Method for constructing fracture evolution trend pattern of floor failure zone based on image recognition

Abstract

The application discloses a floor failure zone crack evolution trend graph construction method based on image recognition and relates to the field of crack evolution monitoring, and comprises the following steps: obtaining floor failure zone images in different monitoring stages and performing imaging specification processing and stage marking; performing crack target recognition and texture anomaly recognition on the floor failure zone stage images; performing relationship determination on crack spatial interaction states in crack image element data; identifying crack intersection sheltering directions, and inversely layering crack formation stages in combination with prior constraint conflict resolution; screening active states of late-stage through layers; constructing crack end sector search areas, and continuously scanning and envelope checking weakening features in the sectors; configuring graph nodes and arranging relationship edges, and generating a floor failure zone crack evolution trend graph. The application utilizes multi-stage image recognition and crack relationship determination to construct a crack evolution trend graph, and has the advantages of clear layering and accurate extension determination.

Description

Technical Field

[0001] This invention relates to the field of fracture evolution monitoring, and in particular to a method for constructing a fracture evolution trend map of a base plate failure zone based on image recognition. Background Technology

[0002] Current methods for monitoring fractures in the foundation slab typically rely on manual interpretation, single-stage image segmentation, or fracture skeleton extraction to identify fracture areas formed under mining influence. These methods then combine image features such as fracture length, width, intersection location, grayscale variations, and fracture texture distribution to analyze the extent of foundation slab damage and the degree of fracture development. While these methods can reflect the current morphology of fractures to some extent, they primarily focus on single images or single monitoring stages.

[0003] In continuous monitoring scenarios, base plate cracks exhibit complex interactive states such as cutting, covering, dislocation, overlapping, and boundary rewriting. Existing methods struggle to resolve the crack formation sequence based on the shading direction of crack intersections and the conflict of constraints. They also struggle to uniformly correlate early crack layers, mid-term modified layers, late-term penetrating layers, currently active crack layers, and areas to be extended. This results in unclear division of crack relative age layers, inaccurate identification of active cracks, and insufficient expression of potential extension directions, making it impossible to generate a map-based result that reflects the evolution trend of cracks in the base plate damage zone.

[0004] Therefore, how to provide a method for constructing an image recognition-based map of the evolution trend of cracks in the base plate is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] One objective of this invention is to propose a method for constructing a fracture evolution trend map of a base plate failure zone based on image recognition. This invention utilizes multi-stage image recognition and fracture relationship determination to construct a fracture evolution trend map, which has the advantages of clear hierarchy and accurate extension judgment.

[0006] The method for constructing a fracture evolution trend map of a base plate failure zone based on image recognition according to an embodiment of the present invention includes the following steps: Images of the base plate failure zone at different monitoring stages were acquired, and the images were processed according to imaging standards and marked with stages to generate stage images of the base plate failure zone. Crack target identification and texture anomaly identification are performed on stage images of the base plate failure zone to generate crack image element data; Determine the relationship between the spatial interaction state of fractures in fracture image feature data, and generate fracture superposition rewriting relationship; The direction of occlusion at fracture confluence is identified based on the fracture superposition rewriting relationship, and the fracture formation stage is hierarchically inverted by combining the sequential constraint conflict resolution to generate fracture relative age layer data. Active state screening is performed on late-stage interconnected layers in relative age-stratigraphic data of fractures to generate current active fracture layer identification data; Based on the current active fracture layer identification data, a fracture end sector search region is constructed, and the weakened features within the sector are continuously scanned and envelope verified to generate potential fracture extension links and potential extension relationships. Based on relative age layer data of fractures, identification data of currently active fracture layers, potential extension links of fractures, and potential extension relationships, a map of fracture evolution trend in the basement failure zone is generated by configuring map nodes and arranging relationship edges.

[0007] Optionally, the imaging standardization processing includes image denoising, grayscale equalization, scale unification, and distortion correction.

[0008] Optionally, the generation of the crack image feature data specifically includes: The continuous dark areas and linear texture areas in the stage images of the base plate failure zone are jointly extracted, and suspected crack areas and suspected fracture texture areas are identified by combining the monitoring stage markers. Connectivity domain arrangement, extended morphology screening and fracture structure element extraction are performed on suspected fracture regions to generate fracture target identification data; The texture extension direction, local gray-level abrupt state, and local boundary integrity of suspected broken texture areas are identified to generate texture anomaly identification data. The crack target identification data and texture anomaly identification data are collected according to the monitoring stage and spatial location to generate crack image element data.

[0009] Optionally, the generation of the fracture superposition rewriting relationship specifically includes: Based on the crack image element data, cracks that have spatial interaction within the same monitoring stage are combined and screened to obtain candidate interactive crack pairs and their local interactive regions. The continuous state of the skeleton is traced along the direction of the fracture centerline in the candidate interactive fracture pair, and when the continuous state of the skeleton and the continuity feature across another fracture boundary exist at the same time, the corresponding fracture spatial interaction state is determined as the cutting relationship. Within the local interaction region of the candidate interactive crack pair, the position of the texture anomaly region coincides with the original crack boundary, and the crack spatial interaction state where the coincident position changes discontinuously with the original crack boundary is determined as the coverage relationship. Within a local interaction region, the location of the misalignment of the fracture centerline and the path of another fracture centerline are traced. When the misalignment location is crossed by the path, the corresponding fracture spatial interaction state is determined as a fault relationship. The distribution of texture anomaly regions in different fracture target regions is compared within the local interaction region. When the texture anomaly regions overlap each other without forming a covering, discontinuity or cutting, the corresponding fracture spatial interaction state is determined as an overlay relationship. Within the local interaction area, the original crack boundary disappearance position is compared with the newly formed broken texture occupation position and the new crack boundary occupation position. When any occupation position coincides with the boundary disappearance position, the corresponding crack space interaction state is determined as the boundary rewriting relationship. The cutting relationship, covering relationship, discontinuing relationship, overlay relationship, and boundary rewriting relationship are aggregated according to the monitoring stage, candidate interactive fracture pairs, and relationship type to generate fracture overlay rewriting relationship.

[0010] Optionally, the generation of the fracture relative age layer data specifically includes: Based on the crack overlay rewriting relationship, the continuous state of the skeleton, the discontinuous position of the boundary and the coverage direction of the broken texture at the crack junction are extracted to generate junction occlusion recognition data; Based on the intersection occlusion recognition data, determine the first and second cracks in the candidate interactive crack pairs, and generate crack sequence labeling data; Based on the fracture overlay rewriting relationship and fracture sequence marking data, the formation order direction between candidate interactive fracture pairs is determined, and the formation order direction is used as the fracture sequence constraint. For the same candidate interactive fracture pair, the direction of the fracture constraints is compared and the type is verified, and the constraint state with opposite constraint directions or multiple relationship types is determined as a constraint conflict. In candidate interactive gap pairs with conflicting sequential constraints, the sequential constraints of the gaps are selected and discarded in the order of boundary rewriting relationship, covering relationship, discontinuation relationship, cutting relationship and overlay relationship to obtain the effective sequential constraints. Based on the effective priority constraint, the distribution of the number of times of the target fracture area is statistically analyzed, and the fracture is hierarchically classified according to the number of times it occurred first, the state of modification and acceptance, and the performance of the penetration, to generate fracture hierarchical classification data. The fracture hierarchical classification data includes early fracture layer, mid-term modification layer and late penetration layer. Based on the fracture hierarchy data and effective sequential constraints, the interlayer progression relationship between different fracture layers is determined, and the fracture hierarchy data and interlayer progression relationship are aggregated to generate fracture relative age layer data.

[0011] Optionally, the generation of the currently active fracture layer identification data specifically includes: The end states of late-stage continuous fractures in relative age-determined fracture data are arranged to determine the end aperture, end boundary, and end orientation of late-stage continuous fractures, thereby generating end state data of late-stage continuous fractures. Based on the end state data of late-stage through-cracks, the changes in end opening and the clear state of end boundaries are identified, and weak texture areas, local fragmentation areas and gray-scale abrupt change areas in the neighborhood of the end are identified along the end orientation of the late-stage through-cracks. Based on the increased end opening, clear end boundary, and the extension and change of weak texture area, local broken area and gray-scale abrupt change area, the active state of the fractures in the late-stage penetrating layer is screened to determine the current active fracture layer. The fractures in the current active fracture layer are associated with the fractures in the later-stage through-layer to obtain the active inheritance relationship. The current active fracture layer, active end state, end neighborhood weakening state and active inheritance relationship are collected to generate the current active fracture layer identification data.

[0012] Optionally, the generation of the potential extension links and potential extension relationships of the crack specifically includes: Based on the current active fracture layer identification data, the fracture end to be extended for search is determined, and the fracture end sector search area is constructed with the fracture end as the starting point. A continuous scanning band is set along the end direction within the search area of ​​the sector at the end of the crack, and the location of the weak texture area, local broken area and gray-scale change area within the continuous scanning band is identified to generate sector weakening feature distribution data. Weak texture areas, locally fragmented areas, or gray-scale abrupt change areas within adjacent continuous scan bands are connected sequentially according to their end orientation to form a weakening path, and candidate extension paths are determined based on the continuous penetration state of the weakening path. Based on the end position and end orientation of the candidate extension path, the region in front of the end to which the candidate extension path points is determined, and the region in front of the end is taken as the region to be extended. An end-closed envelope region is constructed around the candidate extension path, and envelope verification is performed on the continuity of weak texture regions, the arrangement direction of local broken regions, and the continuity state of gray-scale abrupt change regions within the end-closed envelope region. Candidate extension paths are selected based on the envelope verification results. Candidate extension paths that satisfy envelope continuity are identified as potential extension links for the fracture. Potential extension relationships are generated based on the link pointing relationship between the current active fracture layer and the region to be extended.

[0013] Optionally, the generation of the fracture evolution trend map of the base plate failure zone specifically includes: The relative age layer data of fractures, the identification data of currently active fracture layers, the potential extension links of fractures and potential extension relationships are sorted out to generate fracture evolution map composition data. Based on the fracture evolution map composition data, map nodes are configured, with the early fracture layer, the intermediate modified layer, the late through layer, the currently active fracture layer, and the region to be extended respectively configured as map nodes; Connect the graph nodes according to the interlayer progression relationship, active succession relationship and potential extension relationship to obtain the graph relationship edge; The graph nodes and graph relationship edges are collected according to the order of fracture evolution to generate a graph of the fracture evolution trend of the base plate failure zone.

[0014] The beneficial effects of this invention are: This invention performs image standardization processing, stage marking, fracture target identification, and texture anomaly identification on base plate failure zone images at different monitoring stages. This enables correlation analysis of fracture target areas, fracture centerlines, fracture boundaries, fracture ends, fracture junctions, and texture anomaly areas under a unified scale and stage marking. Compared to methods relying solely on single-stage grayscale segmentation or manual interpretation, this invention simultaneously utilizes fracture morphology features and texture anomaly features, reducing confusion between blocky dark areas, locally fragmented textures, and actual fractures. This improves the completeness and consistency of fracture image element extraction from base plate failure zones, providing a more stable data foundation for subsequent fracture interaction relationship determination.

[0015] This invention further determines the cutting, covering, discontinuing, overlapping, and boundary rewriting relationships of the fracture space interaction state, and combines the continuous state of the skeleton at the fracture confluence, the discontinuous position of the boundary, and the direction of the fracture texture coverage to invert the fracture formation stage. By resolving conflicts arising from sequential constraints, the problem of confusion in determining the formation sequence caused by the coexistence of multiple relationships in the same fracture confluence area can be avoided, making the division of the early fracture layer, the intermediate modified layer, and the late through-layer clearer. Therefore, this invention can not only identify the current morphology of the fracture, but also express the overlapping rewriting process and relative formation sequence between fractures, improving the continuity and traceability of the fracture evolution process analysis in the basement failure zone.

[0016] This invention also screens currently active fracture layers based on late-stage penetrating layers, and generates potential fracture extension links and relationships through fracture end sector search, continuous scanning of weakened features, and envelope verification. This unifies the representation of existing fracture levels, currently active fractures, and regions that may continue to extend into the map. The resulting floor failure zone fracture evolution trend map simultaneously presents interlayer progression relationships, active succession relationships, and potential extension relationships, making the development direction, evolution stage, and subsequent expansion trend of floor failure zone fractures more intuitive. This improves the accuracy of floor stability analysis, failure zone development assessment, and subsequent monitoring decisions. Attached Figure Description

[0017] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1This is a flowchart of the method for constructing a fracture evolution trend map of the base plate based on image recognition, as proposed in this invention. Figure 2 This is a flowchart illustrating the generation of relative age layer data for the image recognition-based method for constructing a fracture evolution trend map of the base plate failure zone proposed in this invention. Detailed Implementation

[0018] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0019] refer to Figures 1-2 A method for constructing a fracture evolution trend map of the base plate based on image recognition includes the following steps: Images of the base plate failure zone at different monitoring stages were acquired, and the images were processed according to imaging standards and marked with stages to generate stage images of the base plate failure zone. Crack target identification and texture anomaly identification are performed on stage images of the base plate failure zone to generate crack image element data; Determine the relationship between the spatial interaction state of fractures in fracture image feature data, and generate fracture superposition rewriting relationship; The direction of occlusion at fracture confluence is identified based on the fracture superposition rewriting relationship, and the fracture formation stage is hierarchically inverted by combining the sequential constraint conflict resolution to generate fracture relative age layer data. Active state screening is performed on late-stage interconnected layers in relative age-stratigraphic data of fractures to generate current active fracture layer identification data; Based on the current active fracture layer identification data, a fracture end sector search region is constructed, and the weakened features within the sector are continuously scanned and envelope verified to generate potential fracture extension links and potential extension relationships. Based on relative age layer data of fractures, identification data of currently active fracture layers, potential extension links of fractures, and potential extension relationships, a map of fracture evolution trend in the basement failure zone is generated by configuring map nodes and arranging relationship edges.

[0020] In this embodiment, the imaging standardization processing includes image denoising, grayscale equalization, scale unification, and distortion correction. Specifically, the image of the base plate damage zone is first replaced with isolated bright spots and dark spots using a 3×3 window, and then smoothed with a 5×5 window to remove fine noise outside the suspected linear dark area of ​​the fracture. After denoising, the image is divided into 8×8 grayscale processing blocks, and the grayscale enhancement limit value is set to 2.0 to create a stable grayscale difference between the fracture boundary, fracture texture, and adjacent rock texture. Subsequently, the image is uniformly configured to 1024×1024 pixels to form a unified scale benchmark for subsequent recording of fracture width, fracture end opening, and texture anomaly area. When the straight line offset of the image edge exceeds 6 pixels, the image is distorted, and the corrected effective recognition area is determined in the unified scale image. The effective recognition area is configured to 960×960 pixels. Stage markers are used to record the acquisition time and monitoring stage corresponding to the base plate damage zone image.

[0021] In this embodiment, the generation of crack image feature data specifically includes: The continuous dark areas and linear texture areas in the stage images of the base plate failure zone are jointly extracted, and suspected crack areas and suspected fracture texture areas are identified by combining the monitoring stage markers. The joint extraction includes identifying continuous dark areas with gray values ​​lower than those of the neighboring rock mass texture in the stage image of the base plate failure zone, as well as linear texture areas extending in the same direction. Continuous dark areas are identified according to a 15×15 neighborhood window. Areas with a difference between the gray value and the mean gray value of the neighborhood window greater than 20 and with no less than 12 consecutive pixels are marked as continuous dark areas. Linear texture areas are identified according to the main extension direction. Areas with a deviation of no more than 15 degrees in the main extension direction, a single extension length of no less than 18 pixels, and a breakpoint spacing of no more than 3 pixels are marked as linear texture areas. Areas where continuous dark areas and linear texture areas overlap and the extension length is more than 3 times the area width are marked as suspected fracture areas. Local areas with no less than 3 gray value change directions, no less than 5 boundary turning points, a main extension length less than 3 times the area width, and no continuous linear extension are marked as suspected fracture texture areas. Connectivity domain arrangement, extended morphology screening and fracture structure element extraction are performed on suspected fracture regions to generate fracture target identification data; The connected component arrangement includes using an eight-neighbor connected component labeling method to connect suspected gap regions, retaining connected regions with no less than 30 pixels and a long side of the circumscribed rectangle of no less than 20 pixels; The extended shape screening includes marking connected regions with a length greater than 3 times the width and continuous extension direction as crack target regions, removing dark areas that are distributed in a point-like or block-like isolated distribution, and ensuring that the extension direction is continuous, with the angle between the directions of adjacent center line segments not exceeding 20 degrees and the distance between the center line breakpoints not exceeding 4 pixels. The structural elements of a crack include the crack centerline, crack boundary, crack end, and crack junction. The crack centerline is obtained by refining the crack target area using a single-pixel skeleton. The crack end is obtained from the position where the number of skeleton pixels adjacent to each other is 1. The crack junction is obtained from the position where the number of skeleton pixels adjacent to each other is not less than 3. The crack boundary is obtained from the outer contour of the crack target area. The texture extension direction, local gray-level abrupt state, and local boundary integrity of suspected broken texture areas are identified to generate texture anomaly identification data. Texture extension direction identification includes determining the extension direction of the suspected broken texture region relative to the crack end, crack boundary and crack junction. The longest external direction of the suspected broken texture region is taken as the initial texture direction. When the angle between the initial texture direction and the direction of the adjacent crack end does not exceed 30 degrees, the initial texture direction is marked as the texture extension direction. The identification of local gray-scale abrupt changes includes comparing the gray-scale differences between the suspected fractured texture area and the texture of the adjacent rock mass. The suspected fractured texture area is divided into 5×5 gray-scale statistical units. When the gray-scale mean difference between adjacent gray-scale statistical units is greater than 25 and occurs continuously at least 3 times, and forms discontinuous changes, block changes, or band changes, it is marked as a local gray-scale abrupt change. Discontinuous changes are defined as the interval between gray-scale abrupt change units not exceeding 2 statistical units. Block changes are defined as the number of gray-scale abrupt change units concentrated in the local area not less than 6. Band changes are defined as gray-scale abrupt change units arranged continuously in the same direction for a length of not less than 4 statistical units. Local boundary integrity identification includes identifying the continuous and discontinuous states of the edges of suspected broken texture regions, and marking suspected broken texture regions with discontinuous boundaries that are adjacent to or overlap with the crack target region as texture anomalous regions. Discontinuous boundaries are defined as continuous boundary segments with a length less than 60% of the total edge length, adjacent boundaries are defined as the shortest distance between the suspected broken texture region and the crack target region not exceeding 5 pixels, and overlapping boundaries are defined as the overlapping area of ​​the two pixels being greater than 10% of the area of ​​the suspected broken texture region. The crack target identification data and texture anomaly identification data are collected according to the monitoring stage and spatial location to generate crack image element data. The crack image element data includes crack target area, crack center line, crack boundary, crack end, crack intersection, texture anomaly area, texture extension direction, local gray-scale change state and monitoring stage.

[0022] In this embodiment, the generation of the fracture superposition rewriting relationship specifically includes: Based on the crack image element data, cracks that have spatial interaction within the same monitoring stage are combined and screened to obtain candidate interactive crack pairs and their local interactive regions. The combined screening involves retrieving the target region of the crack, the center line of the crack, the crack boundary, the crack end, and the texture anomaly region from the crack image feature data. Two target regions of cracks are identified as candidate interactive crack pairs if the crack center lines intersect, the crack boundaries are adjacent, the shortest distance between the two crack ends does not exceed 8 pixels and the angle between the ends does not exceed 30 degrees, or the texture anomaly regions overlap. Among these, adjacent crack boundaries are defined as the shortest distance between the two crack boundaries not exceeding 5 pixels, and overlapping texture anomaly regions are defined as the overlapping area of ​​the two texture anomaly regions accounting for no less than 15% of the area of ​​their respective texture anomaly regions. The local interaction region is the local image range in the candidate interaction crack pair where the center line intersects, the boundary is adjacent, the end is close, or the texture abnormal region overlaps. It is formed by extending outward by 12 pixels based on the interaction position within the local image range. When there are multiple local image ranges in the same candidate interaction crack pair and the distance between adjacent local image ranges does not exceed 10 pixels, the adjacent local image ranges are merged into the same local interaction region. The continuous state of the skeleton is traced along the direction of the fracture centerline in the candidate interactive fracture pair, and when the continuous state of the skeleton and the continuity feature across another fracture boundary exist at the same time, the corresponding fracture spatial interaction state is determined as the cutting relationship. The continuous state of the skeleton is characterized by the extension direction of one crack center line crossing the boundary of another crack not exceeding 18 degrees, and a continuous skeleton segment with a length of not less than 12 pixels is retained on both sides of the boundary of another crack, and the angle between the line connecting the endpoints of the skeleton segments on both sides and the main extension direction of the crack center line does not exceed 20 degrees. The continuity feature is characterized by the breakpoint spacing between the skeleton segment before and after the crossover not exceeding 4 pixels, the difference in skeleton width before and after the crossover not exceeding 3 pixels, and the continuous extension length after the crossover not less than 10 pixels. Within the local interaction region of the candidate interactive crack pair, the position of the texture anomaly region coincides with the original crack boundary, and the crack spatial interaction state where the coincident position changes discontinuously with the original crack boundary is determined as the coverage relationship. The overlapping state includes the texture abnormal region covering the original crack boundary, or continuously spreading along the outside of the crack boundary or the direction of crack intersection. Among them, when the overlapping area of ​​the texture abnormal region and the original crack boundary is not less than 25% of the area of ​​the original crack boundary neighborhood, it is determined to be the covering state; when the length of the texture abnormal region continuously spreading along the outside of the crack boundary or the direction of crack intersection is not less than 16 pixels, it is determined to be the continuous spreading state. The discontinuous change of the original fracture boundary includes the original fracture boundary changing from a continuous outer contour to a discontinuous outer contour within the texture anomaly region, and the discontinuity position coincides with the coverage area of ​​the texture anomaly region. Specifically, when the traceable boundary length of the original fracture boundary within the texture anomaly region is less than 60% of the boundary length before entering the texture anomaly region, and the number of discontinuous segments is not less than 3 segments, it is determined to be a change from a continuous outer contour to a discontinuous outer contour; when the overlap ratio between the discontinuous position and the coverage area of ​​the texture anomaly region is not less than 40%, the discontinuous change is retained. Within a local interaction region, the location of the misalignment of the fracture centerline and the path of another fracture centerline are traced. When the misalignment location is crossed by the path, the corresponding fracture spatial interaction state is determined as a fault relationship. The position of the fracture centerline shift includes the position where the fracture centerline of the same fracture target area is offset on both sides of the local interaction area. The position offset is such that the lateral distance between the endpoints of the centerlines on both sides of the local interaction area of ​​the same fracture target area is not less than 5 pixels, and the angle between the main extension directions of the centerlines on both sides does not exceed 25 degrees. The other crack centerline penetration path includes a crack centerline that passes through the offset position of the crack centerline and maintains continuous extension. The shortest distance between the other crack centerline and the offset position is no more than 3 pixels. The continuous extension means that the length of the continuous skeleton segment before and after the offset position of the other crack centerline is no less than 10 pixels, and the angle between the directions of the skeleton segments before and after the offset position is no more than 20 degrees. The distribution of texture anomaly regions in different fracture target regions is compared within the local interaction region. When the texture anomaly regions overlap each other without forming a covering, discontinuity or cutting, the corresponding fracture spatial interaction state is determined as an overlay relationship. Texture anomaly regions overlap to form different crack target regions. Texture anomaly regions are located in the same local interaction region and form spatial overlap or staggered distribution. Among them, spatial overlap means that the overlap area of ​​two texture anomaly regions accounts for no less than 20% of the area of ​​their respective texture anomaly regions. Staggered distribution means that the distance between the outer contours of two texture anomaly regions does not exceed 4 pixels, and the angle between their texture extension directions is greater than 35 degrees. The non-coverage includes the overlap ratio between the texture abnormal area and the original crack boundary neighborhood being less than 25%, and the continuous length along the outside of the crack boundary or the crack intersection direction being less than 16 pixels, or the traceable boundary length of the original crack boundary in the texture abnormal area being not less than 60% of the boundary length before entering the texture abnormal area, or the number of discontinuous segments being less than 3 segments. The absence of a fault includes a lateral distance of less than 5 pixels between the center lines on both sides of the local interaction area, or an angle greater than 25 degrees between the main extension directions of the center lines on both sides, or the absence of another center line of the fault that passes through the fault position and continues to extend continuously. Failure to form a cut includes: the length of the continuous skeleton segment after crossing the boundary is less than 10 pixels, or the angle between the directions of the skeleton segments before and after crossing the boundary is greater than 20 degrees, or the distance between the breakpoints before and after crossing the boundary is more than 4 pixels, or the difference in skeleton width before and after crossing the boundary is more than 3 pixels. Within the local interaction area, the original crack boundary disappearance position is compared with the newly formed broken texture occupation position and the new crack boundary occupation position. When any occupation position coincides with the boundary disappearance position, the corresponding crack space interaction state is determined as the boundary rewriting relationship. The newly formed broken texture is a texture abnormal area in which the original crack boundary disappears within a local interactive region, and it is characterized by gray-scale changes, irregular boundaries, and no continuous linear extension. Overlap comparison includes identifying the location where the original crack boundary changes from a continuous outer contour to a missing boundary in the local interaction area, and identifying the occupancy status of newly formed broken textures or new crack boundaries at the location where the boundary disappears. The occupied state includes a new fracture texture and a boundary disappearance position with an overlap area ratio of not less than 35%, or a new crack boundary and a boundary disappearance position with an directional angle of not more than 25 degrees and a minimum distance of not more than 4 pixels. The boundary disappearance location is the location where the length of the traceable boundary of the original crack boundary within the local interaction area is less than 50% of the length of the boundary before entering the local interaction area, and the length of the missing boundary segment is not less than 8 pixels. The cutting relationship, covering relationship, discontinuing relationship, overlay relationship, and boundary rewriting relationship are aggregated according to the monitoring stage, candidate interactive fracture pairs, and relationship type to generate fracture overlay rewriting relationship.

[0023] In this embodiment, the generation of fracture relative age layer data specifically includes: Based on the crack overlay rewriting relationship, the continuous state of the skeleton, the discontinuous position of the boundary and the coverage direction of the broken texture at the crack junction are extracted to generate junction occlusion recognition data; The continuous state of the skeleton includes the degree of continuity and the change in the extension direction of the fracture centerline before and after passing through the fracture junction. The fracture junction is centered on the skeleton points with no less than 3 adjacent fracture centerlines, and the junction analysis window is formed by extending outward from the skeleton points by 15 pixels. When the length of the centerline segments on both the front and back sides is no less than 10 pixels, the directional angle does not exceed 20 degrees, and the distance between the breakpoints does not exceed 4 pixels, the skeleton is determined to be continuous. The location of a discontinuous boundary includes the location where the fracture boundary is interrupted, missing, or covered within the fracture confluence. When the length of the traceable boundary within the confluence analysis window is less than 60% of the length of the boundary before entering the confluence, and the number of discontinuous segments is not less than 2, it is determined as a discontinuous boundary location. The broken texture coverage direction includes the direction in which the texture abnormal region spreads from one crack boundary to another crack boundary. When the overlap ratio between the texture abnormal region and the crack boundary is not less than 25%, and the continuous spreading length along one crack boundary to another crack boundary is not less than 12 pixels, the spreading direction is determined as the broken texture coverage direction. Based on the intersection occlusion recognition data, determine the first and second cracks in the candidate interactive crack pairs, and generate crack sequence labeling data; The back crack is a crack that continuously crosses the crack junction and maintains an extended boundary. The determination of the back crack includes the crack centerline continuously crossing the crack junction, the included angle between the directions of the centerline segments before and after the junction not exceeding 20 degrees, and the retention of continuous skeleton segments with a length of not less than 10 pixels on both sides of the junction. The prior crack is defined as a crack whose boundary is interrupted or covered by broken texture at the crack confluence. The determination of the prior crack includes the crack boundary being interrupted, missing, or covered by broken texture within the confluence analysis window. Interruption or missing means that the length of the traceable boundary is less than 60% of the length of the boundary before entering the confluence. Covered by broken texture means that the overlap ratio between the texture anomaly area and the neighborhood of the crack boundary is not less than 25%. Based on the fracture overlay rewriting relationship and fracture sequence marking data, the formation order direction between candidate interactive fracture pairs is determined, and the formation order direction is used as the fracture sequence constraint. The determination of the formation order direction includes taking the earlier fracture in the fracture sequence marking data as the starting end and the later fracture as the pointing end to form the formation order direction from the earlier fracture to the later fracture; when the same candidate interactive fracture pair has not formed a complete sequence marking, the formation order direction is determined according to the relationship type in the fracture superposition rewriting relationship and the occlusion, penetration, displacement or boundary replacement state in the local interactive area. For the same candidate interactive fracture pair, the direction of the fracture constraints is compared and the type is verified, and the constraint state with opposite constraint directions or multiple relationship types is determined as a constraint conflict. Orientation comparison includes comparing the constraint directions formed by the same candidate interaction gap pair in different local interaction regions. When the center distance between different local interaction regions is greater than 10 pixels, they are treated as independent local interaction regions and participate in orientation comparison. When one constraint direction indicates that the first gap is earlier than the second gap, and another constraint direction indicates that the second gap is earlier than the first gap, they are determined to be opposite constraint directions. Type verification includes verifying the sequential constraints of fractures formed by different relation types in the same candidate interactive fracture pair. When the same candidate interactive fracture pair has sequential constraints formed by no less than two relation types, it is determined that multiple relation types coexist. In candidate interactive gap pairs with conflicting sequential constraints, the sequential constraints of the gaps are selected and discarded in the order of boundary rewriting relationship, covering relationship, discontinuation relationship, cutting relationship and overlay relationship to obtain the effective sequential constraints. The selection process includes prioritizing the retention of fracture constraints formed by boundary rewriting relationships when there are conflicting sequential constraints in the same candidate interactive fracture pair; when there are no boundary rewriting relationships, sequential constraints of fractures formed by covering relationships, discontinuing relationships, cutting relationships, or overlapping relationships are retained in sequence, and fracture constraints that are opposite to the direction of the retained valid sequential constraints are eliminated. Based on the effective priority constraint, the distribution of the number of times of the target fracture area is statistically analyzed, and the fracture is hierarchically classified according to the number of times it occurred, the status of the modification and acceptance, and the performance of the penetration. The fracture hierarchical classification data includes the early fracture layer, the mid-term modification layer, and the late penetration layer. Hierarchical classification includes counting the number of times each fracture target region appears before and after the effective priority constraints; fractures with more appearances before the priority constraints than after the priority constraints and without accompanying overlay, discontinuity, or boundary rewriting relationships are classified into the early fracture layer, where more appearances before the priority constraints are at least two times more than after the priority constraints; fractures with both appearances before and after the priority constraints and accompanied by overlay, discontinuity, or boundary rewriting relationships are classified into the intermediate modification layer, where both appearances before and after the priority constraints are at least one time each. Fracts with more subsequent occurrences than earlier occurrences and continuous penetration characteristics are classified into the late penetration layer. "More subsequent occurrences than earlier occurrences" means that the number of subsequent occurrences is at least two more than the number of earlier occurrences. Continuous penetration characteristics include the fracture centerline continuously crossing the fracture junction, and the preservation of continuous skeleton segments on both sides of the local interaction area, with each continuous skeleton segment on both sides of the local interaction area having a length of at least 12 pixels. When the same fracture target area simultaneously meets two level conditions, the final classification level is determined according to the order of late penetration layer, intermediate modification layer, and early fracture layer. The interlayer progression relationship between different fracture layers is determined based on fracture hierarchy data and effective sequential constraints, and the fracture hierarchy data and interlayer progression relationship are aggregated to generate fracture relative age layer data. The determination of interlayer progression relationships includes: when the earlier fracture in the effective sequential constraints belongs to the first fracture layer and the later fracture belongs to the second fracture layer, and the first fracture layer and the second fracture layer do not belong to the same fracture layer, the sequential bearing direction from the first fracture layer to the second fracture layer is determined as the interlayer progression relationship; when multiple effective sequential constraints form the same fracture layer pointing relationship, the same fracture layer pointing relationship is merged into the same interlayer progression relationship.

[0024] In this embodiment, the generation of currently active fracture layer identification data specifically includes: The end states of late-stage continuous fractures in relative age-determined fracture data are arranged to determine the end aperture, end boundary, and end orientation of late-stage continuous fractures, thereby generating end state data of late-stage continuous fractures. End state arrangement includes calling the fracture target region, fracture centerline, fracture end, and fracture boundary in the late-stage through layer. First, the extension direction of the continuous skeleton segment at the end of the fracture centerline is determined as the end orientation. Then, the fracture end is extended forward by 20 pixels along the end orientation and 10 pixels on each side of the end to form the fracture end neighborhood. The shortest distance between the two fracture boundaries in the fracture end neighborhood is determined as the end opening. The continuous state of the fracture boundary in the fracture end neighborhood is determined as the end boundary. Based on the end state data of late-stage through-cracks, the changes in end opening and the clear state of end boundaries are identified, and weak texture areas, local fragmentation areas and gray-scale abrupt change areas in the neighborhood of the end are identified along the end orientation of the late-stage through-cracks. End aperture change includes comparing the end aperture of the current monitoring stage with that of the previous monitoring stage. When the end aperture of the current monitoring stage is greater than that of the previous monitoring stage and the increase is not less than 2 pixels, it is determined to be an increase in end aperture. The clear end boundary status includes identifying the traceable length and the number of discontinuous segments of the fracture boundary in the vicinity of the fracture end. When the length of the traceable boundary accounts for no less than 70% of the total length of the end boundary and the number of discontinuous segments does not exceed 2, the end boundary is determined to be clear. The weak texture area is a region in the neighborhood of the fracture end that has a lower grayscale variation than the grayscale variation of the adjacent rock mass texture and has a continuously weakened boundary. The grayscale variation of the weak texture area is calculated according to 5×5 pixel statistical units. When the grayscale range of the consecutive statistical units in the neighborhood of the end is less than 60% of the grayscale range of the adjacent rock mass texture, and the number of consecutive statistical units is not less than 4, the area where the consecutive statistical units are located is determined as the weak texture area. The adjacent rock mass texture is selected from the annular neighborhood area outside the fracture end neighborhood that is 5 to 15 pixels away from the fracture boundary, and pixels that overlap with the fracture target area, the local fracture area, and the grayscale change area are removed. The local broken zone is an area with irregular boundaries in the neighborhood of the crack end, no less than 3 gray-scale change directions and no less than 5 local turning points. The number of local turning points is counted according to the outer contour of the local broken zone. When the angle between the directions of adjacent contour segments is not less than 35 degrees and the distance between adjacent turning points is not less than 2 pixels, it is counted as 1 local turning point. Continuous turning points with a distance of less than 2 pixels are merged into 1 local turning point. The gray-scale abrupt change area is the area in the neighborhood of the crack end where the gray-scale mean difference between adjacent gray-scale statistical units is greater than 25 and the number of consecutive occurrences is no less than 3. The gray-scale statistical unit is configured as 5×5 pixels. Adjacent gray-scale statistical units are arranged in sequence according to the direction of the crack end. When adjacent gray-scale statistical units share a boundary or the spacing does not exceed 1 pixel, they are determined to be adjacent gray-scale statistical units. Based on the increased end opening, clear end boundary, and the extension and change of weak texture area, local broken area and gray-scale abrupt change area, the active state of the fractures in the late-stage penetrating layer is screened to determine the current active fracture layer. The active state screening includes retaining cracks with increased end opening, clear end boundaries, expanded weak texture area, enhanced local broken area, or continued gray-scale abrupt change area; and removing cracks that are covered, discontinuous, or rewritten by boundaries and do not exhibit increased end opening, expanded weak texture area, enhanced local broken area, or continued gray-scale abrupt change area. The weak texture area expansion state is that the weak texture area extends continuously along the end of the crack for a length of not less than 12 pixels, and the angle between the expansion direction and the end direction does not exceed 25 degrees. The enhanced state of the local fracture zone is characterized by an increase in the area of ​​the local fracture zone in the vicinity of the fracture end compared to the previous monitoring stage, and the shortest distance between the newly added local fracture zone and the fracture end does not exceed 6 pixels. The grayscale abrupt change area is a continuous arrangement of grayscale abrupt change areas along the crack end with a length of not less than 10 pixels, and the interval between adjacent grayscale abrupt change areas does not exceed 4 pixels. The fractures in the current active fracture layer are associated with the fractures in the later-stage penetrating layer to obtain the active inheritance relationship. The current active fracture layer, active end state, end neighborhood weakening state and active inheritance relationship are collected to generate the current active fracture layer identification data. The active connection relationship is the connection relationship formed by the fractures in the current active fracture layer after being screened and preserved by the late-stage through layer; the active end state is characterized by the end opening, end boundary, and end orientation; the weakening state of the end neighborhood is characterized by the expansion state of the weak texture area, the enhancement state of the local broken area, and the continuation state of the gray-scale abrupt change area.

[0025] In this embodiment, the generation of potential extension links and potential extension relationships of the crack specifically includes: Based on the current active fracture layer identification data, the fracture end to be extended for search is determined, and the fracture end sector search area is constructed with the fracture end as the starting point. The fracture end sector search area starts from the fracture end to be extended and extends forward along the direction of the fracture end with the end orientation as the center, for a period of not less than 40 pixels and not more than 80 pixels, and expands to both sides of the end orientation with a maximum angle of 35 degrees. When the fracture end sector search area coincides with the fracture target area in the current active fracture layer, the area in front of the fracture end to be extended and not occupied by the current active fracture layer is reserved as the fracture end sector search area. A continuous scanning band is set along the end direction within the search area of ​​the sector at the crack end, and the location of the weak texture area, local broken area and gray-scale change area within the continuous scanning band is identified to generate sector weakening feature distribution data. The continuous scanning bands are arranged sequentially along the end orientation, and the interval between adjacent continuous scanning bands does not exceed 2 pixels; the width of the continuous scanning band is 6 pixels, the center line of each continuous scanning band is perpendicular to the end orientation, and the center line of adjacent continuous scanning bands is shifted forward 2 pixels along the end orientation; when the continuous scanning band exceeds the boundary of the crack end sector search area, only the scanning band portion located within the crack end sector search area is retained to participate in position recognition. Location identification includes determining the location of the corresponding weakened feature within the search area of ​​the sector at the end of the crack, based on the identification conditions of weak texture area, locally broken area and gray-scale abrupt change area. The sector weakening feature distribution data includes the location of weak texture area, local broken area, gray-scale abrupt change area, and the arrangement order of weak texture area, local broken area, and gray-scale abrupt change area along the end direction; when the weak texture area, local broken area, or gray-scale abrupt change area is located outside the boundary of the sector search area at the end of the crack, the weakening feature part located outside the boundary is removed. Weak texture areas, locally fragmented areas, or gray-scale abrupt change areas within adjacent continuous scan bands are connected sequentially according to their end orientation to form a weakening path, and candidate extension paths are determined based on the continuous penetration state of the weakening path. The continuous through state includes a center distance of no more than 8 pixels between connected regions within adjacent continuous scan bands, an angle of no more than 25 degrees between the connecting line direction of the connected regions and the end orientation, and a continuous extension length of the weakening path along the end orientation of no less than 18 pixels; the center of the connected region is the center position of the bounding rectangle of the region. The weakening paths are connected sequentially from the nearest to the farthest end of the crack to be extended; when there are multiple weakening features in the same continuous scan band, all weakening features whose center distance from the connected area in the previous continuous scan band is no more than 8 pixels and whose connection direction is at an angle of no more than 25 degrees with the end orientation are included in the weakening path; when there are multiple weakening paths that meet the continuous penetration condition, each weakening path is used as a candidate extension path. Based on the end position and end orientation of the candidate extension path, determine the area in front of the end to which the candidate extension path points, and take this area in front of the end as the area to be extended; The region to be extended is a local image region formed by extending the candidate extension path 20 pixels along the end direction. When the overlap area between the region to be extended and the identified fracture target region is not less than 20% of the area of ​​the region to be extended, the region to be extended is updated to be the region located in front of the end of the candidate extension path and not occupied by the current active fracture layer. When the length of the updated region to be extended along the end direction is less than 10 pixels, the corresponding candidate extension path is removed. When there are weakened features in the region to be extended and the area of ​​the weakened features accounts for not less than 15% of the area of ​​the region to be extended, the corresponding weakened features are included in the envelope verification range of the region to be extended. An end-closed envelope region is constructed around the candidate extension path, and envelope verification is performed on the continuity of weak texture regions, the arrangement direction of local broken regions, and the continuity state of gray-scale abrupt change regions within the end-closed envelope region. The end-closed envelope region is formed by extending 6 pixels to both sides of the candidate extension path as the center line, and covers the continuous local range between the fracture end to be extended, the candidate extension path, and the region to be extended. When the end-closed envelope region exceeds the fracture end sector search area, the envelope portion located within the fracture end sector search area is retained to participate in the envelope verification. When the end-closed envelope region coincides with the fracture target region in the current active fracture layer, except for the region adjacent to the fracture end to be extended, the overlapping portion in front of the fracture end to be extended is retained, and the overlapping portion behind the fracture end to be extended is discarded. The continuity of weak texture regions is defined as the continuous distribution of weak texture regions along the candidate extension path within the end closed envelope region, and the interval between adjacent weak texture regions does not exceed 5 pixels. The arrangement direction of the localized fractured zones is such that the localized fractured zones are arranged along the candidate extension path, and the angle between the arrangement direction and the end orientation does not exceed 25 degrees; The gray-scale abrupt change region continuation state is when the gray-scale abrupt change region appears continuously along the candidate extension path, and the interval between adjacent gray-scale abrupt change regions does not exceed 4 pixels; Candidate extension paths are selected based on the envelope verification results. Candidate extension paths that satisfy the envelope continuity are identified as potential extension links of the fracture. Potential extension relationships are generated based on the link pointing relationship between the current active fracture layer and the region to be extended. Envelope continuity is achieved when any one of the following states is met: weak texture region continuity, local broken region arrangement direction, or gray-scale abrupt change region continuity, and the corresponding state covers at least 60% of the candidate extension path length; when multiple candidate extension paths satisfy envelope continuity, each candidate extension path is generated as a potential fracture extension link without path priority selection. The potential extension link of the fracture points from the end of the fracture to be extended to the region to be extended; The potential extension relationship is the relationship between the currently active fracture layer and the region to be extended via the potential extension link of the fracture.

[0026] In this embodiment, the generation of the fracture evolution trend map of the base plate failure zone specifically includes: The relative age layer data of fractures, the identification data of currently active fracture layers, the potential extension links of fractures and potential extension relationships are organized into map components to generate fracture evolution map composition data. The organization of map components includes extracting early fracture layers, intermediate modified layers, late connected layers and interlayer progression relationships from fracture relative age layer data; extracting currently active fracture layers and active connection relationships from currently active fracture layer identification data; extracting areas to be extended from potential fracture extension links; and extracting the link pointing relationship between currently active fracture layers and areas to be extended from potential extension relationships. Based on the fracture evolution map composition data, map nodes are configured, with the early fracture layer, the intermediate modified layer, the late through layer, the currently active fracture layer, and the region to be extended respectively configured as map nodes; The map node configuration includes configuring the early fracture layer as the early fracture layer node, the intermediate modified layer as the intermediate modified layer node, the late penetrating layer as the late penetrating layer node, the currently active fracture layer as the currently active fracture layer node, and the region to be extended as the region to be extended node. Map nodes carry the node source and spatial location range. The node source is used to indicate that the map node is generated from relative fracture chronology data, currently active fracture layer identification data, or potential fracture extension links; the spatial location range is determined by the outer boundary of the fracture target area or the area to be extended contained in the corresponding map node. Graph nodes are connected according to inter-layer progression, active succession, and potential extension relationships to obtain graph relationship edges; The graph relationship edges include inter-layer progression relationship edges, active succession relationship edges, and potential extension relationship edges. Inter-layer progression relationship edges are formed by connecting early fracture layer nodes, intermediate modified layer nodes, and late-stage connected layer nodes based on the inter-layer progression relationship in the relative age layer data of fractures. Active succession relationship edges are formed by connecting late-stage connected layer nodes and currently active fracture layer nodes based on the active succession relationship in the currently active fracture layer identification data. Potential extension relationship edges are formed by connecting currently active fracture layer nodes and nodes of the area to be extended based on the potential extension relationship corresponding to the potential extension link of the fracture. The graph nodes and graph relationship edges are collected according to the order of fracture evolution to generate a graph of the fracture evolution trend of the base plate failure zone. The atlas collection includes arranging nodes in the order of early fracture layer nodes, mid-term modified layer nodes, late-term through-layer nodes, currently active fracture layer nodes, and nodes in areas to be extended, and then collecting the atlas nodes and atlas relationship edges into an atlas of the evolution trend of fractures in the basement failure zone.

[0027] Example 1: To verify the feasibility of this invention in practice, it was applied to a scenario of monitoring the evolution of fractures in the floor failure zone of a coal mine longwall face. During the advancement of this longwall face, the floor was affected by mining and formed multiple stages of fractures. Some early fractures were covered by later fracture textures, and some fractures were cut, dislocated, and rewritten at their intersections. It was difficult to determine the order of fracture formation based on a single-stage image, and it was also difficult to distinguish between stable fractures and active fractures that might continue to extend.

[0028] During implementation, images of the base plate failure zone were collected before the working face advanced, and at 20m, 40m, 60m, and 80m of advancement. Twenty-four images were collected at each monitoring stage, with a single original image resolution of 1280×960 pixels. After inputting the images from each stage into the method of this invention, the images were first denoised, grayscale equalized, scaled, and distortion corrected, resulting in a uniform 1024×1024 pixel processed image, with an effective identification area of ​​960×960 pixels. Subsequently, continuous dark areas, linear texture areas, and suspected fracture texture areas were identified in the images from each stage. Dark areas with a grayscale difference greater than 20 and at least 12 consecutive pixels were used as the basis for crack identification, while linear texture areas with a single segment extension length of at least 18 pixels and a breakpoint spacing of no more than 3 pixels were used as the basis for crack extension.

[0029] In the crack relationship analysis, cracks with intersecting centerlines, adjacent boundaries, and end distances not exceeding 8 pixels, or overlapping texture abnormal areas within the same monitoring stage, are selected as candidate interactive crack pairs. A local interactive region is formed by extending 12 pixels outward from the interactive position. For crack junctions, a 15-pixel junction analysis window is used to identify the continuous state of the skeleton, the discontinuous position of the boundary, and the direction of broken texture coverage. The sequential constraint conflicts are resolved according to the order of boundary rewriting relationship, coverage relationship, discontinuity relationship, cutting relationship, and overlay relationship, thereby dividing the early crack layer, the intermediate modification layer, and the late through layer.

[0030] In active fracture identification, the fracture end of the late-stage through-layer is extended forward by 20 pixels and to each side by 10 pixels to form an end neighborhood. This identifies features such as increased end opening, clear end boundaries, expansion of weak texture areas, enhancement of local fragmentation areas, and continuation of grayscale abrupt change areas. For the retained currently active fracture ends, a sector search area of ​​40 to 80 pixels is constructed forward with the end orientation as the center. A continuous scan band with a width of 6 pixels and an interval of 2 pixels is set, connecting weakened features within the sector to form candidate extension paths. The continuity of weak texture areas, the arrangement direction of local fragmentation areas, and the continuation of grayscale abrupt change areas are then verified through the closed envelope area of ​​the end. Finally, the early fracture layer, intermediate modified layer, late-stage through-layer, currently active fracture layer, and areas to be extended are configured as map nodes, and the interlayer progression relationship, active connection relationship, and potential extension relationship are configured as map relationship edges, forming a fracture evolution trend map of the base plate failure zone.

[0031] Using the aforementioned 120 images of the base plate damage zone as the same batch of data, and comparing them with a traditional method that combines single-stage grayscale thresholding with crack skeleton extraction, the following performance statistics were obtained: Table 1. Performance Comparison of the Invention and Traditional Mapping Methods

[0032] As can be clearly seen from Table 1, the method of the present invention is superior to the traditional method in many indicators.

[0033] The method of this invention improves the accuracy of crack target identification from 86.4% to 91.7% of the traditional method, an improvement of 5.3 percentage points. The main reason is that this invention does not identify cracks based solely on gray-scale abrupt changes, but uses continuous dark areas, linear texture areas, crack center lines, crack boundaries, and crack ends together for crack target identification, making the distinction between crack target areas and blocky dark areas and isolated noise more stable.

[0034] Regarding the accuracy of texture anomaly recognition, the traditional method achieves 82.9%, while the method of this invention achieves 88.1%, an improvement of 5.2 percentage points. This is because the present invention further identifies the texture extension direction, local gray-scale abrupt change state, and local boundary integrity, which can separate broken textures from ordinary rock mass textures and reduce the situation of misjudging non-fracture textures as fracture anomalies.

[0035] In terms of the accuracy of determining the fracture overlay rewriting relationship, the method of the present invention reaches 85.4%, which is 6.8 percentage points higher than the 78.6% of the traditional method. This is because the present invention determines the cutting relationship, covering relationship, discontinuity relationship, overlay relationship and boundary rewriting relationship separately, instead of making a rough judgment based solely on whether the fractures intersect. Therefore, it expresses the relationship of complex intersection areas more completely.

[0036] Regarding the accuracy of relative age layer classification of fractures, the method of this invention improved from 76.8% to 83.2%, an increase of 6.4 percentage points. This is because the invention combines the occlusion direction of fracture intersections and the resolution of conflicts between sequential constraints, unifying multiple local relationships into early fracture layers, intermediate modified layers, and late connected layers, thereby reducing the level error caused by the judgment of a single intersection.

[0037] Regarding the accuracy of current active crack identification, the method of this invention achieves 86.0%, which is 5.5 percentage points higher than the 80.5% of the traditional method. This is because the present invention further screens the end opening, end boundary, weak texture area expansion, local broken area enhancement, and gray-scale abrupt change area continuation status of the late-stage through-layer, which can exclude cracks that have been covered, discontinuous, or whose boundaries have been rewritten and no longer extend.

[0038] Regarding the accuracy of identifying potential extended links, the method of this invention improved from 74.9% to 81.6%, an increase of 6.7 percentage points. This is because the invention uses a joint constraint of the fracture end sector search area, continuous scanning band, and end closed envelope area, ensuring that potential extended links simultaneously satisfy directional continuity and envelope continuity. The completeness rate of map relation edges improved from 79.3% to 86.5%, an increase of 7.2 percentage points. This is because the invention unifies the interlayer progression relationship, active succession relationship, and potential extension relationship into map relation edges, making the node connections in the fracture evolution trend map of the base plate failure zone more complete. The single-stage image processing time was reduced from 3.8s to 3.3s, a reduction of 0.5s. This is because the invention performs end sector search based on the current active fracture layer, reducing the processing workload of extending the scan across the entire map.

[0039] The above are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for constructing a fracture evolution trend map of a base plate failure zone based on image recognition, characterized in that, Includes the following steps: Images of the base plate failure zone at different monitoring stages were acquired, and the images were processed according to imaging standards and marked with stages to generate stage images of the base plate failure zone. Crack target identification and texture anomaly identification are performed on stage images of the base plate failure zone to generate crack image element data; Determine the relationship between the spatial interaction state of fractures in fracture image feature data, and generate fracture superposition rewriting relationship; The direction of occlusion at fracture confluence is identified based on the fracture superposition rewriting relationship, and the fracture formation stage is hierarchically inverted by combining the sequential constraint conflict resolution to generate fracture relative age layer data. Active state screening is performed on late-stage interconnected layers in relative age-stratigraphic data of fractures to generate current active fracture layer identification data; Based on the current active fracture layer identification data, a fracture end sector search region is constructed, and the weakened features within the sector are continuously scanned and envelope verified to generate potential fracture extension links and potential extension relationships. Based on relative age layer data of fractures, identification data of currently active fracture layers, potential extension links of fractures, and potential extension relationships, a map of fracture evolution trend in the basement failure zone is generated by configuring map nodes and arranging relationship edges.

2. The method for constructing a fracture evolution trend map of a base plate based on image recognition according to claim 1, characterized in that, The imaging standardization processing includes image denoising, grayscale equalization, scale unification, and distortion correction.

3. The method for constructing a fracture evolution trend map of a base plate based on image recognition according to claim 1, characterized in that, The generation of the crack image feature data specifically includes: The continuous dark areas and linear texture areas in the stage images of the base plate failure zone are jointly extracted, and suspected crack areas and suspected fracture texture areas are identified by combining the monitoring stage markers. Connectivity domain arrangement, extended morphology screening and fracture structure element extraction are performed on suspected fracture regions to generate fracture target identification data; The texture extension direction, local gray-level abrupt state, and local boundary integrity of suspected broken texture areas are identified to generate texture anomaly identification data. The crack target identification data and texture anomaly identification data are collected according to the monitoring stage and spatial location to generate crack image element data.

4. The method for constructing a fracture evolution trend map of a base plate based on image recognition according to claim 1, characterized in that, The generation of the fracture superposition rewriting relationship specifically includes: Based on the crack image element data, cracks that have spatial interaction within the same monitoring stage are combined and screened to obtain candidate interactive crack pairs and their local interactive regions. The continuous state of the skeleton is traced along the direction of the fracture centerline in the candidate interactive fracture pair, and when the continuous state of the skeleton and the continuity feature across another fracture boundary exist at the same time, the corresponding fracture spatial interaction state is determined as the cutting relationship. Within the local interaction region of the candidate interactive crack pair, the position of the texture anomaly region coincides with the original crack boundary, and the crack spatial interaction state where the coincident position changes discontinuously with the original crack boundary is determined as the coverage relationship. Within a local interaction region, the location of the misalignment of the fracture centerline and the path of another fracture centerline are traced. When the misalignment location is crossed by the path, the corresponding fracture spatial interaction state is determined as a fault relationship. The distribution of texture anomaly regions in different fracture target regions is compared within the local interaction region. When the texture anomaly regions overlap each other without forming a covering, discontinuity or cutting, the corresponding fracture spatial interaction state is determined as an overlay relationship. Within the local interaction area, the original crack boundary disappearance position is compared with the newly formed broken texture occupation position and the new crack boundary occupation position. When any occupation position coincides with the boundary disappearance position, the corresponding crack space interaction state is determined as the boundary rewriting relationship. The cutting relationship, covering relationship, discontinuing relationship, overlay relationship, and boundary rewriting relationship are aggregated according to the monitoring stage, candidate interactive fracture pairs, and relationship type to generate fracture overlay rewriting relationship.

5. The method for constructing a fracture evolution trend map of a base plate based on image recognition according to claim 1, characterized in that, The generation of the relative age-stratigraphic data of the fractures specifically includes: Based on the crack overlay rewriting relationship, the continuous state of the skeleton, the discontinuous position of the boundary and the coverage direction of the broken texture at the crack junction are extracted to generate junction occlusion recognition data; Based on the intersection occlusion recognition data, determine the first and second cracks in the candidate interactive crack pairs, and generate crack sequence labeling data; Based on the fracture overlay rewriting relationship and fracture sequence marking data, the formation order direction between candidate interactive fracture pairs is determined, and the formation order direction is used as the fracture sequence constraint. For the same candidate interactive fracture pair, the direction of the fracture constraints is compared and the type is verified, and the constraint state with opposite constraint directions or multiple relationship types is determined as a constraint conflict. In candidate interactive gap pairs with conflicting sequential constraints, the sequential constraints of the gaps are selected and discarded in the order of boundary rewriting relationship, covering relationship, discontinuation relationship, cutting relationship and overlay relationship to obtain the effective sequential constraints. Based on the effective priority constraint, the distribution of the number of times of the target fracture area is statistically analyzed, and the fracture is hierarchically classified according to the number of times it occurred first, the state of modification and acceptance, and the performance of the penetration, to generate fracture hierarchical classification data. The fracture hierarchical classification data includes early fracture layer, mid-term modification layer and late penetration layer. Based on the fracture hierarchy data and effective sequential constraints, the interlayer progression relationship between different fracture layers is determined, and the fracture hierarchy data and interlayer progression relationship are aggregated to generate fracture relative age layer data.

6. The method for constructing a fracture evolution trend map of a base plate based on image recognition according to claim 1, characterized in that, The generation of the currently active fracture layer identification data specifically includes: The end states of late-stage continuous fractures in relative age-determined fracture data are arranged to determine the end aperture, end boundary, and end orientation of late-stage continuous fractures, thereby generating end state data of late-stage continuous fractures. Based on the end state data of late-stage through-cracks, the changes in end opening and the clear state of end boundaries are identified, and weak texture areas, local fragmentation areas and gray-scale abrupt change areas in the neighborhood of the end are identified along the end orientation of the late-stage through-cracks. Based on the increased end opening, clear end boundary, and the extension and change of weak texture area, local broken area and gray-scale abrupt change area, the active state of the fractures in the late-stage penetrating layer is screened to determine the current active fracture layer. The fractures in the current active fracture layer are associated with the fractures in the later-stage through-layer to obtain the active inheritance relationship. The current active fracture layer, active end state, end neighborhood weakening state and active inheritance relationship are collected to generate the current active fracture layer identification data.

7. The method for constructing a fracture evolution trend map of a base plate based on image recognition according to claim 1, characterized in that, The generation of the potential extension links and potential extension relationships of the crack specifically includes: Based on the current active fracture layer identification data, the fracture end to be extended for search is determined, and the fracture end sector search area is constructed with the fracture end as the starting point. A continuous scanning band is set along the end direction within the search area of ​​the sector at the crack end, and the location of the weak texture area, local broken area and gray-scale change area within the continuous scanning band is identified to generate sector weakening feature distribution data. Weak texture areas, locally fragmented areas, or gray-scale abrupt change areas within adjacent continuous scan bands are connected sequentially according to their end orientation to form a weakening path, and candidate extension paths are determined based on the continuous penetration state of the weakening path. Based on the end position and end orientation of the candidate extension path, the region in front of the end to which the candidate extension path points is determined, and the region in front of the end is taken as the region to be extended. An end-closed envelope region is constructed around the candidate extension path, and envelope verification is performed on the continuity of weak texture regions, the arrangement direction of local broken regions, and the continuity state of gray-scale abrupt change regions within the end-closed envelope region. Candidate extension paths are selected based on the envelope verification results. Candidate extension paths that satisfy envelope continuity are identified as potential extension links for the fracture. Potential extension relationships are generated based on the link pointing relationship between the current active fracture layer and the region to be extended.

8. The method for constructing a fracture evolution trend map of a base plate based on image recognition according to claim 1, characterized in that, The generation of the fracture evolution trend map of the base plate failure zone specifically includes: The relative age layer data of fractures, the identification data of currently active fracture layers, the potential extension links of fractures and potential extension relationships are sorted out to generate fracture evolution map composition data. Based on the fracture evolution map composition data, map nodes are configured, with the early fracture layer, the intermediate modified layer, the late through layer, the currently active fracture layer, and the region to be extended respectively configured as map nodes; Connect the graph nodes according to the interlayer progression relationship, active succession relationship and potential extension relationship to obtain the graph relationship edge; The graph nodes and graph relationship edges are collected according to the order of fracture evolution to generate a graph of the fracture evolution trend of the base plate failure zone.

Images