitute a historical clustering cluster. Figure 2 By splitting the initial clustering graph represented by the first clustering graph (b) in , a historical clustering graph containing 5 connected graphs can be obtained, and 5 historical clustering clusters can be obtained.

[0080] The fifth step is to add the pixels in the current SAR image to the historical clustering map to obtain the candidate clustering map.

[0081] For example, Figure 2 As shown in the figure, if the solid circle represents the historical pixel point, and the solid square represents the current pixel point, Figure 2 The solid circle distribution diagram (a) represents the distribution of historical pixel points. Figure 2 The first clustering graph (b) in represents the initial clustering graph, Figure 2 The second cluster graph (c) in represents the historical cluster graph, then Figure 2 The third clustering graph (d) in can characterize the candidate clustering graph.

[0082] Step S2, determining the target distance characteristics between each current pixel and each historical pixel according to the phase, radar wave wavelength and radar reflection intensity corresponding to each current pixel and each historical pixel, and determining the target distance characteristics between the current pixels.

[0083] The current pixel point may be a pixel point in the current SAR image, and the historical pixel point may be a pixel point in the historical SAR image.

[0084] It should be noted that each pixel in a SAR image is usually composed of three main features: phase, radar wave wavelength and radar reflection intensity. Therefore, by comprehensively considering the phase, radar wave wavelength and radar reflection intensity, the target distance characteristics between pixels can be quantified. The larger the value, the greater the characteristic difference between the pixels, which often means that these pixels are more likely to not belong to the same category.

[0085] As an example, the target distance feature between the current pixel point and the historical pixel points, and the target distance feature between any two current pixel points can be determined by the method for determining the target distance feature between any two pixel points listed as an example in the second step included in step S1.

[0086] Step S3: If the target distance feature between the pixel and the current pixel is less than or equal to the preset feature threshold, these pixel points are connected in the candidate clustering graph to generate new connection edges to obtain the current clustering graph.

[0087] The current clustering graph may be a graph formed by connecting pixels in the candidate clustering graph whose target distance feature between the current pixel and the pixel is less than or equal to a preset feature threshold.

[0088] It should be noted that the pixel points that generate connecting edges are often pixel points with relatively small feature differences. Secondly, the current clustering graph obtained can characterize the connection between all pixel points in the current SAR image and the historical SAR image set, and when determining the current clustering graph, the connection between the historical pixel points is directly used without repeated calculation, thereby improving the calculation efficiency.

[0089] As an example, if the target distance feature between the current pixel and the historical pixel is less than or equal to the preset feature threshold, the current pixel and the historical pixel are connected in the candidate clustering map. If the target distance feature between two current pixels is less than or equal to the preset feature threshold, the two current pixels are connected in the candidate clustering map.

[0090] For example, Figure 2 As shown in the figure, if the solid circle represents the historical pixel point, and the solid square represents the current pixel point, Figure 2 The solid circle distribution diagram (a) represents the distribution of historical pixel points. Figure 2 The first clustering graph (b) in represents the initial clustering graph, Figure 2 The second cluster graph (c) in represents the historical cluster graph, and Figure 2The third cluster graph (d) in represents the candidate cluster graph, then Figure 2 The fourth clustering graph (e) in the graph can represent the current clustering graph, representing the current clustering graph Figure 2 Each connected graph in the fourth clustering graph (e) in can be recorded as the current connected graph, Figure 2 The fourth clustering graph (e) in FIG. 5 represents a current clustering graph in which there are two current connected graphs.

[0091] Step S4, determining the current correlation degree between each two historical clusters in the historical cluster set according to the connection between the two historical clusters and the current pixel in the current cluster map.

[0092] It should be noted that since the current pixel point may affect the correlation between historical clustering clusters, based on the connection between each two historical clustering clusters and the current pixel point in the current clustering map, the greater the current correlation degree between the two historical clustering clusters, the higher the correlation between the two historical clustering clusters under the influence of the current pixel point.

[0093] As an example, this step may include the following steps:

[0094] The first step is to determine the target confidence corresponding to each current pixel according to the connection between the pixels in the current clustering map, which can include the following sub-steps:

[0095] In the first sub-step, any current pixel point is determined as the current marked pixel point, and the historical cluster clusters that have connecting edges with the current marked pixel point in the current cluster graph are selected from the above-mentioned historical cluster cluster set as the reference cluster clusters of the above-mentioned current marked pixel point, thereby obtaining the reference cluster cluster set corresponding to the above-mentioned current marked pixel point.

[0096] In the second sub-step, historical pixels connected to the current marked pixel point in the current clustering diagram are selected from the reference cluster set corresponding to the current marked pixel point as candidate points of the current marked pixel point.

[0097] The third sub-step is to determine the target confidence corresponding to the current marked pixel point according to the target distance characteristics between all candidate points of the current marked pixel point and the current marked pixel point in the reference cluster set corresponding to the current marked pixel point.

[0098] In the second step, the current pixel points that have connecting edges with the two historical clusters in the current clustering graph are selected from all the current pixel points as the intersection pixel points between the two historical clusters.

[0099] Among them, the intersection pixel point between two historical clusters is the current pixel point that has a connection edge with both historical clusters.

[0100] The third step is to determine the current degree of association between the two historical clusters based on the target confidence corresponding to all intersection pixels between the two historical clusters and the connection between the two historical clusters and all intersection pixels between them in the current clustering graph.

[0101] For example, the formulas for determining the target confidence corresponding to the current pixel point and the current correlation degree between two historical clusters are:

[0102] ;

[0103] ;in, It is a The target confidence corresponding to the current pixel. a Is the sequence number of the current pixel. It is a The number of reference clusters in the reference cluster set corresponding to the current pixel. r It is a The sequence number of the reference cluster in the reference cluster set corresponding to the current pixel. It is a The first one in the reference cluster set corresponding to the current pixel r The reference clusters and a The number of connected edges between the current pixels. It is an exponential function with a natural constant as its base. It is a The first one in the reference cluster set corresponding to the current pixel r Among the reference clusters, a All candidate points of the current pixel and the a The mean of the target distance features between the current pixels. It is a In the reference cluster set corresponding to the current pixel point, a All candidate points of the current pixel and the a The mean of the target distance features between the current pixels. It is p The historical clusters and q The current degree of association between historical clusters. p and q is the sequence number of the historical cluster in the set of historical clusters, and . It is p The historical clusters and q The number of intersection pixels between historical clusters. tIt is p The historical clusters and q The sequence number of the intersection pixel between the historical clusters. is a normalization function. It is p The historical clusters and q The first t The target confidence corresponding to the intersection pixel. It is the absolute value function. It is p The historical clusters and q The first t The intersection pixel point and the p The first t The accumulated value of the target distance features between all candidate points of the intersection pixels. It is p The historical clusters and q The first t The intersection pixel point and the q The first t The accumulated value of the target distance features between all candidate points of the intersection pixels. It is p The historical clusters and q All the intersection pixels between the historical clusters and the q The number of connecting edges between historical clusters. It is p The historical clusters and q All the intersection pixels between the historical clusters and the p The number of connecting edges between historical clusters.

[0104] It should be noted that when The larger the a The current pixel and its corresponding r The more connections between the reference clusters, the more a The current pixel and its corresponding r The greater the correlation between the reference clusters. Can be used as The weight of The smaller the time, the more likely it is that a The current pixel and its corresponding r The first a The average of the target distance features between all candidate points of the current pixel, the closer to the aThe current pixel and its corresponding reference clusters a The average of the target distance features between all candidate points of the current pixel; it often indicates that the a The current pixel and its corresponding r The closer the distance between reference clusters is to the average distance between all corresponding reference clusters, the closer the distance between reference clusters is to the average distance between all corresponding reference clusters. The larger the a The distance difference between the current pixel and its corresponding multiple reference clusters with large correlation is relatively small, which often indicates that the a The current pixel has a greater impact on multiple historical clusters, which often indicates that it should be a The current pixel is given a higher degree of attention, which often indicates that the a The higher the confidence of the current pixel, the higher the confidence of the current pixel. The larger the p The historical clusters and q The first t The influence of the intersection pixels on multiple historical clusters is relatively large, which often indicates that the p The historical clusters and q The first t The higher the confidence of the intersection pixel, the higher the confidence of the intersection pixel. Can be used as The weight of . Can characterize the t The intersection pixel pair p The impact of historical clusters, and when The smaller the time, the more likely it is that p The historical clusters and q The first t The intersection pixel point and the p The historical clusters and t The target distance feature between the historical pixels connected by the intersection pixels is relatively smaller. Can characterize the t The intersection pixel pair q The impact of historical clusters, and when The smaller the time, the more likely it is that p The historical clusters and q The first t The intersection pixel point and the q The historical clusters and t The target distance feature between the historical pixels connected by the intersection pixels is relatively small. The smaller the time, the more likely it is that tThe intersection pixel pair q The historical clusters and p The smaller the difference in influence between the historical clusters, the more q The historical clusters and p The more historical clusters are relatively related. The larger the p The historical clusters and q There are more intersection pixels with high confidence between the historical clusters. q The historical clusters and p The difference in influence between the historical clusters is small, which often indicates that q The historical clusters and p The more relevant the historical clusters are, the more likely they are to be q The historical clusters and p The more historical clusters should be merged into the same cluster.

[0105] Step S5, determining the original tolerance coefficient threshold corresponding to each current connected graph according to the number of connected edges and pixels in each current connected graph in the current clustering graph.

[0106] The current connectivity graph may be a connectivity graph in the current clustering graph. The original tolerance coefficient threshold may be a preliminarily calculated tolerance coefficient threshold. The tolerance coefficient threshold is also called a splitting threshold.

[0107] It should be noted that in the process of dynamic splitting and clustering of the connectivity graph, it is often necessary to quantify the tolerance coefficient threshold for subsequent connectivity graph splitting.

[0108] As an example, the ratio of the number of connected edges to the number of pixels in each current connected graph can be determined as the target ratio corresponding to each current connected graph, and the original tolerance coefficient threshold corresponding to each current connected graph can be determined based on the target ratio corresponding to each current connected graph.

[0109] For example, the formula for determining the original tolerance threshold corresponding to the current connectivity graph can be:

[0110] ;in, It is the first h The original tolerance coefficient threshold corresponding to the current connected graph. h is the sequence number of the current connected graph in the current clustering graph. It is an exponential function with a natural constant as its base. It is the first h The number of connected edges in the current connected graph. It is the first hThe number of pixels in the current connected graph. It is the first h The target ratio corresponding to the current connected graph.

[0111] It should be noted that It can be roughly characterized that h The tolerance coefficient threshold of the current connected graph when splitting. The larger the value, the less likely it is to be split.

[0112] Step S6, according to the current correlation degree between all historical clusters in each current connected graph, the original tolerance coefficient threshold corresponding to each current connected graph is corrected to obtain the target tolerance coefficient threshold corresponding to each current connected graph.

[0113] The target bearing coefficient threshold may be the bearing coefficient threshold finally calculated.

[0114] It should be noted that, based on the current correlation degree between all historical clusters in the current connectivity graph, the original tolerance coefficient threshold corresponding to the current connectivity graph is corrected, which can improve the accuracy of determining the target tolerance coefficient threshold corresponding to the current connectivity graph.

[0115] As an example, this step may include the following steps:

[0116] In the first step, the target weight corresponding to each current connected graph is determined according to the current degree of association between all historical clustering clusters in each current connected graph.

[0117] In the second step, the product of the target weight corresponding to each current connected graph and the original tolerance coefficient threshold is determined as the target tolerance coefficient threshold corresponding to each current connected graph.

[0118] For example, the formula for determining the target tolerance coefficient threshold corresponding to the current connectivity graph can be:

[0119] ; ;in, It is the first h The target tolerance coefficient threshold corresponding to the current connected graph. h is the sequence number of the current connected graph in the current clustering graph. It is the first h The target weight corresponding to the current connected graph. It is the first h The original tolerance coefficient threshold corresponding to the current connected graph. is a normalization function. It is the first h The cumulative value of the current degree of association between all historical clusters in the current connectivity graph.

[0120] It should be noted that It can be roughly characterized that h The tolerance coefficient threshold for splitting the current connected graph, the larger the value, the less likely it is to be split. The larger the h The greater the current correlation between the historical clusters in the current connectivity graph, the greater the current correlation between the historical clusters in the current connectivity graph. h There are more intersection pixels with high confidence between the historical clusters in the current connectivity graph, and the impact difference between these historical clusters is small, which often indicates that the h The more relevant the historical clusters in the current connectivity graph are, the less likely it is that the first h The current connected graph is split. So, when The larger the h The more relevant the historical clusters in the current connectivity graph are, the less likely it is that the first h The current connected graph is split. Therefore, by right Corrected It can accurately characterize the h The tolerance coefficient threshold of the current connected graph when splitting. The larger the value, the less likely it is to be split.

[0121] Step S7, according to the target tolerance coefficient threshold corresponding to the current clustering graph and each current connected graph, each current connected graph is split by a connected graph dynamic splitting clustering algorithm to obtain a target cluster cluster.

[0122] It should be noted that, based on the target tolerance coefficient threshold corresponding to the current clustering graph and each current connected graph, each current connected graph is split through the connected graph dynamic splitting clustering algorithm to obtain the target clustering cluster, thereby realizing the clustering of pixels in the current SAR image and all historical SAR images.

[0123] As an example, each current connected graph can be split according to the target tolerance coefficient threshold corresponding to the current clustering graph and each current connected graph through the connected graph dynamic splitting clustering algorithm, and all pixel points in each sub-connected graph obtained by splitting each current connected graph constitute a target cluster cluster.

[0124] It should be noted that the final clustering result is the clustering of each point in the SAR data at different times. Similar surface deformation patterns are classified into one category. By changing the clustering results of each cluster at different times, the deformation of the surface can be detected, thereby helping to identify potential seismic activities.

[0125] For example, Figure 2As shown in the figure, if the solid circle represents the historical pixel point, and the solid square represents the current pixel point, Figure 2 The solid circle distribution diagram (a) represents the distribution of historical pixel points. Figure 2 The first clustering graph (b) in represents the initial clustering graph, Figure 2 The second cluster graph (c) in represents the historical cluster graph, and Figure 2 The third clustering graph (d) in represents the candidate clustering graph, Figure 2 The fourth clustering graph (e) in represents the current clustering graph, Figure 2 There are two current connected graphs in the current clustering graph represented by the fourth clustering graph (e). According to the target tolerance coefficient threshold corresponding to each current connected graph in the current clustering graph, each current connected graph in the current clustering graph is split by the connected graph dynamic splitting clustering algorithm. The final splitting result can be obtained as follows: Figure 2 The fifth cluster graph (f) in represents, Figure 2 All pixels in each connected graph in the fifth clustering graph (f) can form a target cluster.

[0126] In summary, compared with directly clustering all pixel points in the current SAR image and all historical SAR images through the dynamic splitting clustering algorithm of the connectivity graph, the present invention directly uses the connection status between historical pixel points when determining the current clustering graph, and does not repeat the calculation again, thereby improving the clustering efficiency. Secondly, the present invention also considers the influence of the current pixel point on the correlation between the historical clustering clusters, and adaptively determines the target tolerance coefficient threshold corresponding to each current connectivity graph, thereby improving the accuracy of clustering all pixel points in the current SAR image and all historical SAR images.

[0127] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments can still be modified, or some of the technical features can be replaced by equivalents. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A seismic data intelligent analysis system based on synthetic aperture radar technology, comprising a memory and a processor, characterized in that: The processor executes the computer program stored in the memory to implement the following steps: Based on synthetic aperture radar technology, the current SAR image corresponding to the area to be detected is obtained, as well as the historical clustering graph and the historical clustering cluster set corresponding to the historical SAR image set, and the pixel points in the current SAR image are added to the historical clustering graph to obtain the candidate clustering graph; Determine the target distance feature between each current pixel point and each historical pixel point, and determine the target distance feature between the current pixel points according to the phase, radar wave wavelength and radar reflection intensity corresponding to each current pixel point and each historical pixel point, wherein the current pixel point is a pixel point in the current SAR image and the historical pixel point is a pixel point in the historical SAR image; If the target distance feature between the pixel and the current pixel is less than or equal to the preset feature threshold, these pixels are connected in the candidate clustering graph to generate new connection edges and obtain the current clustering graph; According to the connection between each two historical clustering clusters in the historical clustering cluster set and the current pixel point in the current clustering graph, the current correlation degree between the two historical clustering clusters is determined; According to the number of connected edges and pixels in each current connected graph in the current clustering graph, determine the original tolerance coefficient threshold corresponding to each current connected graph; According to the current correlation degree between all historical clusters in each current connected graph, the original tolerance coefficient threshold corresponding to each current connected graph is modified to obtain the target tolerance coefficient threshold corresponding to each current connected graph; According to the target tolerance coefficient threshold corresponding to the current clustering graph and each current connected graph, each current connected graph is split through the connected graph dynamic split clustering algorithm to obtain the target cluster cluster; The determining of the original tolerance coefficient threshold corresponding to each current connected graph according to the number of connected edges and pixels in each current connected graph in the current clustering graph includes: Determine the ratio of the number of connected edges to the number of pixels in each current connected graph as a target ratio corresponding to each current connected graph, and determine the original tolerance coefficient threshold corresponding to each current connected graph according to the target ratio corresponding to each current connected graph; The formula corresponding to the original tolerance coefficient threshold corresponding to the current connectivity graph is: ;in, It is the first h The original tolerance threshold corresponding to the current connected graph; h is the sequence number of the current connected graph in the current clustering graph; It is an exponential function with a natural constant as base; It is the first h The number of connected edges in the current connected graph; It is the first h The number of pixels in the current connected graph; It is the first h The target ratio corresponding to the current connected graph.

2. The intelligent seismic data analysis system based on synthetic aperture radar technology according to claim 1 is characterized in that: The step of obtaining the current SAR image corresponding to the area to be detected, and the historical clustering graph and the historical clustering cluster set corresponding to the historical SAR image set includes: The SAR image of the area to be detected by earthquake at the current moment is obtained as the current SAR image, and the SAR image of the area to be detected by earthquake at each historical moment in the historical time period is obtained as the historical SAR image, so as to obtain a historical SAR image set; Determine the target distance characteristics between any two historical pixel points based on the phase, radar wave wavelength and radar reflection intensity corresponding to the historical pixel points; If the target distance feature between the historical pixels is less than or equal to the preset feature threshold, these historical pixels are connected to generate connection edges, and the graph obtained after all historical pixels are connected is used as the initial clustering graph, where the initial clustering graph includes all historical pixels; The initial clustering graph is split by the connected graph dynamic splitting clustering algorithm, and all the pixels included in each connected graph obtained by splitting the initial clustering graph constitute a historical clustering cluster to obtain a historical clustering cluster set, and all the connected graphs obtained by splitting the initial clustering graph constitute a historical clustering graph.

3. The intelligent seismic data analysis system based on synthetic aperture radar technology according to claim 2 is characterized in that: Determine the target distance characteristics between any two pixels, including: Randomly determine any two pixel points as the first reference point and the second reference point. If the first reference point and the second reference point are pixel points in the same frame of SAR image, determine the target distance feature between the first reference point and the second reference point according to the phase, radar wave wavelength and radar reflection intensity corresponding to the first reference point and the second reference point, wherein the radar wave wavelength corresponding to all pixel points in all SAR images corresponding to the area to be detected by earthquake is equal; If the first reference point and the second reference point are pixel points in different frames of SAR images, the SAR image where the first reference point is located is determined as the first SAR image, and the target distance characteristics between the first reference point and the second reference point are determined according to the phase corresponding to the pixel point in the first SAR image with the same position as the second reference point, the phase and radar wave wavelength corresponding to the first reference point, and the radar reflection intensity corresponding to the first reference point and the second reference point.

4. The intelligent analysis system for seismic data based on synthetic aperture radar technology according to claim 3 is characterized in that: The formulas corresponding to the target distance features between two pixels include: Jordi i Pixels and j pixels are pixels in the same frame of SAR image, then i Pixels and j The formula corresponding to the target distance feature between pixels is: ;in, It is i Pixels and j Target distance feature between pixels; i and j is the sequence number of the pixel points in all SAR images corresponding to the earthquake detection area, and ; It is i The value corresponding to the phase corresponding to each pixel point; It is j The value corresponding to the phase corresponding to each pixel point; is the arc value of 180°; It is the value corresponding to the wavelength of radar wave corresponding to the pixel point in the SAR image corresponding to the area to be detected by earthquake; It is i The value corresponding to the radar reflection intensity of each pixel point; It is j The value corresponding to the radar reflection intensity of each pixel point; Jordi i Pixels and j pixels are pixels in different frames of SAR images, then i Pixels and j The formula corresponding to the target distance feature between pixels is: ;in, It is i The SAR image where the pixel point is located is j The value corresponding to the phase corresponding to the pixel with the same pixel position.

5. The intelligent analysis system for seismic data based on synthetic aperture radar technology according to claim 1, characterized in that: The step of determining the current degree of association between each two historical clustering clusters in the historical clustering cluster set according to the connection between the two historical clustering clusters and the current pixel in the current clustering graph includes: According to the connection between the pixels in the current clustering map, determine the target confidence corresponding to each current pixel; Filter out the current pixel points that have connecting edges with the two historical clusters in the current cluster graph from all current pixel points as the intersection pixel points between the two historical clusters; The current correlation degree between the two historical clusters is determined according to the target confidence corresponding to all the intersection pixels between the two historical clusters and the connection between the two historical clusters and all the intersection pixels between them in the current clustering graph.

6. The intelligent analysis system for seismic data based on synthetic aperture radar technology according to claim 5 is characterized in that: Determining the target confidence corresponding to each current pixel point according to the connection between the pixels in the current clustering map includes: Determine any current pixel point as the current marked pixel point, and select from the historical cluster cluster set a historical cluster cluster that has a connection edge with the current marked pixel point in the current cluster graph as a reference cluster cluster of the current marked pixel point, thereby obtaining a reference cluster cluster set corresponding to the current marked pixel point; Filtering historical pixel points connected to the current marked pixel point in the current clustering diagram from the reference cluster set corresponding to the current marked pixel point as candidate points of the current marked pixel point; The target confidence corresponding to the currently marked pixel is determined according to the target distance characteristics between all candidate points of the currently marked pixel and the currently marked pixel in the reference cluster set corresponding to the currently marked pixel.

7. The intelligent seismic data analysis system based on synthetic aperture radar technology according to claim 6 is characterized in that: The target confidence corresponding to the current pixel point and the formulas corresponding to the current degree of association between the two historical clusters are: ； ;in, It is a The target confidence corresponding to the current pixel; a is the serial number of the current pixel; It is a The number of reference clusters in the reference cluster set corresponding to the current pixel point; r It is a The serial number of the reference cluster in the reference cluster set corresponding to the current pixel point; It is a The first one in the reference cluster set corresponding to the current pixel r The reference clusters and a The number of connected edges between the current pixels; It is an exponential function with a natural constant as base; It is a The first one in the reference cluster set corresponding to the current pixel r Among the reference clusters, a All candidate points of the current pixel and the a The mean of the target distance features between the current pixels; It is a In the reference cluster set corresponding to the current pixel point, a All candidate points of the current pixel and the a The mean of the target distance features between the current pixels; It is p The historical clusters and q The current degree of association between historical clusters; p and q is the sequence number of the historical cluster in the set of historical clusters, and ; It is p The historical clusters and q The number of intersection pixels between historical clusters; t It is p The historical clusters and q The sequence number of the intersection pixel between the historical clusters; is the normalization function; It is p The historical clusters and q The first t The target confidence corresponding to the intersection pixel; It is the absolute value function; It is p The historical clusters and q The first t The intersection pixel, and p The historical clusters t The accumulated value of the target distance feature between all candidate points of the intersection pixels; It is p The historical clusters and q The first t The intersection pixel, and q The first t The accumulated value of the target distance feature between all candidate points of the intersection pixels; It is p The historical clusters and q All the intersection pixels between the historical clusters and the q The number of connecting edges between historical clusters; It is p The historical clusters and q All the intersection pixels between the historical clusters and the p The number of connecting edges between historical clusters.

8. The intelligent seismic data analysis system based on synthetic aperture radar technology according to claim 1 is characterized in that: The method of correcting the original tolerance coefficient threshold corresponding to each current connectivity graph according to the current degree of association between all historical clusters in each current connectivity graph to obtain the target tolerance coefficient threshold corresponding to each current connectivity graph includes: According to the current correlation degree between all historical clusters in each current connected graph, the target weight corresponding to each current connected graph is determined; The product of the target weight corresponding to each current connected graph and the original tolerance coefficient threshold is determined as the target tolerance coefficient threshold corresponding to each current connected graph.

9. The intelligent seismic data analysis system based on synthetic aperture radar technology according to claim 8, characterized in that: The formula corresponding to the target tolerance threshold of the current connectivity graph is: ; ;in, It is the first h The target tolerance coefficient threshold corresponding to the current connected graph; h is the sequence number of the current connected graph in the current clustering graph; It is the first h The target weight corresponding to the current connected graph; It is the first h The original tolerance threshold corresponding to the current connected graph; is the normalization function; It is the first h The cumulative value of the current degree of association between all historical clusters in the current connectivity graph.

Images