Method, device and medium for adaptive segmentation of blast pile images based on rock shape

Abstract

The present invention discloses a method, device and medium for adaptive segmentation of a blast pile image based on the shape of rock blocks. The method comprises: step 1, preprocessing the blast pile image to obtain a binary image of the blast pile; step 2, performing distance transformation on the binary image of the blast pile to obtain a distance transformation image of the blast pile image; step 3, using a seed point marking method based on the solidity of the rock block contour, adaptively marking seed points according to the distance transformation image, and obtaining a seed point map of the blast pile image; step 4, using a watershed algorithm, and performing image segmentation on the rock blocks in the blast pile image according to the seed point map. The present invention can reduce the phenomenon of over-segmentation and under-segmentation of rock blocks in the prior art, and realize the effective segmentation of adhered rock blocks in the blast pile.

Description

Technical Field

[0001] The present invention belongs to the technical field of machine vision applications, and in particular relates to a method, device and medium for adaptively segmenting a blast pile image based on rock block shapes. Background Art

[0002] Blasting is widely used in mining and civil engineering due to its economic and high efficiency. The distribution of blasting fragmentation is an important technical indicator for measuring the blasting effect. It directly affects the cost and efficiency of subsequent shoveling, crushing and grinding processes, and also provides a necessary basis for optimizing blasting parameters. Therefore, establishing a fast and accurate detection method for blasting fragmentation has important theoretical significance and practical value for guiding blasting construction and improving blasting efficiency.

[0003] The blast pile has the characteristics of large scale, serious adhesion of rock blocks and irregular shapes, large particle size differences, and small grayscale differences. The watershed algorithm is widely used in rock particle segmentation because of its good response to weak edges. However, due to the presence of many meaningless local minima, that is, redundant seed points, in the blast pile image, the watershed algorithm produces serious mis-segmentation, especially over-segmentation. Summary of the invention

[0004] The present invention provides a method, device and medium for adaptive segmentation of blast pile images based on rock block shapes, so as to solve the technical problem in the prior art that redundant seed points lead to erroneous segmentation of a watershed algorithm.

[0005] In order to achieve the above technical objectives, the present invention adopts the following technical solutions:

[0006] A method for adaptive segmentation of blast pile images based on rock block shapes, comprising:

[0007] Step 1, preprocessing the burst pile image to obtain a burst pile binary image;

[0008] Step 2, performing distance transformation on the burst binary image to obtain a distance transformed image of the burst image;

[0009] Step 3, using a seed point marking method based on the solidity of the rock block contour, adaptively marking the seed points according to the distance transformation image, and obtaining a seed point map of the blast pile image;

[0010] Step 4: Use the watershed algorithm to segment the rock blocks in the blast pile image according to the seed point map.

[0011] Furthermore, the specific method of obtaining the seed point map in step 3 is:

[0012] Step 3.1, create a new canvas according to the size of the distance transformation image and set the initial value of the grayscale threshold;

[0013] Step 3.2, performing grayscale threshold processing on the distance transformation image according to the current grayscale threshold to obtain a grayscale threshold cross-sectional image;

[0014] Step 3.3, using contour detection technology to perform contour detection on the gray threshold cross-sectional image, and obtaining the solidity of each contour in each gray threshold cross-sectional image;

[0015] Step 3.4, determine whether the solidity of each contour is greater than a given solidity threshold. If so, draw the contour into the newly created canvas and fill it as a seed point;

[0016] Step 3.5, increase the grayscale threshold according to the preset method, and determine whether the increased grayscale threshold is greater than 255. If so, complete the drawing of the seed point map, otherwise return to step 3.2.

[0017] Furthermore, after determining in step 3.4 that the solidity of a certain contour is greater than a given solidity threshold, it is further determined whether there are background points in the contour. Only when there are no background points in the contour or the contour is irrelevant to the background points, the contour is drawn into the newly created canvas and filled as a seed point.

[0018] Furthermore, the solidity of the rock block contour refers to the ratio of the area of ​​the rock block to the area of ​​the convex hull of the rock block contour.

[0019] Furthermore, the given solidity threshold is greater than 0.82.

[0020] Furthermore, the preprocessing of step 1 includes in sequence: image grayscale, bilateral filtering denoising, contrast-limited adaptive histogram equalization, image binarization, morphological optimization and area filtering.

[0021] Furthermore, the Phansalkar method based on local image properties is used to binarize the explosion pile image.

[0022] An electronic device comprises a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor implements any one of the above-mentioned methods for adaptively segmenting blast pile images based on rock block shapes.

[0023] A computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, the method for adaptively segmenting a blast pile image based on rock shape described in any one of the above is implemented.

[0024] Beneficial Effects

[0025] The present invention first pre-processes and distance transforms the blast pile image to obtain a distance transform image; then, a seed point marking method based on the rock block contour solidity is used to adaptively mark seed points according to the distance transform image, that is, an appropriate grayscale threshold is selected to separate the adhesion part in the distance transform image, and the seed point is determined by calculating the solidity, thereby obtaining a seed point map of the blast pile image; finally, a watershed algorithm is used to perform image segmentation on the rock blocks in the blast pile image according to the seed point map. Therefore, the present invention can reduce the phenomenon of over-segmentation and under-segmentation of rock blocks in the prior art, and realize the effective segmentation of adhesion rock blocks in the blast pile. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 is a flowchart of image preprocessing according to an embodiment of the present invention;

[0027] Figure 2 : This is the effect of the preprocessing of the blast pile image by the embodiment of the present invention, where a is the grayscale image of the limestone blast pile, b is the bilateral filtering denoising, c is the histogram equalization, d is the binary image, e is the morphological optimization, and f is the area filtering;

[0028] Figure 3 It is a two-dimensional topographic diagram of the watershed;

[0029] Figure 4 are the images before and after the distance transformation, where a is the burst binary image and b is the distance transformation image;

[0030] Figure 5 It is the distance transformation detail map and seed point map of the rock block;

[0031] Figure 6 is the solidity distribution histogram;

[0032] Figure 7 is a three-dimensional schematic diagram of a distance image and a two-dimensional cross-sectional diagram under different grayscale thresholds, wherein a is a three-dimensional schematic diagram of a distance image, b1 is a three-dimensional schematic diagram of a distance image with a grayscale threshold of 50, b2 is a cross-sectional diagram with a grayscale threshold of 50, c1 is a three-dimensional schematic diagram of a distance image with a grayscale threshold of 100, and c2 is a cross-sectional diagram with a grayscale threshold of 100;

[0033] Figure 8 The concave structure formed in the background area and the two-dimensional cross-sectional contour diagram, where a is the concave structure formed in the background area, and b is the contour of the background area;

[0034] Fig. 9 is a granite burst pile image and a binary image after preprocessing, where a is a granite burst pile image and b is a granite burst pile binary image;

[0035] Fig.10The seed point marking effect and segmentation effect diagram are shown in Figure 1, where a is the overlay of the limestone seed point diagram and the original diagram, b is the limestone blast pile segmentation effect diagram, c is the overlay of the granite seed point diagram and the original diagram, and d is the granite blast pile segmentation effect diagram;

[0036] Fig.11 are artificial segmentation maps, where a is the artificial segmentation map of limestone and b is the artificial segmentation map of granite;

[0037] Fig.12 is the cumulative distribution curve of limestone area;

[0038] Fig.13 This is the cumulative distribution curve of granite area. DETAILED DESCRIPTION

[0039] The following is a detailed description of an embodiment of the present invention. This embodiment is based on the technical solution of the present invention, and provides a detailed implementation method and a specific operation process to further explain the technical solution of the present invention.

[0040] Example 1

[0041] The present invention provides an adaptive segmentation method for a blast pile image based on the shape of rock blocks. First, the blast pile image is preprocessed to obtain a binary image of the blast pile; then the binary image of the blast pile is distance transformed to obtain a distance transformed image of the blast pile image; then a seed point marking method based on the solidity of the rock block contour is adopted to adaptively mark the seed points according to the distance transformed image to obtain a seed point map of the blast pile image; finally, a watershed algorithm is adopted to segment the rock blocks in the blast pile image according to the seed point map. Based on the existing distance transformation optimized watershed algorithm for image segmentation, the present invention takes into account the characteristics of the blast pile, such as large scale, serious adhesion of rock blocks, irregular shape, large particle size difference, small grayscale difference, etc., and uses an adaptive grayscale threshold to mark the seed points based on the solidity of the rock block contour, forming a watershed algorithm based on the solidity of the rock block contour, reducing over-segmentation and under-segmentation phenomena, and realizing effective segmentation of adhered rock blocks in the blast pile.

[0042] Specifically, the method for adaptively segmenting a blast pile image based on rock shape provided in this embodiment includes the following steps:

[0043] Step 1: preprocess the burst pile image to obtain a burst pile binary image.

[0044] Due to the complex environment in which the blast pile images are collected, the blast pile images collected by the machine vision system have serious noise, and the rock blocks are seriously piled up and adhere to each other, the background difference is small, and the color information is not obvious. In order to effectively segment the blast pile rocks, the blast pile images need to be preprocessed. The preprocessing process is as follows: Figure 1As shown, the process includes: image grayscale, bilateral filtering denoising, contrast-limited adaptive histogram equalization, image binarization, morphological optimization and area filtering.

[0045] In this embodiment, bilateral filtering is first used to denoise the grayscale image of the blast pile, and then the edge of the rock block is made more obvious by limiting the contrast of adaptive histogram equalization. Then, the blast pile image is binarized using the Phansalkar method based on local image properties. Finally, the binary image of the blast pile is morphologically optimized and area filtering is performed. Figure 2 is the effect diagram after preprocessing, where Figure 2 (a) is the grayscale image of the explosion pile; Figure 2 (b) is the image after bilateral filtering and denoising; Figure 2 (c) is the effect diagram after histogram equalization; Figure 2 (d) is the binary effect diagram of the blast pile image. It can be seen from the figure that the Phansalkar method can accurately distinguish the background from the ore particles, and the effect is relatively ideal; Figure 2 (e) is the binary image of the explosion pile after morphological optimization. Figure 2 (d) By comparison, it can be found that morphological optimization can eliminate smaller "holes" and noise points in binary images and smooth the edges of targets, but there are still some large "hole" areas and noise points that cannot be removed. Figure 2 (f) is the binary image of the explosion pile after area filtering is used to eliminate large "holes" and noise points, achieving a relatively ideal result.

[0046] Step 2: Perform distance transformation on the burst binary image to obtain a distance transformed image of the burst image.

[0047] The principle of the watershed algorithm is to visualize an image as a topographic map in three dimensions. In this "topographic" interpretation, three types of points are considered, such as Figure 3 As shown in the figure: ① local minimum point or minimum surface, which corresponds to the lowest point of a basin; ② other position points in the basin; ③ edge points of the basin, that is, the intersection point between the basin and other basins. For a specific regional minimum, the set of points that meet condition ② is called the catchment basin or watershed of the minimum. The points that meet condition ③ form the front line of the ground, which is called the dividing line or watershed. The main goal of the segmentation algorithm based on these concepts is to find the watershed, that is, the edge of the rock in the image.

[0048] The traditional watershed algorithm is usually a markerless segmentation, and the input object is a gradient image, which is based on brightness changes. It only reflects the edge information of the image, and will have unreasonable segmentation due to its sensitivity to noise, resulting in the following disadvantages of the watershed algorithm: ① The noise in the original image causes the segmentation contour to shift; ② For images with low contrast, the contour of the target area is easily lost during watershed segmentation; ③ Because there are many meaningless local minima in the image, the watershed produces over-segmentation. Therefore, the most commonly used method in the prior art is to perform distance transformation on the binary image.

[0049] Distance transform refers to the distance from each pixel in a binary image to its nearest zero-value pixel. Assuming a binary image with a target set and a background set, and a distance map of , the definition of distance transform is as follows:

[0050] D(p)=min(disf(p, q))p∈O, q∈B (1)

[0051] The distance function is generally selected as Euclidean distance.

[0052] Step 3: Adopt a seed point marking method based on the rock block contour solidity, perform seed point adaptive marking according to the distance transformation image, and obtain a seed point map of the blast pile image.

[0053] Figure 4 The images after distance transformation of the burst binary image are given, where (a) is the binary image of the burst, and (b) is the distance image of the binary image. It can be seen from (b) that the distance image of the binary image is similar to the skeleton of the image and still retains the general shape of the image.

[0054] Figure 5 The distance transformation detail image and seed point image of a rock block in the explosion pile image, where (a) is the grayscale image of a rock block in the explosion pile image; (b) is the distance transformation image of the rock block; (c) is the local maximum point of the distance image, i.e., the seed point. Figure 5 It can be seen in (c) that there are multiple maximum points in the center of the rock block (in the rectangular box), that is, redundant seed points, which will cause serious over-segmentation. In addition, there are also many extreme points in the adhesion part of the two rock blocks (in the elliptical box), which seriously affects the segmentation effect of the adhesion rock blocks.

[0055] For the redundant seed point problem, the commonly used methods are mainly to merge adjacent maximum points and expand the maximum points. Figure 5 As shown in (c), the distance between the maximum point inside the rock block and the maximum point in the adhesion part of the rock block is large, and it is difficult to merge them accurately by the above method. Moreover, for complex ore images, the distribution of maximum points is very irregular, and the above method of eliminating maximum points is less effective.

[0056] In the segmentation process of some special adhesion objects, such as adhesion cells, spherical particles, etc., some prior knowledge, such as the color of the cell nucleus, the shape of the spherical particles, etc., can be used in turn to correct the seed points. However, when the segmentation object of the present invention is the explosive pile rock block, there is no other prior knowledge available except the shape, and the shape of the explosive pile rock block is irregular, and most of them are irregular polygons. In order to explore the shape law of the explosive pile rock block, the present invention manually segments the explosive pile image, and the segmentation effect is shown in FIG. Fig.11 (a).

[0057] Through image contour detection technology, the information of rock blocks in the blast pile is counted, including the area of ​​the rock block, the convex hull of the contour, and the solidity of the rock block contour. It should be noted that the solidity of the rock block contour in the present invention is the ratio of the area of ​​the rock block to the area of ​​the convex hull of the rock block contour. The information of some larger rock blocks is shown in Table 1. Figure 6 is the solidity distribution histogram.

[0058] Table 1 Rock information of blast pile image

[0059]

[0060] from Figure 6 It can be seen that the contour solidity of the larger rock blocks in the figure ranges from 0.82 to 0.97, mainly concentrated between 0.87 and 0.95. Therefore, the contour solidity of the explosive pile rock blocks is relatively high. It can be concluded that the shape of the explosive pile rock blocks is close to or equal to a convex polyhedron.

[0061] Based on the above research, in order to solve the technical problem of redundant seed points, the present invention proposes a seed point marking method based on the rock block contour solidity, which is used to adaptively mark seed points according to the distance transformation image to obtain the seed point map of the blast pile image. This method selects an appropriate grayscale threshold to separate the adhesion part (i.e., the adhesion rock block in the blast pile image) in the distance transformation image, and determines the seed point by calculating the contour solidity through contour detection. Figure 7 As shown, (a) is a three-dimensional schematic diagram of the distance image, where the height is the grayscale value of the corresponding point; (b1) is a three-dimensional schematic diagram of the distance image when the grayscale threshold is 50, and (b2) is its corresponding two-dimensional cross-sectional view; (c1) is a three-dimensional schematic diagram of the distance image when the grayscale threshold is 100, and (c2) is its corresponding two-dimensional cross-sectional view.

[0062] from Figure 7 (b1) and (b2) show that when the grayscale threshold is too small, although the smaller rocks can be separated and seed points can be obtained, many large rocks are still stuck together, and most of them do not meet the solidity conditions and cannot be used as seed points. Figure 7(c1) and (c2) show that when the grayscale threshold is too large, although the large rock blocks can be separated, the small rock blocks will be ignored. Due to the large size differences of the rock blocks in the explosive pile, a single grayscale threshold cannot meet the requirements of the seed point marking of the rock blocks in the explosive pile. Therefore, multiple grayscale thresholds are required for processing, and the contours in the two-dimensional cross-section image processed by each grayscale threshold are judged. If they meet the seed point conditions, the inside of the contour is filled and marked as a seed point.

[0063] Based on the above analysis, the steps of the seed point marking method based on the rock block contour solidity in step 3 of this embodiment are as follows:

[0064] Step 3.1, create a new canvas according to the size of the distance transformation image, and set the initial value of the grayscale threshold, such as setting the initial grayscale threshold to 10;

[0065] Step 3.2, perform grayscale threshold processing on the distance transformation image according to the current grayscale threshold to obtain a grayscale threshold cross-sectional image; Figure 7 (b2), (c2) 2D cross-section images;

[0066] Step 3.3, using contour detection technology to perform contour detection on the gray threshold cross-sectional image, and obtaining the solidity of each contour in each gray threshold cross-sectional image;

[0067] Step 3.4, determine whether the solidity of each contour is greater than a given solidity threshold. If so, draw the contour into the newly created canvas and fill it as a seed point;

[0068] Since the seed point marking method is based on the distance transformation image, when the contour solidity of a background area or noise point in the binary image is too high, the background area will be formed after the distance transformation processing. Figure 8 The concave structure in (a) will form the following when the grayscale threshold is used to extract the contour: Figure 8 (b) Therefore, in a more preferred embodiment, after determining in step 3.4 that the solidity of a certain contour is greater than a given solidity threshold, it is further determined whether there is a background point in the contour (first determine the position of the contour in the binary image, and then determine whether there is a background pixel value of 0 inside the contour). Only when there is no background point in the contour (there is no background pixel value of 0 inside the contour) or the contour is irrelevant to the background point (determine how many background points there are in the contour, case 1: if the number is greater than 1, then the contour is irrelevant to the background point, case 2: if the number is equal to 1, then determine whether the centroid of the contour is inside the background point area, if not, then it is irrelevant, if so, then it is relevant), the contour is drawn into the newly created canvas and filled as a seed point.

[0069] Step 3.5, increase the grayscale threshold according to a preset method, and determine whether the increased grayscale threshold is greater than 255. If so, the drawing of the seed point map is completed, otherwise, return to step 3.2. The grayscale threshold can be increased by equal step size or by other functions, and the present invention does not make specific restrictions.

[0070] Solidity is related to rock type, so the solidity threshold is given different values ​​according to the rock type. Figure 6 It can be seen that the contour solidity of the larger rock blocks in this figure ranges from 0.82 to 0.97.

[0071] Step 4, using the watershed algorithm, the rock blocks in the blast pile image are segmented according to the seed point map. Specifically, the image obtained by the grayscale image of the blast pile is equalized by the restricted contrast histogram according to the seed point map ( Figure 2 (c) Segmentation. Since the use of the watershed algorithm to segment an image is a prior art, the present invention will not elaborate on it in detail.

[0072] The following experiments are conducted to demonstrate the effect of the method of the present invention on the segmentation of rock blocks in the blast pile image.

[0073] 1. Seed point marking and segmentation results

[0074] In order to explore the segmentation effect of the explosive pile image adaptive watershed method based on the rock block contour solidity of the present invention on the explosive pile rock block, this experiment collected the explosive pile images of two different rocks, limestone and granite, and applied the segmentation method of the present invention to segment them. Figure 2 As shown in (a), the binary image is Figure 4 As shown in (a), the granite explosion pile image is as follows Fig. 9 As shown in (a), the binary image is Fig. 9 (b) as shown.

[0075] The seed point marking method based on the rock block contour solidity is used to mark the seed points of the binary images of the two rock explosion piles, where the solidity threshold is 0.85, and the watershed algorithm is used to segment the seed point map. The seed point marking effect and segmentation effect are shown in Fig.10 , where (a) is the overlay of limestone seed point map and grayscale map, (b) is the limestone blast pile segmentation effect map, (c) is the overlay of granite seed point map and grayscale map, and (d) is the granite blast pile segmentation effect map.

[0076] 2. Analysis of seed point marking results

[0077] from Fig.10 (a) and (c) show that the seed point marking method based on the rock block contour solidity can effectively mark the rock blocks, especially the larger ones. Fig.10(a) and (c) are compared with the pre-processed binary images of the blast pile, and it is found that the seed point marking method of the present invention can avoid the influence of the internal noise of the binary image rock block on the segmentation. Fig. 9 (b) The black box.

[0078] 3. Analysis of explosion pile image segmentation results

[0079] from Fig.10 As can be seen from (b) and (d), the adaptive watershed method for blast pile images based on the rock block contour solidity of the present invention can achieve relatively accurate segmentation of blast pile images with severe stacking and adhesion, especially limestone blast pile images with less noise. However, due to the influence of rock block shadows and other problems, the segmentation effect of granite blast pile images is poorer than that of limestone blast pile images. In order to evaluate the segmentation results with quantitative indicators, this experiment uses the artificial segmentation map as the segmentation standard to evaluate the segmentation effect of the two blast pile images, where the artificial segmentation map of the limestone blast pile is shown in Fig.11 (a), artificial segmentation diagram of granite blast pile Fig.11 (b).

[0080] (1) Analysis of the overall segmentation effect of the explosion pile image

[0081] Fig.12 and 13 They are the cumulative distribution curves of rock blocks in the limestone and granite explosion pile images, respectively. The calculation formula is shown in formula (2).

[0082]

[0083] Where, P is the cumulative area ratio of rock blocks; S total is the sum of the areas of all identified rock blocks in the entire blast pile image; S x It is the cumulative area of ​​the rock blocks at the area classification.

[0084] from Fig.12 and 13 It can be seen that the area cumulative distribution curves of the segmentation method of the present invention and the manual segmentation results are highly similar, especially for limestone, indicating that the method of the present invention can accurately segment rock blocks in the blast pile image.

[0085] (2) Analysis of the segmentation effect of a single rock block in the blast pile image

[0086] In order to evaluate the segmentation effect of the adaptive watershed algorithm of the blast pile image based on the solidity of the rock block contour on a single rock block, 15 rock blocks with an area of ​​more than 2000 and severe adhesion or shadows on the surface were extracted from the two blast pile images. The position distribution and corresponding area of ​​the extracted limestone blocks are shown in Figure 2. Fig.11 (a) and Table 2, the location distribution and corresponding area of ​​granite blocks are shown in Fig.11(b) and Table 3. From Table 2, it can be seen that the maximum relative error of the 15 limestone blocks extracted is 4.26% of No. 4, while the minimum relative error is only 0.27%, and the average relative error is 1.85%. From Table 3, it can be seen that the maximum relative error of the 15 granite blocks extracted is 13.30% of No. 3, while the minimum relative error is only 0.91%, and the average relative error is 5.62%.

[0087] Table 2 Area of ​​limestone blocks

[0088]

[0089] Table 3 Granite block area

[0090]

[0091] The above embodiments are preferred embodiments of the present application. Ordinary technicians in this field can also make various changes or improvements on this basis. Without departing from the overall concept of the present application, these changes or improvements should fall within the scope of protection required by the present application.

Claims

1. A method for adaptive segmentation of blast pile images based on rock shape, characterized in that: include: Step 1, preprocessing the burst pile image to obtain a burst pile binary image; Step 2, performing distance transformation on the burst binary image to obtain a distance transformed image of the burst image; Step 3, using a seed point marking method based on the solidity of the rock block contour, adaptively marking the seed points according to the distance transformation image, and obtaining a seed point map of the blast pile image; The specific method of obtaining the seed point map in step 3 is: Step 3.1, create a new canvas according to the size of the distance transformation image and set the initial value of the grayscale threshold; Step 3.2, performing grayscale threshold processing on the distance transformation image according to the current grayscale threshold to obtain a grayscale threshold cross-sectional image; Step 3.3, using contour detection technology to perform contour detection on the gray threshold cross-sectional image, and obtaining the solidity of each contour in each gray threshold cross-sectional image; Step 3.4, determine whether the solidity of each contour is greater than a given solidity threshold. If so, draw the contour into the newly created canvas and fill it as a seed point; Step 3.5, increase the grayscale threshold according to the preset method, and determine whether the increased grayscale threshold is greater than 255. If so, complete the drawing of the seed point map, otherwise return to step 3.2; Step 4: Use the watershed algorithm to segment the rock blocks in the blast pile image according to the seed point map.

2. The method according to claim 1, characterized in that After determining in step 3.4 that the solidity of a certain contour is greater than a given solidity threshold, further determine whether there is a background point in the contour. Only when there is no background point in the contour or the contour is irrelevant to the background point, draw the contour into the newly created canvas and fill it as a seed point.

3. The method according to claim 1, characterized in that The solidity of a rock block contour refers to the ratio of the area of ​​the rock block to the area of ​​the convex hull of the rock block contour.

4. The method according to claim 1, characterized in that: The solidity threshold is given as greater than 0.

82.

5. The method according to claim 1, characterized in that The preprocessing of step 1 includes: image grayscale, bilateral filtering denoising, contrast-constrained adaptive histogram equalization, image binarization, morphological optimization and area filtering.

6. The method according to claim 1, characterized in that The Phansalkar method based on local image properties is used to binarize the explosion pile image.

7. An electronic device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the computer program is executed by the processor, the processor is caused to implement the method according to any one of claims 1 to 6.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images