Loess collapsibility evaluation and calculation method based on image semantic segmentation

Abstract

The invention relates to the technical field of collapsibility evaluation, in particular to a loess collapsibility evaluation calculation method based on image semantic segmentation, which comprises the following steps of: analyzing spatial heterogeneous characteristics of a monitoring area, adjusting sampling step length and resolution, and calculating temperature, humidity, stress and fracture form change based on parameters such as reflectivity, chromaticity and texture entropy of a remote sensing image; and judging the relationship between fracture communication and a pore structure, and completing collapsibility grading. According to the method, multi-dimensional fusion is realized through reflectivity, chromaticity and texture feature information of remote sensing images, a spatial heterostructure is automatically analyzed, a high-change area and a stable area are automatically identified in different monitoring areas, a sampling strategy is adaptively adjusted according to information distribution, and the resolution and the sampling density are linked to respond to regional dynamics. Real-time tracking and grading judgment of structural changes such as microscopic fracture extension and pore aggregation are achieved, automatic marking of spatial partition attributes and risk categories is achieved in combination with multi-source data, and spatial information content and grading types are greatly enriched.

Description

Technical Field

[0001] This invention relates to the field of collapsibility evaluation technology, and in particular to a method for evaluating and calculating the collapsibility of loess based on image semantic segmentation. Background Technology

[0002] Collapsibility assessment technology involves a comprehensive evaluation method for collapsible soils, especially in foundation engineering in loess regions, including collapsibility determination, classification, structural damage identification, and bearing capacity analysis. This technical field includes quantitative or qualitative analysis of the structural, water-bearing, fracture evolution, micropore distribution, and foundation bearing capacity and settlement engineering characteristics of collapsible loess. Traditional collapsibility assessment involves conducting indoor geotechnical tests on manually collected soil samples, such as collapsibility tests, moisture content tests, and triaxial shear tests, combined with empirical criteria to determine the collapsibility level or calculate settlement. This typically involves physical parameter determination and empirical formula extrapolation, primarily based on sample deformation, critical moisture content, and collapsibility sensitivity under external loads.

[0003] Traditional techniques assess collapsibility through physical sampling and experimental testing, but can only obtain soil parameters based on point distribution. This cannot express the changing trends over a large area. Collapsibility zoning relies on static samples and limited cross-sections, making it difficult to fully depict environmental changes and risk development processes. Block structure and spatial characteristics are not distinguished in the judgment. When the sampling interval is large, some spatial details are easily missed. The results are presented as a single level and limited description, with blurred risk boundaries and insufficient spatial expressiveness, resulting in one-sided and lagging level classification and structural identification in complex scenarios. Summary of the Invention

[0004] To address the shortcomings of existing technologies that assess collapsibility through physical sampling and experimental testing, which only provide soil parameters based on point distribution and fail to express changes over a large spatial area, collapsibility zoning relies on static samples and limited cross-sections, making it difficult to fully characterize environmental changes and risk development processes. Furthermore, it lacks differentiation between block structure and spatial characteristics in the assessment, and some spatial details are easily overlooked when sampling intervals are large. The results are characterized by single-level and limited descriptions, blurred risk boundaries, and insufficient spatial expressiveness, leading to one-sided and lagging technical problems in grade classification and structural identification in complex scenarios. This invention provides a method for evaluating and calculating loess collapsibility based on image semantic segmentation. The technical solution is as follows: On the one hand, a method for evaluating and calculating the collapsibility of loess based on image semantic segmentation is provided, including the following steps: S1: Based on remote sensing image reflectance data, collect chromaticity and texture entropy data, analyze the reflectance fluctuation and chromaticity distribution differences of each grid cell, normalize the texture entropy, screen the key blocks with parameter variation amplitude, and calculate the spatial distribution characteristics to obtain the spatial heterogeneous distribution structure. S2: Based on the spatial heterogeneous distribution structure, determine the level changes of each monitoring block, analyze the spatial location of blocks with frequent fluctuations, shorten the sampling step size and increase the resolution for high-change blocks, and increase the step size and decrease the resolution for low-change blocks to obtain the optimal sampling configuration. S3: Based on the optimal sampling configuration, calculate the temperature change trend of each block, analyze the dynamics of humidity change, soil stress and water content, compare the morphology of the crack boundaries in the images during the abrupt change period, determine the regional expansion, and obtain the crack change response factor. S4: Based on the crack change response factor, analyze the crack path connectivity of each block, determine whether the crack connects the block boundary, calculate the crack length and boundary ratio, compare the proportion of pore pixels, integrate the judgment rules, and obtain the connected pore structure index.

[0005] On the other hand, the spatial heterogeneous distribution structure includes a set of heterogeneous parameters, aggregated distribution information, and spatial structural features; the crack change response factor includes sensitivity parameters, extended criteria, and response type; and the connected pore structure index includes structural classification, correlation characteristics, and index system.

[0006] On the other hand, the reflectance fluctuation trend refers to the magnitude and direction of change of the remote sensing reflectance of pixels within the same monitoring grid over time or within a certain period or spatial range. The chromaticity distribution difference refers to the distribution changes, comparisons, or deviations of chromaticity parameters between different pixels or different monitoring grids in the same area, reflecting differences in surface characteristics.

[0007] On the other hand, the texture entropy refers to the measurement of the gray-scale distribution pattern of pixels in a remote sensing image, and the sampling distribution refers to the spatial distribution pattern and distribution method of sampling points or sampling units within the evaluation area.

[0008] On the other hand, the specific steps for obtaining the spatially heterogeneous distribution structure are as follows: S101: Based on remote sensing image reflectance data, the reflectance sequence of each pixel in the monitoring grid is differentially analyzed in a continuous observation period, the trend of reflectance change of each pixel over time is statistically analyzed, the change gradient of pixels in each region is compared, the spatial distribution location of the cluster of change trends in the region is determined, and the change gradient distribution is obtained. S102: Based on the change gradient distribution, compare the chroma channel distribution of each block of pixels, calculate the difference between the pixels in the block and the average chroma level of the block, sort by chroma offset, optimize and select representative sample blocks, and normalize and organize the block texture distribution to obtain a representative texture factor group. S103: Based on the representative texture factor group, calculate the reflectance change trend, chromaticity shift and normalized texture distribution of the corresponding block, jointly analyze the spatial distribution pattern of the three parameters, screen the monitoring blocks with spatial representativeness, determine the aggregation relationship between their spatial coordinates, and obtain the spatial heterogeneous distribution structure.

[0009] On the other hand, the steps for obtaining the optimal sampling configuration are as follows: S201: Based on the spatial heterogeneous distribution structure, analyze the heterogeneous distribution characteristics of the monitoring blocks, statistically analyze the change data of each block in the continuous sampling period, compare the heterogeneous change between each block and adjacent blocks, and classify and sort the blocks according to the change trend and the degree of data dispersion to obtain spatial classification parameters. S202: Based on the spatial classification parameters, determine the spatial location of the blocks with active heterogeneous changes, extract the remote sensing sampling primitives of the associated blocks, optimize and adjust the sampling step size of the blocks, calculate the spatial density of the sampling point distribution before and after adjustment, organize the sampling parameters of each block after adjustment, and obtain the distribution density factor. S203: Based on the distribution density factor, screen monitoring blocks with small level fluctuations, extend the sampling step size of the corresponding blocks, analyze the continuity of remote sensing area coverage after adjusting the sampling interval, integrate the distribution of blocks with different sampling densities, judge the overall spatial coverage balance, and obtain the optimal sampling configuration.

[0010] On the other hand, the specific steps for obtaining the crack change response factor are as follows: S301: Based on the optimal sampling configuration, calculate the temperature change of each block in a continuous time period, analyze the average temperature level and fluctuation trend in the sampling period, determine the direction of temperature increase or decrease in each period, organize the trend parameters, screen the key change stages, and obtain the temperature change coefficient group. S302: Based on the temperature change coefficient group, analyze the fluctuation structure of humidity data in each block within the cycle, identify the boundary points of phased rise or fall in the data, statistically analyze the humidity change characteristics within the segment, compare the correlation between the segment and the temperature trend, and obtain the segmented humidity characteristic set. S303: Based on the segmented humidity characteristic set, determine the pixel boundary changes of the crack region image before and after the mutation, compare the arrangement and spatial coverage of the crack pixel boundary structure, screen the key change groups of the region expansion, calculate the change amplitude of its region boundary, and obtain the crack change response factor.

[0011] On the other hand, the steps for obtaining the interconnected pore structure index are as follows: S401: Based on the crack change response factor, analyze the spatial extension of the crack path within the monitoring block, determine whether the crack cell arrangement is continuously distributed, compare the contact situation of the crack path on the boundary, calculate the degree of penetration of the crack path within the block, filter the path that penetrates to the block boundary, and obtain the spatial path connectivity. S402: Based on the spatial path connectivity, compare the matching between the crack path length and the boundary length of each monitoring block, screen the blocks with key path extension in the boundary direction, analyze the number of pore pixels in the same block and their coverage area, and statistically calculate the proportion of pores in the crack path-covered block to obtain the structural overlap ratio coefficient. S403: Based on the structural overlap ratio coefficient, determine the spatial combination distribution of crack and pore structures, analyze the arrangement pattern of the intersecting areas of crack paths and pore boundaries, screen the densely intersecting spatial structures, statistically analyze the spatial distribution of structural arrangement, and obtain the connected pore structure index.

[0012] On the other hand, the method also includes: S5: Based on the connected pore structure index, determine the crack penetration and pore aggregation performance in the monitoring unit, analyze the spatial overlap between the two, optimize the wet collapse level classification standard, identify the level type of each unit, and obtain the spatial wet collapse classification type. The spatial wetland classification types include grade identification, risk identification, and zoning classification.

[0013] On the other hand, the specific steps for obtaining the spatial sinkhole classification type are as follows: S501: Based on the connected pore structure index, determine the spatial extension of the crack path within the monitoring unit, analyze the number and arrangement direction of the intersection points between the crack penetration trajectory and the boundary, count the coverage of the crack intersection points, and compare the distribution of the continuous penetration structure within the unit to obtain the spatial continuity parameter. S502: Based on the spatial continuity parameter, determine the aggregation of pore pixel distribution in each monitoring unit, count the arrangement density and center distance of pores in space, calculate the overlap ratio of cracks and pores in the image, identify the spatial distribution of crack penetration and pore aggregation and co-occurrence, and obtain the degree of spatial co-occurrence coupling. S503: Based on the spatial co-occurrence coupling degree, adjust the collapsibility classification parameters, determine whether the spatial structure of the monitoring unit has a continuous trend, classify and assign the spatial characteristics of the monitoring unit, and obtain the spatial collapsibility classification type.

[0014] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: By leveraging the reflectivity, chromaticity, and texture features of remote sensing images to achieve multi-dimensional fusion, automatic analysis of spatial heterogeneous structures, automatic identification of high-variance and stable areas in different monitoring zones, adaptive adjustment of sampling strategies based on information distribution, and linkage between resolution and sampling density to respond to regional dynamics, the system enables real-time tracking and hierarchical determination of structural changes such as micro-fracture expansion and pore aggregation. Combined with multi-source data, the system automatically labels spatial zoning attributes and risk categories, and simultaneously extracts the evolution process of surface structures and the boundaries of collapsible zones, greatly enriching the spatial information content and hierarchical types, and supporting fine-grained expression of risk status in continuous space. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart of the main steps of the present invention; Figure 2 This is a flowchart of steps S1 of the present invention; Figure 3 This is a flowchart of steps S2 of the present invention; Figure 4 This is a flowchart of steps S3 of the present invention; Figure 5 This is a flowchart of step S4 of the present invention; Figure 6 This is a flowchart of step S5 of the present invention. Detailed Implementation

[0017] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0018] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0019] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0020] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0021] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0022] This invention provides a method for evaluating and calculating the collapsibility of loess based on image semantic segmentation, such as... Figure 1 As shown, it includes the following steps: S1: Based on the reflectance data of remote sensing images, analyze the reflectance fluctuation trend of each pixel in the monitoring grid, compare the chromaticity distribution differences of the differential grid blocks, normalize the sampling group of texture entropy, screen blocks with large differences in reflectance and chromaticity, optimize the parameter configuration of representative blocks in combination with texture entropy data, and calculate the spatial distribution characteristics of each block to obtain the spatial heterogeneous distribution structure. S2: Based on the spatial heterogeneous distribution structure, determine the level changes of each monitoring block, analyze the spatial location of blocks with frequent heterogeneous fluctuations, shorten the sampling step size of the primitive blocks in the remote sensing image sampling, adjust the spatial resolution to the optimal density, increase the sampling step size and reduce the resolution for blocks with small changes, reconfigure the sampling distribution, and obtain the optimal sampling configuration. S3: Based on the optimal sampling configuration, calculate the temperature change trend curve of each block, analyze the stage change pattern of the humidity data sequence, determine the dynamic amplitude of soil stress and moisture content data, and use morphological change comparison to determine the regional expansion range of the crack pixel boundary structure in the images before and after the abrupt change, and obtain the crack change response factor. S4: Based on the crack change response factor, analyze the continuity status of crack paths in each block of the monitoring area, determine whether cracks have crossed regions, calculate the ratio of the total length of the crack path to the boundary length, compare the proportion of pore pixels in each block, optimize the judgment rules for the combination of cracks and pore structures, and obtain the connected pore structure index. S5: Based on the connectivity pore structure index, determine the crack penetration state and pore structure aggregation behavior within the monitoring unit, calculate the spatial structure where crack penetration and pore aggregation phenomena occur simultaneously, optimize the classification standard for wet collapse levels, identify the high, medium or low level of wet collapse spatial type for each monitoring unit, and obtain the spatial wet collapse classification type.

[0023] Spatial heterogeneous distribution structure includes heterogeneous parameter set, aggregated distribution information, and spatial structural characteristics; fracture change response factor includes sensitivity parameter, extension criterion, and response type; connected pore structure index includes structural classification, correlation characteristics, and index system; spatial collapse classification type includes level label, risk label, and zoning assignment.

[0024] In S1, a monitoring grid refers to a small rectangular or polygonal area formed after spatially dividing the evaluation area. Each small area serves as an independent monitoring and calculation unit, facilitating the systematic acquisition and processing of remote sensing data. A pixel refers to the smallest unit (i.e., a "pixel") in a remote sensing image or digital image, typically used to describe a specific point or area on the Earth's surface in a remote sensing image. The reflectance fluctuation trend refers to the magnitude and direction of change in the remote sensing reflectance of pixels within the same monitoring grid over time or within a certain period or spatial range. The chromaticity distribution difference refers to the chromaticity (color attribute) between different pixels or different monitoring grids in the same area. The distribution, variation, comparison, or deviation of parameters reflect differences in surface characteristics; texture entropy refers to the measurement of the gray-scale distribution pattern of pixels in remote sensing images, often used to measure the complexity and uniformity of surface images, with a larger value indicating a more complex structure; representative block parameter configuration refers to the selection of typical blocks that can represent the change characteristics of a region for a set of monitoring grids or pixels by analyzing parameter fluctuations (such as reflectance, chromaticity, texture entropy, etc.), thereby determining its main parameter set; spatial distribution characteristics refer to the spatial statistical characteristics of each monitoring grid under different parameter indices within the evaluation area, such as distribution pattern, distribution density, spatial clustering or dispersion.

[0025] In S2, the grade change refers to the analysis of the heterogeneity (or other grading standards) change level corresponding to each monitoring block based on the results of S1, clarifying which level each block belongs to and its change trend; blocks with frequent heterogeneity fluctuations refer to monitoring blocks where the heterogeneity index shows significant changes or frequent fluctuations in the spatial or temporal dimensions, often indicating active changes in surface attributes or structures; the sampling step size of the primitive block refers to the distance between continuously selected sampling points in space during the remote sensing image sampling process. The smaller the step size, the denser the sampling and the better the details can be captured; the shortening operation refers to adjusting the sampling step size to make it smaller, thereby increasing the sampling density; the optimal density refers to the most suitable density setting of the sampling resolution (i.e., spatial resolution) under the premise of meeting evaluation requirements and computational efficiency; the sampling distribution refers to the spatial distribution mode and distribution pattern of sampling points or sampling units in the entire evaluation area.

[0026] In S3, the temperature change trend curve refers to the continuous change curve of temperature data in the collected monitoring block over time or space, used to analyze environmental change characteristics; the stage change pattern refers to the stage change characteristics (such as periodic rise or fall) of parameters such as humidity within the monitoring period; soil stress refers to the stress state in loess caused by external forces, rainfall, wet-dry cycles, etc., and is an important parameter reflecting the physical and mechanical behavior of soil; dynamic amplitude refers to the amplitude and fluctuation of a parameter (such as soil stress, water content) during the monitoring process, not a static value; crack pixel boundary structure refers to the set of pixel boundary information formed by cracks inside or on the surface of soil after remote sensing or image segmentation, used to describe the crack morphology, distribution and structural relationship; morphological change comparison refers to comparing crack pixel boundary structures at different times or spaces to analyze whether the crack morphology, range, size, etc. have changed.

[0027] In S4, the fracture path refers to the spatial connection route of soil fractures in the loess region, formed by the serial connection of pixels, reflecting the extension, direction, and connectivity characteristics of the fractures; the continuity status refers to whether the fracture path can cross the boundary of the monitoring block to form a continuous connected structure, which is an important indicator for judging the impact range and structural integrity of the fractures; the proportional relationship refers to the quantitative relationship between the total length of the fracture path and parameters such as the length of the monitoring area boundary, used to quantitatively analyze the extensibility of the fractures; the porosity pixel ratio refers to the ratio of the number of pixels representing pores to the total number of pixels in a monitoring block, reflecting the degree of porosity development in the soil; the judgment rule refers to the standard or logical conditions established when jointly and simultaneously judging fracture and porosity parameters, used to comprehensively determine the risk level or zoning attributes.

[0028] In S5, a monitoring unit refers to a basic block that serves as an independent evaluation object after spatial division, and is the basic analytical unit of the entire evaluation calculation process; clustering behavior refers to the phenomenon of spatial structures such as pores or fissures agglomerating or clustering within a certain monitoring unit, manifested as spatial characteristics such as increased density and concentrated quantity; simultaneous spatial structures refer to the state in which the phenomena of fissure penetration and pore aggregation overlap or coexist in the same monitoring unit; the collapsibility level classification standard refers to the quantitative or qualitative standard used to classify the collapsibility level (such as high, medium, and low) of each monitoring unit, including various parameter judgment standards for fissure and pore structures.

[0029] like Figure 2 As shown, the specific steps for obtaining spatially heterogeneous distribution structures are as follows: S101: Based on remote sensing image reflectance data, the reflectance sequence of each pixel in the monitoring grid is differentially analyzed in a continuous observation period, the trend of reflectance change of each pixel over time is statistically analyzed, the change gradient of pixels in each region is compared, the spatial distribution location of the cluster of change trends in the region is determined, and the change gradient distribution is obtained. The evaluation area is divided into monitoring grids according to rules. Then, the historical reflectance values ​​of all pixels within each grid are extracted. For example, the reflectance of three time phases is extracted from images of a certain block over five days. Subsequently, the reflectance difference between two time points is calculated for each pixel in chronological order to obtain the magnitude and direction of reflectance increase or decrease. The differences are then organized by grid to determine the overall trend of reflectance change in the area where the pixel is located. For example, if the reflectance of a pixel decreases on the first and second days, and also decreases on the second and third days, it is determined to be a continuous decreasing trend. This result is then combined with other pixels for trend statistics. Finally, the reflectance change of all pixels within each monitoring area is calculated. The average value is used to express the overall rate of change in the region. Then, spatial clustering is performed according to the rate of change, that is, pixels with consistent trends and similar change magnitudes are grouped together. The pixels are marked according to their spatial location, and the distance between the actual coordinates of the pixels in each group is calculated to check whether the pixels have spatial concentration. For example, if a large number of pixels in a certain group are concentrated in multiple neighboring grids and the average distance between them is less than one sampling step, it is determined that it belongs to the trend clustering area. Finally, multiple clustering units with representative reflectance change trends are divided in the entire region, and the number of pixels in the unit and the average reflectance change are counted to form a set of change gradient distribution data.

[0030] S102: Based on the change gradient distribution, compare the chroma channel distribution of each block of pixels, calculate the difference between the pixels in the block and the average chroma level of the block, sort by chroma offset, optimize and select representative sample blocks, and normalize and organize the block texture distribution to obtain representative texture factor groups. First, the color value of each pixel in the image is extracted, and the average color value of all pixels in each block is calculated. Then, the numerical deviation between each pixel and the average color value of the block is calculated. The deviation values ​​are sorted and the average deviation of the block is calculated to reflect the color consistency of the entire block. Then, all blocks are arranged from high to low deviation, and the blocks with larger deviation values ​​are selected as representative sample blocks. For example, if a total of 200 blocks are extracted, the top 20 blocks with the largest color deviations are selected for the next step of processing. Subsequently, the texture information of the sample blocks is analyzed by reading the grayscale relationship of the image. The structure involves extracting image features from each sample block and then linearly adjusting the results according to the range of maximum and minimum values, so that all sample texture values ​​are mapped to between 0 and 1. For example, if the original texture value of a block is 2.6, which is at a medium to high level between the maximum and minimum values ​​of all sample blocks, it will be adjusted to about 0.63. Finally, the chromaticity deviation value, the adjusted texture value, and its spatial coordinate information of each sample block are integrated into a structured set of sample parameters and recorded in the form of a list or matrix for the next step of spatial filtering, forming a representative texture factor group.

[0031] S103: Based on the representative texture factor group, calculate the reflectance change trend, chromaticity shift and normalized texture distribution of the corresponding block, jointly analyze the spatial distribution pattern of the three parameters, screen the monitoring blocks with spatial representativeness, determine the aggregation relationship between their spatial coordinates, and obtain the spatial heterogeneous distribution structure. The reflectance change results, chromaticity deviation, and texture values ​​are reread according to the block division. Then, these three different dimensions of indicators are combined into a combined feature. Next, the difference between the combined features of each pair of blocks is measured to determine which blocks have smaller differences in these three indicators. Blocks with differences less than a certain reference level are grouped together. Then, the spatial distribution of each group of blocks is statistically analyzed. By calculating the center point coordinates of each block, it is analyzed whether it has spatial aggregation. For example, if multiple blocks in a group are distributed in a certain area and are close to each other, it is determined that the blocks belong to a spatial aggregation structure. Then, it is checked whether the spatial range of the aggregation structure is compact, that is, whether the maximum extension range is less than a predetermined standard. If this condition is met, it is confirmed that the blocks constitute a stable spatial heterogeneous structure, and its spatial distribution information and corresponding feature values ​​are recorded. Finally, blocks that have consistency in the three dimensions of reflectance change, chromaticity deviation, and texture features and have a high degree of spatial aggregation are extracted and defined as spatial heterogeneous distribution structure units.

[0032] like Figure 3 As shown, the specific steps for obtaining the optimal sampling configuration are as follows: S201: Based on the spatial heterogeneous distribution structure, analyze the heterogeneous distribution characteristics of the monitoring blocks, statistically analyze the change data of each block in the continuous sampling period, compare the heterogeneous change between each block and its adjacent blocks, and classify and sort the blocks according to the change trend and the degree of data dispersion to obtain spatial classification parameters. Each monitoring block is numbered and labeled according to the preset spatial partition boundaries. Then, the heterogeneity index change sequence within each block in a continuous sampling period is extracted. The time series data of each block is organized by day, week, or ten-day period. The heterogeneity parameters at each time point are interpolated to obtain information on the amount of time difference and the direction of change. Then, the mean, maximum, and fluctuation range of the heterogeneity change amplitude for each block are calculated, and the results are compared one-to-one with the corresponding values ​​of the same block in adjacent periods. For example, if the heterogeneity index values ​​of block A in weeks 1, 2, and 3 are 0.52, 0.48, and 0.60, then the change from week 1 to week 2 is −0.04, and from week 2 to week 3 is +0.12. The direction of change within the block reverses, and changes exceeding 0.1 are judged as high volatility. Then, the heterogeneity between adjacent blocks is further analyzed. The difference in the change value of the index is calculated as the absolute difference, and then it is judged whether it exceeds the set change threshold. If the difference in change between two blocks exceeds 0.08, the heterogeneity change between the two blocks is significant. Then, all blocks are sorted according to the total magnitude of heterogeneity change, and the dispersion of the heterogeneity change data of each block is calculated. The dispersion is described by the standard deviation of all change values, and then the standard deviation values ​​are divided into intervals. For example, less than 0.02 is defined as low dispersion, 0.02 to 0.06 is medium dispersion, and greater than 0.06 is high dispersion. Then, the change trend (referring to the increase or decrease of the index value) and the dispersion level are combined to score each block. Blocks with reversed trends score higher than blocks with unidirectional trends, and blocks with high dispersion score higher than blocks with low dispersion. A set of spatial level ranking results arranged from high to low scores is formed as spatial classification parameters.

[0033] S202: Based on spatial hierarchical parameters, determine the spatial location of blocks with active heterogeneous changes, extract remote sensing sampling primitives of related blocks, optimize and adjust the sampling step size of the blocks, calculate the spatial density of the sampling point distribution before and after adjustment, organize the sampling parameters of each block after adjustment, and obtain the distribution density factor. Extracting high-scoring blocks from the classification results, identifying areas with frequent heterogeneous changes, and extracting the spatial coordinates of these blocks, we extract corresponding sampling primitives from the original remote sensing image. We then read the sampling start point, sampling path, and sampling interval data of these primitives to determine the current sampling step size. For example, if the initial step size is 60 meters, and the heterogeneity level within the block is high, the step size is reduced to 30 meters. After adjustment, we recalculate the number of sampling points and the actual spatial coverage density within each block. The density is calculated by dividing the number of sampling points by the block area. For example, a block originally had 16 sampling points, covering an area of ​​3600 square meters, with a density of one point per 225 square meters. After adjustment, the number of sampling points increased to 36, with a density of one point per 100 square meters. Subsequently, for each adjusted block, new parameters such as sampling step size, primitive distribution coordinates, and sampling coverage were recorded. The parameters were then compiled into a comparison table with the data before adjustment, forming a parameter set of density changes obtained by each block after optimization. These parameters were sorted from largest to smallest sampling density and marked as distribution density factors for subsequent coverage consistency analysis.

[0034] S203: Based on the distribution density factor, screen monitoring blocks with small level fluctuations, extend the sampling step size of the corresponding blocks, analyze the continuity of remote sensing area coverage after adjusting the sampling interval, integrate the distribution of blocks with different sampling densities, judge the overall spatial coverage balance, and obtain the optimal sampling configuration. In the sampling density sorting, blocks with small density changes are screened out, i.e., blocks where the ratio of density change before and after adjustment is less than a set value. The heterogeneity level fluctuation of these blocks is then assessed to determine if it is low. If the heterogeneity parameter fluctuation amplitude of a block does not exceed 0.05 within three periods and there is no directional reversal, it is marked as a block with small fluctuation amplitude. Subsequently, the sampling step size for these blocks is extended, for example, from 30 meters to 60 meters. The sampling point spacing between primitives is then reset accordingly. Finally, all the changed sampling primitive positions are stitched together to simulate the continuity check of the remote sensing image coverage area, determining whether spatial discontinuities are formed between sampling points. If spatial discontinuities still exist between adjacent blocks after adjustment... If the spacing between at least one point is less than twice the step size, it is considered a continuous coverage. Then, the sampling density between each block is analyzed to see if there are abrupt changes. The abrupt change points are marked. After unifying the scale, the entire map is divided into high-density, medium-density, and low-density zones. By statistically analyzing the proportion of each type of block, it is determined whether there are local areas with excessively concentrated density or sparse edge areas in the current sampling configuration. If the proportion balance meets the preset balance standard (for example, the area ratio of the three types of regions does not differ by more than 15%), then the result of this round of adjustment is confirmed as the optimal sampling configuration. The final sampling spacing and corresponding distribution coordinates of each monitoring block are output to form the optimal sampling configuration of the continuous remote sensing image.

[0035] like Figure 4 As shown, the specific steps for obtaining the crack change response factor are as follows: S301: Based on the optimal sampling configuration, calculate the temperature change of each block in a continuous time period, analyze the average temperature level and fluctuation trend within the sampling period, determine the direction of temperature increase or decrease in each period, organize the trend parameters, screen the key change stages, and obtain the temperature change coefficient group. First, surface temperature data for each monitoring block is extracted at various points within a continuous time period based on the coordinates of its sampling points. The temperature values ​​at the same time point for all sampling points in each block are averaged to obtain the representative temperature value for that block at each time point. These values ​​are then arranged chronologically to form a temperature time series. Next, the temperature difference between adjacent time points in the series is calculated to determine the direction and magnitude of the change. Segments with positive temperature differences are marked as an upward trend, and those with negative differences are marked as a downward trend. The entire time series is then divided into several sub-stages based on segments with the same trend direction. For example, if the temperature increases daily for the first 3 days and decreases daily for the next 4 days, two stages are formed: an upward stage and a downward stage. The average temperature level and temperature change magnitude of each stage are then statistically analyzed. Finally, it is determined whether the stage trend of each block is stable. The trend judgment is based on the daily temperature increase or decrease within a certain stage. The differences did not exceed the set threshold, which can be 1.2 times the daily average change. For example, if the average change is 1.8℃, the threshold is set to 2.16℃. If the daily temperature difference in a certain period is 2.4℃, then the trend of that period is determined to be unstable. Then, the periodic data of all blocks are organized into parameter groups according to the direction and magnitude of temperature increase or decrease. The start and end times of the period, the direction of temperature change, the average change, and the trend fluctuation level are recorded. Then, it is determined whether it belongs to a critical change period based on whether the trend fluctuation level in the parameter group continuously exceeds the medium and high thresholds. The medium and high threshold standards can be set to a change magnitude greater than 3.0℃ and a continuous number of days not less than 3 days. If the temperature of a certain block rises from 26.1℃ to 29.8℃ in 3 consecutive days, then this period is selected as a critical change period. All the selected critical periods are summarized and their corresponding block numbers, time ranges, and trend characteristics are labeled to construct a temperature change coefficient group.

[0036] S302: Based on the temperature change coefficient group, analyze the fluctuation structure of humidity data in each block within the period, identify the boundary points of phased rise or fall in the data, statistically analyze the humidity change characteristics within the segment, compare the correlation between the segment and the temperature trend, and obtain the segmented humidity characteristic set. The formula is used to compare the arrangement and spatial coverage of the boundary structure of the cracked pixels: ; Calculate the response characteristic value of boundary arrangement change, calculate the magnitude of regional boundary change, and obtain the fracture change response factor, where... Representing the Response characteristic values ​​of boundary arrangement changes in each fracture sub-region Represents the total number of pixel boundary pairs. Representing the The first fracture sub-region The change in boundary orientation angle between pixel boundaries. Representing the The first fracture sub-region The spatial coordinate offset of a pixel boundary in the x-direction. Representing the The first fracture sub-region The spatial coordinate offset of a pixel boundary in the y-direction. Representing the The first fracture sub-region The boundary structure value of each pixel before the mutation. Representing the The first fracture sub-region The boundary structure value of each pixel boundary after the mutation Representing the The first fracture sub-region The mean value of the pixel boundary structure before and after the abrupt change; Boundary arrangement change response characteristic value is a numerical index used to quantitatively describe the arrangement and spatial structure changes of the boundary of a fracture region before and after abrupt changes. It reflects the overall response degree of the boundary direction, spatial displacement and structural features of fracture pixels within a certain region.

[0037] Indicates the first The response intensity of each crack sub-region to the changes in the image boundary structure before and after the mutation reflects the degree of mutation in the boundary arrangement and spatial coverage of the region. It is the first The total number of pixel boundary pairs participating in the analysis within the region is usually extracted by edge detection and then analyzed in pairs according to coordinates. It is the first The angle difference (in degrees) of the pixel boundary orientation before and after the abrupt change is obtained by calculating the orientation field using an image gradient method (such as Sobel). , It is the first The offset (in pixels) of each boundary pixel in the x and y coordinate directions is obtained based on the coordinate differences after image spatial registration; , The numbers before and after the mutation are respectively The structural values ​​of each boundary pixel (such as boundary gradient magnitude or texture intensity) are obtained through region structure quantization after image segmentation. The result is the average of the two. In the formula calculation, the absolute value sign handles the variation of the direction angle and the structural value, the square root operation measures the coupling of the two-dimensional spatial offset, and the fractional term reflects the intensity of structural variation. The three are weighted and integrated to form a multidimensional expression of the boundary response. Finally, the summation and averaging reflect the overall regional change level.

[0038] The following is the first image scene. Each crack sub-region Normalized data for pixel boundary points: First pair of pixel boundaries: , , , , , ; 2nd pair of pixel boundaries: , , , , , ; The third pair of pixel boundaries: , , , , , ; Calculate the representation value for each pair of pixel boundaries item by item: Item 1: ; Item 2: ; Item 3: ; Substituting the above three terms into the formula, calculate the characteristic value of the boundary arrangement change response in this region: ; This result indicates that the boundary arrangement change response characteristic value of the first fracture sub-region Falling into the preset high response range This indicates that the region underwent significant boundary orientation adjustment, spatial position shift, and structural feature reconstruction before and after the mutation. The formula introduces a spatial coupling offset term and a normalized ratio of structural changes, and performs unified aggregation of multi-dimensional features, making the expression adaptable and sensitive to different types of boundary deformation scenarios, and achieving high-dimensional quantitative response modeling in the process of identifying abrupt changes in crack structures.

[0039] The extremely low response range is Meaning: The fracture boundary changes very weakly, tending towards stability or error interference; the low response range is... Meaning: The fracture boundary has been slightly adjusted, and the regional boundary has changed weakly; the medium response range is... Meaning: The fracture boundary shows some changes, with significant morphological adjustments; the high-response range is... Meaning: Significant changes in fracture boundaries, with reconstruction of arrangement and structure; the extremely high response range is... Meaning: The fracture boundary changes drastically, accompanied by regional expansion or connection.

[0040] S303: Based on the segmented humidity characteristic set, determine the pixel boundary changes of the crack region image before and after the mutation, compare the arrangement and spatial coverage of the crack pixel boundary structure, screen the key change groups of the region expansion, calculate the change amplitude of the region boundary, and obtain the crack change response factor. For each temperature trend phase in a given area, the humidity observation data sequence for that area within the same period is retrieved. The humidity values ​​at each time point are arranged and plotted as a daily curve. The daily humidity change is then calculated using point-by-point differences to determine the period of continuous humidity increase or decrease. The inflection point where the trend direction changes is used as the segment boundary. For example, if humidity increases continuously from day 2 to day 5 and begins to decrease on day 6, day 5 is set as the boundary, dividing the area into an increasing segment and a decreasing segment. The average humidity level, maximum fluctuation amplitude, fluctuation direction, and duration within each segment are then statistically analyzed. For instance, if humidity rises from 12% to 18% over 4 days in a segment, with a change amplitude of 6%, this is considered a significant increasing segment, and this segment is recorded as a humidity characteristic record. Subsequently, its impact on the humidity is retrieved within each humidity segment. The system checks whether the time period completely or partially overlaps with a certain stage in the temperature change coefficient group. If an intersection occurs, it is marked as an "associated segment". Then, the temperature direction and humidity direction of each associated segment are extracted and their consistency of change is compared. If the two directions are the same (e.g., both are increasing), they are marked as positively correlated; if the directions are opposite, they are marked as negatively correlated. The number and ratio of positive and negative correlations occurring in each block throughout the entire observation period are then counted. For example, if a block experiences 4 stages of change, 3 of which are positively correlated and 1 is negatively correlated, the positive correlation ratio is 75%. The ratio information and segmentation features are then integrated into a block humidity segmentation characteristic set. Each record includes parameters such as block number, segment start and end dates, humidity change amplitude, temperature and humidity direction consistency indicator, and number of days of humidity change, forming a segmented humidity characteristic set.

[0041] like Figure 5 As shown, the specific steps for obtaining the connectivity pore structure index are as follows: S401: Based on the crack change response factor, analyze the spatial extension of crack paths within the monitoring block, determine whether the crack cell arrangement is continuous, compare the contact situation of crack paths at the boundary, calculate the degree of penetration of crack paths within the block, filter paths that penetrate to the block boundary, and obtain the spatial path connectivity. The set of fractured pixels within each monitoring block is extracted from the remote sensing image segmentation results. All continuous pixels are chained according to their adjacency in the 2D raster. The connectivity of the fracture path is identified by determining whether the difference between the row and column coordinates of adjacent pixels is equal to 1 or √2. If there is a gap of more than two pixels between two pixels or an interruption spanning multiple rows and columns in a path, the path is marked as discontinuous. Subsequently, the coordinates of the start and end points of each fracture path are extracted to determine whether it touches the block boundary. If either the start or end point coordinate value is equal to the maximum or minimum value of the block boundary (e.g., the block range is x: 0–60, y: 0–60, and the path endpoint is located at x=0 or x=60), the path is recorded. For each edge path, the number of pixels along the entire path is counted as the crack length. The degree of connectivity is determined by the ratio of the total path length to the contact points at the block boundary. For example, if the path length is 50 pixels and the endpoints reach the left and right boundaries, it is marked as a lateral connectivity path. If at least one section of the path connects to the boundary and the number of path interruptions is less than 3 pixels, it can be considered weak connectivity. Connectivity levels are set as strong connectivity, weak connectivity, and non-connectivity. The connectivity level and number of all crack paths are counted for each block. Then, the paths that are strong connectivity are selected as core samples, and their start and end directions and spatial extension trajectories are recorded. Finally, the path connectivity parameters are constructed based on the number of strong connectivity paths, average length, boundary contact rate, and other data in each block to obtain the spatial path connectivity.

[0042] S402: Based on spatial path connectivity, compare the matching between the crack path length and boundary length of each monitoring block, screen the blocks with key path extension in the boundary direction, analyze the number of pore pixels and their coverage area in the same block, and statistically calculate the proportion of pores in the crack path-covered block to obtain the structural overlap ratio coefficient. The length of the strongly penetrating fracture path in each monitoring block is compared with the block boundary perimeter. The percentage of the path length to the boundary length is calculated. If this percentage is greater than 30% and the directional extension is mainly along the longest side of the boundary, the block is marked as a critical extension block. For example, if a block has a side length of 60 meters and a path length of 24 meters, the percentage is 40%, and it is considered a critical extension block in the boundary direction. Then, the spatial distribution information of pore pixels in the block is extracted. The block image is then traversed again to statistically label the coordinates and number of all pore pixels. The pore area is matched according to the fracture path coverage area to determine whether there are fracture path pixels. For contact areas where pore cells overlap or are adjacent to each other by no more than one cell, the ratio of the number of contact pores to the total number of pores is calculated. For example, if a block has a total of 180 pore cells, and 76 of them are in contact with or cover the fracture path, then the coverage ratio is 42.2%. Subsequently, this coverage ratio is organized by block to construct a fracture-pore spatial overlap list for each block. Then, it is classified into levels according to preset thresholds. For example, a coverage ratio of less than 20% is weak overlap, 20%–40% is medium overlap, and more than 40% is strong overlap. Finally, all blocks are weighted and summarized according to the number of strong overlap blocks and the coverage ratio to output the structural overlap ratio coefficient.

[0043] S403: Based on the structural overlap ratio coefficient, determine the spatial combination distribution of crack and pore structures, analyze the arrangement pattern of the intersecting areas of crack paths and pore boundaries, screen the densely intersecting spatial structures, statistically analyze the spatial distribution of structural arrangement, and obtain the connected pore structure index. The arrangement pattern of the intersection region of fracture paths and pore boundaries is analyzed using the following formula: ; Calculate the staggered distribution characteristic values, filter out densely staggered spatial structures, statistically analyze the spatial distribution of structural arrangements, and obtain the connectivity porosity structure index. Representing the The monitoring blocks exhibit an interleaved distribution characteristic value. Representing the The number of pixel pairs in the region where fissures and pores intersect within the block. Representing the The pixel length of the crack path segment in an interlaced pixel pair Representing the The mean length of all crack path segments within the block. Representing the The normal distribution density of each staggered pixel to the corresponding pore boundary pixel. Representing the The percentage of local structural overlap of each interleaved pixel in its block. Representing the The standard deviation of the block structure overlap ratio coefficient Representing the Average connectivity of pixels in the block aperture; Interlaced distribution characteristic value is a numerical index used to quantify the spatial distribution characteristics of the interlaced area of ​​fracture paths and pore boundaries within a monitoring block. It can reflect the spatial heterogeneity of different monitoring blocks in terms of fracture connectivity, pore aggregation and spatial structure complexity, and is an important intermediate parameter for subsequent identification of connected pore structure indicators.

[0044] Given a certain number There are in the monitoring block The values ​​of each parameter after sampling and normalization for each of the interleaved pixel pairs are as follows: (normalized value) ), ( ), ( ), ( ), ( ); (normalized mean) ); (normalized value) ), ( ), ( ), ( ), ( ); ( ), ( ), ( ), ( ), ( ); ; ; Calculate according to the formula. The first item: ; ; ; ; ; ; Summation: ; Denominator calculation: ; ; ; ; The result indicates that the first The spatially interwoven structure of fracture paths and pore boundaries within the block exhibits low complexity and variability, with interwoven distribution characteristic values. The low value range indicates that the distribution of fracture path lengths in this block is relatively concentrated, the directional density variation of pore boundaries is small, and the overlapping areas of fracture and pore structures have high consistency and stability.

[0045] like Figure 6 As shown, the specific steps for obtaining the spatial collapse classification type are as follows: S501: Based on the interconnected pore structure index, determine the spatial extension of the crack path within the monitoring unit, analyze the number and arrangement direction of the intersection points between the crack penetration trajectory and the boundary, count the coverage of the crack intersection points, and compare the distribution of the continuous penetration structure within the unit to obtain the spatial continuity parameter. The monitoring area was divided into monitoring units. Each crack path within a monitoring unit was identified and numbered. The coordinates of the start and end points of each path were extracted to confirm its orientation in the image and determine its spatial extension. The path length and the continuity between its endpoints were measured to determine if there were any jumps or discontinuities. If the entire path consisted of continuous pixels and its extension direction remained consistent, it was recorded as a stable extension path. Then, for each path's intersection with the monitoring unit boundary, the number of intersection points on all four sides of the boundary was extracted. It was determined whether the intersection points were concentrated on one side or exhibited multi-sided continuity. If a path's start point was on the left boundary, its end point touched the right boundary, the number of intersection points was greater than or equal to 2, and the arrangement direction was close to horizontal, it was marked as a transverse continuity path. The path is identified, and the coordinates of all intersection points are spatially clustered to form intersection cluster areas. The coverage area is determined by counting the number of intersection points in each cluster area. For example, if there are 8 intersection points within a 20m×20m area, then the area is a high-density fracture intersection zone. Subsequently, the total number of transverse or longitudinal through paths and the density of intersection points in each monitoring unit are counted to see if they exhibit a clustered distribution. If multiple continuous paths are connected and the intersection points are highly concentrated, and the direction of the through paths in the area tends to be consistent, then the monitoring unit is marked as a unit with strong fracture extension continuity. The parameters such as the number of fracture paths, average direction, total number of intersection points, and area of ​​dense intersection areas are summarized to construct a spatial extension continuity index system and output spatial continuity parameters.

[0046] S502: Based on spatial continuity parameters, determine the aggregation of pore pixel distribution in each monitoring unit, statistically analyze the spatial arrangement density and center distance of pores, calculate the overlap ratio of cracks and pores in the image, identify the spatial distribution of crack penetration and pore aggregation and co-occurrence, and obtain the degree of spatial co-occurrence coupling. First, the spatial coordinates of the pixels corresponding to the pores in the image are extracted. Then, the image is divided into grids or unit regions. The pore coordinate sets in each monitoring unit are assigned to the corresponding unit. The proximity distance of this set is calculated to determine whether the pixels are concentrated or clustered. The distance between pore pixels is statistically analyzed, and a distance of less than 5 meters is used as a clustering criterion. The number and area coverage of pore clusters that meet this condition are counted. For example, if a monitoring unit has 200 pore pixels, and 140 of them are distributed in a clustered area with a distance of no more than 5 meters between them, then the pore clustering degree of this unit is 70%. Subsequently, the coordinates of the geometric center of the pore clustered area are extracted, and the distance from each pore pixel to the center point is calculated. The average distance is used as the center distance index to measure the degree of distribution concentration. If the center distance is lower than the set aggregation judgment threshold (e.g., 10 meters), it is classified as a high aggregation state. Then, the above-mentioned pore aggregation region is superimposed on the fracture path image, and the existence of spatial overlap is judged pixel by pixel. That is, whether the fracture pixel and the pore pixel overlap in space or are no more than 1 pixel apart. The ratio of the number of pixels in the overlapping region to the total number of pores is used as the fracture-pore overlap ratio. For example, if 64 pore pixels are adjacent to the fracture path, accounting for 32% of the total number of pores, it is considered as a medium-level overlap region. Finally, the co-occurrence region is judged according to the joint result of the fracture continuity level and the pore aggregation level in each unit. If the fracture continuity is strong, the pore aggregation degree exceeds 60%, and the overlap ratio exceeds 30%, it is marked as a fracture-pore co-occurrence region, and the spatial co-occurrence coupling degree level of each unit is output.

[0047] S503: Based on the degree of spatial co-occurrence coupling, adjust the collapsibility classification parameters, determine whether the spatial structure of the monitoring unit has a continuous trend, classify and assign the spatial characteristics of the monitoring unit to obtain the spatial collapsibility classification type. The coupling level labels of each monitoring unit are read and classified into three categories: low, medium, and high, corresponding to different degrees of structural overlap. An initial collapsibility level classification reference value is set as the collapsibility risk coefficient within the range of 0–1. The adjustment value for low coupling is +0.1, for medium coupling +0.3, and for high coupling +0.5. The original collapsibility level value of each monitoring unit is updated by adding the corresponding adjustment value. Then, it is determined whether the fracture path direction of each unit is consistent with the pore aggregation direction. If the angle between the line connecting the fracture path and the pore center point is less than 30 degrees, they are considered to have directional continuity. The ratio of the number of units satisfying directional continuity to the total number of units is calculated. For example, if 38 out of 60 monitoring units meet this condition, the continuous trend ratio is 63.3%. Units with continuous directional trends are marked as trend-consistent units. Then, based on the adjusted subsidence level value, spatial structure extension continuity, and coupling strength, spatial characteristics are comprehensively classified. The classification rules are set as follows: if the subsidence level value is greater than 0.7 and has both high coupling and trend consistency markings, it belongs to the high-level subsidence type; if the subsidence value is between 0.4 and 0.7 and the coupling is medium-level, it belongs to the medium-level subsidence type; the rest are classified as low-level subsidence areas. All monitoring units are classified into the three categories of high, medium, and low spatial subsidence types according to their numbers, and the spatial subsidence classification type is output.

[0048] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0049] In this invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of a single item or a plurality of items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be a single item or multiple items.

[0050] It should be understood that, in various embodiments of the present invention, the order of the above-mentioned process numbers does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0051] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0052] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, apparatuses, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0053] In the several embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0054] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0055] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0056] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0057] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for evaluating and calculating the collapsibility of loess based on image semantic segmentation, characterized in that, The method includes: S1: Based on remote sensing image reflectance data, collect chromaticity and texture entropy data, analyze the reflectance fluctuation and chromaticity distribution differences of each grid cell, normalize the texture entropy, screen the key blocks with parameter variation amplitude, and calculate the spatial distribution characteristics to obtain the spatial heterogeneous distribution structure. S2: Based on the spatial heterogeneous distribution structure, determine the level changes of each monitoring block, analyze the spatial location of blocks with frequent fluctuations, shorten the sampling step size and increase the resolution for high-change blocks, and increase the step size and decrease the resolution for low-change blocks to obtain the optimal sampling configuration. S3: Based on the optimal sampling configuration, calculate the temperature change trend of each block, analyze the dynamics of humidity change, soil stress and water content, compare the morphology of the crack boundaries in the images during the abrupt change period, determine the regional expansion, and obtain the crack change response factor. S4: Based on the crack change response factor, analyze the crack path connectivity of each block, determine whether the crack connects the block boundary, calculate the crack length and boundary ratio, compare the proportion of pore pixels, integrate the judgment rules, and obtain the connected pore structure index.

2. The method for evaluating and calculating the collapsibility of loess based on image semantic segmentation according to claim 1, characterized in that, The spatial heterogeneous distribution structure includes a set of heterogeneous parameters, aggregated distribution information, and spatial structural features. The crack change response factor includes sensitivity parameters, extended criteria, and response type. The connected pore structure index includes structural classification, correlation characteristics, and index system.

3. The method for evaluating and calculating the collapsibility of loess based on image semantic segmentation according to claim 1, characterized in that, The reflectance fluctuation trend refers to the magnitude and direction of change of the remote sensing reflectance of pixels within the same monitoring grid over time or within a certain period or spatial range. The chromaticity distribution difference refers to the distribution changes, comparisons, or deviations of chromaticity parameters between different pixels or different monitoring grids in the same area, reflecting differences in surface characteristics.

4. The method for evaluating and calculating the collapsibility of loess based on image semantic segmentation according to claim 1, characterized in that, The texture entropy refers to the measurement of the gray-scale distribution pattern of pixels in a remote sensing image, and the sampling distribution refers to the spatial distribution pattern and distribution method of sampling points or sampling units within the evaluation area.

5. The method for evaluating and calculating the collapsibility of loess based on image semantic segmentation according to claim 1, characterized in that, The specific steps for obtaining the spatially heterogeneous distribution structure are as follows: S101: Based on remote sensing image reflectance data, the reflectance sequence of each pixel in the monitoring grid is differentially analyzed in a continuous observation period, the trend of reflectance change of each pixel over time is statistically analyzed, the change gradient of pixels in each region is compared, the spatial distribution location of the cluster of change trends in the region is determined, and the change gradient distribution is obtained. S102: Based on the change gradient distribution, compare the chroma channel distribution of each block of pixels, calculate the difference between the pixels in the block and the average chroma level of the block, sort by chroma offset, optimize and select representative sample blocks, and normalize and organize the block texture distribution to obtain a representative texture factor group. S103: Based on the representative texture factor group, calculate the reflectance change trend, chromaticity shift and normalized texture distribution of the corresponding block, jointly analyze the spatial distribution pattern of the three parameters, screen the monitoring blocks with spatial representativeness, determine the aggregation relationship between their spatial coordinates, and obtain the spatial heterogeneous distribution structure.

6. The method for evaluating and calculating the collapsibility of loess based on image semantic segmentation according to claim 1, characterized in that, The specific steps for obtaining the optimal sampling configuration are as follows: S201: Based on the spatial heterogeneous distribution structure, analyze the heterogeneous distribution characteristics of the monitoring blocks, statistically analyze the change data of each block in the continuous sampling period, compare the heterogeneous change between each block and adjacent blocks, and classify and sort the blocks according to the change trend and the degree of data dispersion to obtain spatial classification parameters. S202: Based on the spatial classification parameters, determine the spatial location of the blocks with active heterogeneous changes, extract the remote sensing sampling primitives of the associated blocks, optimize and adjust the sampling step size of the blocks, calculate the spatial density of the sampling point distribution before and after adjustment, organize the sampling parameters of each block after adjustment, and obtain the distribution density factor. S203: Based on the distribution density factor, screen monitoring blocks with small level fluctuations, extend the sampling step size of the corresponding blocks, analyze the continuity of remote sensing area coverage after adjusting the sampling interval, integrate the distribution of blocks with different sampling densities, judge the overall spatial coverage balance, and obtain the optimal sampling configuration.

7. The method for evaluating and calculating the collapsibility of loess based on image semantic segmentation according to claim 1, characterized in that, The specific steps for obtaining the crack change response factor are as follows: S301: Based on the optimal sampling configuration, calculate the temperature change of each block in a continuous time period, analyze the average temperature level and fluctuation trend in the sampling period, determine the direction of temperature increase or decrease in each period, organize the trend parameters, screen the key change stages, and obtain the temperature change coefficient group. S302: Based on the temperature change coefficient group, analyze the fluctuation structure of humidity data in each block within the cycle, identify the boundary points of phased rise or fall in the data, statistically analyze the humidity change characteristics within the segment, compare the correlation between the segment and the temperature trend, and obtain the segmented humidity characteristic set. S303: Based on the segmented humidity characteristic set, determine the pixel boundary changes of the crack region image before and after the mutation, compare the arrangement and spatial coverage of the crack pixel boundary structure, screen the key change groups of the region expansion, calculate the change amplitude of its region boundary, and obtain the crack change response factor.

8. The method for evaluating and calculating the collapsibility of loess based on image semantic segmentation according to claim 1, characterized in that, The specific steps for obtaining the interconnected pore structure index are as follows: S401: Based on the crack change response factor, analyze the spatial extension of the crack path within the monitoring block, determine whether the crack cell arrangement is continuously distributed, compare the contact situation of the crack path on the boundary, calculate the degree of penetration of the crack path within the block, filter the path that penetrates to the block boundary, and obtain the spatial path connectivity. S402: Based on the spatial path connectivity, compare the matching between the crack path length and the boundary length of each monitoring block, screen the blocks with key path extension in the boundary direction, analyze the number of pore pixels in the same block and their coverage area, and statistically calculate the proportion of pores in the crack path-covered block to obtain the structural overlap ratio coefficient. S403: Based on the structural overlap ratio coefficient, determine the spatial combination distribution of crack and pore structures, analyze the arrangement pattern of the intersecting areas of crack paths and pore boundaries, screen the densely intersecting spatial structures, statistically analyze the spatial distribution of structural arrangement, and obtain the connected pore structure index.

9. The method for evaluating and calculating the collapsibility of loess based on image semantic segmentation according to claim 1, characterized in that, The method further includes: S5: Based on the connected pore structure index, determine the crack penetration and pore aggregation performance in the monitoring unit, analyze the spatial overlap between the two, optimize the wet collapse level classification standard, identify the level type of each unit, and obtain the spatial wet collapse classification type. The spatial wetland classification types include grade identification, risk identification, and zoning classification.

10. The method for evaluating and calculating the collapsibility of loess based on image semantic segmentation according to claim 9, characterized in that, The specific steps for obtaining the spatial sinking classification type are as follows: S501: Based on the connected pore structure index, determine the spatial extension of the crack path within the monitoring unit, analyze the number and arrangement direction of the intersection points between the crack penetration trajectory and the boundary, count the coverage of the crack intersection points, and compare the distribution of the continuous penetration structure within the unit to obtain the spatial continuity parameter. S502: Based on the spatial continuity parameter, determine the aggregation of pore pixel distribution in each monitoring unit, count the arrangement density and center distance of pores in space, calculate the overlap ratio of cracks and pores in the image, identify the spatial distribution of crack penetration and pore aggregation and co-occurrence, and obtain the degree of spatial co-occurrence coupling. S503: Based on the spatial co-occurrence coupling degree, adjust the collapsibility classification parameters, determine whether the spatial structure of the monitoring unit has a continuous trend, classify and assign the spatial characteristics of the monitoring unit, and obtain the spatial collapsibility classification type.

Images