Method for optimizing visual image of surface crack gap

Abstract

The invention discloses a surface crack visual image optimization method, and the method comprises the steps: obtaining a plurality of surface images of a to-be-monitored region, and selecting a to-be-optimized region of the surface images; acquiring historical earth surface data of the to-be-monitored area; extracting historical earth surface data as standard earth surface data according to the coordinates of the to-be-optimized region; updating the to-be-optimized region by taking the standard earth surface data as a replacement region; and carrying out binarization coupling on the boundary of the updated to-be-optimized region through the ground fracture single-boundary repair model to obtain an optimized earth surface image. According to the method for optimizing the visual image of the surface crack, the modal features on the crack are removed, and the real crack image is obtained through optimization.

Description

Technical Field

[0001] The present invention belongs to the technical field of geological image processing methods, and in particular relates to a method for optimizing visual images of surface cracks. Background Art

[0002] Underground mining of coal and metal mines creates goafs, leading to stress imbalances in the overlying rock strata and the emergence of subsidence basins and fault zones on the surface. These cracks often appear in strips or grids, with widths ranging from tens of centimeters to several meters. For example, in major coal-producing regions like Shaanxi, Shanxi, and Inner Mongolia, the incidence of surface cracks in mining areas is significantly higher than in surrounding non-mining areas due to long-term, large-scale mining.

[0003] The existing Chinese patent, "A Crack Size Measurement Method Based on Image Processing" (Publication No. CN118583052A, Publication Date: September 3, 2024), includes the following steps: acquiring a crack image with target points set around the crack; performing binarization processing; obtaining a processed crack image; performing orthophoto and reduction transformations on the processed crack image; acquiring all contour lines in the crack image and ranking and counting them based on a user-set length threshold; measuring the maximum width and length of each crack region; and multiplying the crack position, length, and maximum width by a scaling parameter to obtain the actual crack size in the actual coordinate system. The goal is to calculate the actual crack size from the crack size acquired on the image. However, in the actual acquisition of surface crack images, various natural and human factors can prevent the acquired images from accurately capturing the actual crack scene. Summary of the Invention

[0004] The purpose of the present invention is to provide a method for optimizing the visual image of surface cracks, remove the modal features on the cracks, and optimize to obtain a real ground crack image.

[0005] The technical solution of the present invention is a method for optimizing the visual image of surface cracks, which is specifically implemented by the following steps: S1. Obtain multiple surface images of the area to be monitored and select the area to be optimized of the surface image; S2. Obtain historical surface data of the area to be monitored; S3. According to the coordinates of the area to be optimized, extract the historical surface data in S2 as standard surface data; update the standard surface data as a replacement area to be optimized; S4. Binarize and couple the updated boundary of the area to be optimized using the ground fissure single boundary repair model to obtain an optimized surface image.

[0006] The technical feature of the present invention is that: In S1, a surface image acquisition method is selected according to the type of the area to be monitored. The types of the areas to be monitored include mining areas, natural ecological areas, and special geomorphological areas. The acquisition method includes obtaining multiple surface images at a set height, and the acquisition heights of natural ecological areas, special geomorphological areas, and mining areas decrease in sequence.

[0007] In S1, the area to be optimized includes differentiated features appearing in multiple surface images at the same coordinates. The differentiated features are based on image differences caused by at least one of the influencing factors including light and shadow, angle, time, and human activities. The area with image differences is selected as the area to be optimized.

[0008] In S2, the historical surface data includes the type of intervention event, the size of the intervention event, and the coverage of the intervention event; Intervention event types include man-made features and natural features. The natural features include landforms and natural trajectories. Man-made features include industrial products and industrial trajectories. Industrial products include cement buildings, metal buildings, and oil and gas pipelines. Industrial trajectories include mobile features that involve human participation or mechanization.

[0009] In S2, both human-feature intervention data and natural-feature intervention data have data overlap parameters of the same coordinates. The data overlap parameters are stored in historical time order. According to the scope of human intervention and the scope of natural intervention, the coverage of the intervention event is obtained. The coverage is proportional to the size of the intervention event. The size of the intervention event includes the total number of natural and human-feature features on the image, and the coverage is the total area of ​​natural and human-feature features on the image.

[0010] Where, is the number of artifacts in the intervention event, is the total number of features in the intervention event, is the area of ​​man-made features in the coverage area, is the area of ​​natural features in the coverage area, is the characteristic complexity rate in the data coincidence parameter.

[0011] When the area to be monitored is a mining area, the human feature data includes mining infrastructure and construction progress data. The features are stored through the uploaded infrastructure dimensions and the construction effect dimensions in the construction plan. The features of the same size are marked on the surface image and deleted, and the crack image without human occlusion features is retained.

[0012] In S3, the standard surface data includes the type, size and coverage area of ​​the intervention event. The regional image is reconstructed according to the historical time sequence of the historical surface data, and the reconstructed image is fact-calibrated to obtain the replacement area.

[0013] In S3, the replacement area content is set with human features and natural features, and the replacement area size is set with area, vegetation, depth and boundary shape.

[0014] In S4, the ground fissure single boundary repair model includes the following steps: S41: setting target points at the boundaries of the ground fissures in the updated area to be optimized, and performing binarization processing on the ground fissure boundary image to obtain a binary ground fissure boundary image; S42: removing areas where noise points in the binary crack boundary image are lower than a set standard threshold, and filling the removed areas by multiple boundary morphological dilations; S43: Connect the area within the boundary range with the filling area.

[0015] The beneficial effects of the present invention are: The surface crack visual image optimization method of the present invention is suitable for identifying and judging surface cracks after mining area development. By obtaining construction information of the mining area, the environment of the area before excavation is simulated, and then the appearance of the artificial area after excavation is replaced, thereby restoring the fact whether there are cracks in the area. For areas with blurred surface cracks, standard crack images are used to replace them, optimize the replacement boundaries, increase the realism, and improve the simulation effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 is a flow chart of the surface crack visual image optimization method of the present invention; Figure 2 It is a schematic diagram of a single boundary repair model of ground fissures in the surface fissure visual image optimization method of the present invention. DETAILED DESCRIPTION

[0017] To more clearly illustrate the technical solutions of the embodiments of this specification, the following briefly describes the drawings required for describing the embodiments. Obviously, the drawings described below are merely examples or embodiments of this specification. Those skilled in the art can apply this specification to other similar scenarios based on these drawings without inventive effort. Unless otherwise apparent from the context or otherwise noted, the same reference numerals in the figures represent the same structure or operation.

[0018] Example 1 The surface crack visual image optimization method of the present invention has the following process: Figure 1 As shown, the simulation is performed on the ArcGis pro platform. The specific steps are as follows: S1. Obtain multiple surface images of the area to be monitored and select the area to be optimized of the surface image; S2. Obtain historical surface data of the area to be monitored; S3. According to the coordinates of the area to be optimized, extract the historical surface data in S2 as standard surface data; update the standard surface data as a replacement area to be optimized; S4. Binarize and couple the updated boundary of the area to be optimized using the ground fissure single boundary repair model to obtain an optimized surface image.

[0019] The binarization coupling in S4 is achieved by deriving each team's parameter by taking half of its own value and then extending it to the next measurement. The measurement parameter here is distance and the output is image pixels. The binarization here is the application of a conventional simulation algorithm used by those skilled in the art and will not be elaborated here.

[0020] Example 2 On the basis of Example 1, in S1, a method for collecting surface images is selected according to the type of area to be monitored. The types of areas to be monitored include mining areas, natural ecological areas, and special geomorphological areas. The collection method includes obtaining multiple surface images by shooting with a drone at a set height. The collection heights of natural ecological areas, special geomorphological areas, and mining areas decrease in sequence.

[0021] The area to be optimized includes the differentiated features that appear in multiple surface images at the same coordinate position. The differentiated features are based on the differences in visual images caused by at least one of multiple influencing factors, including light and shadow, angle, time, and human activities. The area with different visual images is selected as the area to be optimized. As long as there are differences between the two images, they are differentiated features.

[0022] In order to avoid repeated calculation of surface data at the same location, when overlapping images appear, only differential feature extraction is performed. For example, if the grayscale values ​​of the two images at coordinates a and b are different, then the images at coordinates a and b in the two images are extracted as the areas to be optimized. In actual applications, the difference in grayscale values ​​may be due to changes in the light and shadow when the two images were taken, or the angles of the drones used for shooting are different, etc. However, in order to avoid missed judgments and misjudgments, differential feature extraction is still performed.

[0023] Example 3 Based on Example 2, further, in S2, the historical surface data includes the type of intervention event, the size of the intervention event, and the coverage of the intervention event; Intervention event types include both human-made features and natural features. Human-made features include industrial artifacts and industrial trajectories, while natural features include landforms such as rivers, gullies, and sand ridges, as well as natural trajectories. Human-made features include both industrial artifacts and industrial trajectories. Industrial artifacts include concrete buildings, metal structures, and oil and gas pipelines. Industrial trajectories include mobile features that involve human intervention or mechanical intelligence, such as cars and trains.

[0024] Both human-feature intervention data and natural-feature intervention data have data coincidence parameters of the same coordinates. The data coincidence parameters are stored in historical time order. According to the scope of human intervention and the scope of natural intervention, the coverage of the intervention event is obtained. The area of ​​the coverage is proportional to the degree of the intervention event. The size of the intervention event includes the total number of natural and human-feature features on the image, and the coverage is the total area of ​​natural and human-feature features on the image.

[0025] Where, is the number of artifacts in the intervention event, is the total number of features in the intervention event, is the area of ​​man-made features in the coverage area, is the area of ​​natural features in the coverage area, is the characteristic complexity rate in the data coincidence parameter.

[0026] Example 4 Building on Example 3, when the monitored area is a mining area, the human-induced feature data includes mining infrastructure and construction progress data. Feature storage is performed using the uploaded infrastructure dimensions and the construction results dimensions in the construction plan. By marking features of equivalent size on the surface image and deleting them, images of cracks without human-induced obstructions are retained. For example, if an excavator is blocking the ground, the excavator can be selected and deleted, retaining the ground crack image without the excavator. This is particularly suitable for capturing ground crack images in mining areas.

[0027] Example 5 On the basis of Example 4, further, in S3, the standard surface data includes the type, size and coverage area of ​​the intervention event, and the regional image is reconstructed according to the historical time sequence of the occurrence of the historical surface data, and the reconstructed image is fact-calibrated to obtain the replacement area.

[0028] Example 6 On the basis of Example 5, further, in S4, the replacement region content is set with human features and natural features, and the replacement region size is set with area, vegetation, depth and boundary shape. Example 7 On the basis of Example 6, further, as Figure 2 As shown in S4, the single boundary repair model of ground fissures includes the following steps: S41: setting target points at the boundaries of the ground fissures in the updated area to be optimized, and performing binarization processing on the ground fissure boundary image to obtain a binary ground fissure boundary image; S42: removing areas where noise points in the binary crack boundary image are lower than a set standard threshold, and filling the removed areas by multiple boundary morphological dilations; S43: Connect the area within the boundary range with the filling area.

[0029] For areas with blurred surface cracks, standard crack images are used to replace them, optimize the replacement boundaries, increase the realism, and improve the simulation effect.

[0030] To prevent the impact of construction and facilities like ditch excavation and pipeline laying within mining areas on crack identification, the proposed surface crack visual image optimization method builds mining area construction information on the ArcGis Pro platform, simulates the area's pre-excavation environment, and then replaces the post-excavation appearance of the artificially created area to restore the presence of cracks. This method is suitable for identifying surface cracks after mining development.

[0031] Example 8: Based on the above embodiments, the present application also includes transfer learning and domain adaptation pre-training models to learn common features on large-scale public data sets (such as Google Earth crack samples), and then fine-tune (Fine-Tuning) for specific mining area data.

[0032] If data in the target mining area is scarce, domain adaptation technology (such as adversarial domain adaptation (ADA)) can be used to reduce the data distribution difference between the source domain and the target domain.

[0033] Table 1 Differences between ground cracks in mining areas and cracks in natural mountainous areas

[0034] The activities in the mining area in Table 1 above are also the industrial trajectories involved in this embodiment, and the cracks in the natural mountainous area are natural trajectories. Due to the different types of the two, when simulating the two areas, there are also different threshold settings for the application of the single-boundary repair model for the ground cracks in the area.

[0035] It is worth noting that drones equipped with high-definition cameras (such as 20 million pixels or above) or multispectral cameras (such as DJI Matrice 350RTK, Parrot Anafi, etc.) can quickly cover large mining areas, obtain high-resolution (centimeter-level) images, flexibly adapt to complex terrain (such as steep slopes and valleys), avoid the risks of manual inspections, support scheduled inspections (such as daily / weekly), record dynamic changes in cracks, plan flight routes (such as grid-like, route coverage), ensure image overlap rate ≥70%, and usually fly at an altitude of 50~200 meters, which is adjusted according to the scope of the mining area and accuracy requirements. GPS positioning is turned on during shooting to provide geographic coordinates for subsequent image stitching.

[0036] The surface fissure visual image optimization method of the present invention obtains a complete and clear ground fissure image by processing remote sensing images. It is particularly suitable for surface monitoring in mining areas. By synchronizing the construction data of the mining area to eliminate features, it avoids characteristic objects and trajectories blocking the cracks, resulting in the image being unable to capture the full picture of the cracks, thereby achieving purification and preservation of the ground fissures.

Claims

1. A surface crack visual image optimization method, characterized in that: Please follow the steps below to implement: S1. Obtain multiple surface images of the area to be monitored and select the area to be optimized of the surface image; S2. Obtain historical surface data of the area to be monitored; S3. According to the coordinates of the area to be optimized, extract the historical surface data in S2 as standard surface data; update the standard surface data as a replacement area to be optimized; S4. Binarize and couple the updated boundary of the area to be optimized using the ground fissure single boundary repair model to obtain an optimized surface image.

2. The surface crack visual image optimization method according to claim 1, characterized in that: In S1, a method for collecting surface images is selected according to the type of the area to be monitored. The types of the area to be monitored include mining areas, natural ecological areas, and special landform areas. The collection method includes obtaining multiple surface images at a set height, and the collection heights of natural ecological areas, special landform areas, and mining areas decrease in sequence.

3. The surface crack visual image optimization method according to claim 1, characterized in that: In S1, the area to be optimized includes differentiated features appearing in multiple surface images at the same coordinates. The differentiated features are based on image differences caused by at least one of the influencing factors including light and shadow, angle, time, and human activities. The area with image differences is selected as the area to be optimized.

4. The surface crack visual image optimization method according to claim 1, characterized in that: In S2, the historical surface data includes the type of intervention event, the size of the intervention event, and the coverage of the intervention event; The intervention event types include human-made features and natural features. The natural features include landforms and natural trajectories. Human-made features include industrial products and industrial trajectories. Industrial products include cement buildings, metal buildings, and oil and gas pipelines. Industrial trajectories include mobile features that involve human participation or mechanization.

5. The surface crack visual image optimization method according to claim 4, characterized in that: In S2, both the human feature intervention data and the natural feature intervention data have data overlap parameters of the same coordinates. The data overlap parameters are stored in historical time order. The coverage of the intervention event is obtained based on the scope of human intervention and the scope of natural intervention. The coverage is proportional to the size of the intervention event. The size of the intervention event includes the total number of natural features and human features on the image, and the coverage is the total area of ​​natural features and human features on the image. Where, is the number of artifacts in the intervention event, is the total number of features in the intervention event, is the area of ​​man-made features in the coverage area, is the area of ​​natural features in the coverage area, is the characteristic complexity rate in the data coincidence parameter.

6. The surface crack visual image optimization method according to claim 5, characterized in that: When the area to be monitored is a mining area, the human feature data includes mining infrastructure and construction progress data. The features are stored through the uploaded infrastructure dimensions and the construction effect dimensions in the construction plan. The features of the same size are marked on the surface image and deleted, and the crack image without human occlusion features is retained.

7. The surface crack visual image optimization method according to claim 5, characterized in that: In S3, the standard surface data includes the type, size and coverage area of ​​the intervention event. The regional image is reconstructed according to the historical time sequence of the historical surface data, and the reconstructed image is fact-calibrated to obtain the replacement area.

8. The surface crack visual image optimization method according to claim 7, characterized in that: In S3, the replacement area content is set with human features and natural features, and the replacement area size is set with area, vegetation, depth and boundary shape.

9. The surface crack visual image optimization method according to claim 1, characterized in that: In S4, the ground fissure single boundary repair model includes the following steps: S41: setting target points at the boundaries of the ground fissures in the updated area to be optimized, and performing binarization processing on the ground fissure boundary image to obtain a binary ground fissure boundary image; S42: removing areas where noise points in the binary crack boundary image are lower than a set standard threshold, and filling the removed areas by multiple boundary morphological dilations; S43: Connect the area within the boundary range with the filling area.

Images