Method for identifying slope loose rock

Abstract

To provide a method for identifying slope loose rock.SOLUTION: After collecting an original image, the method constructs a first detail feature image and a second detail feature image based on the original image. Segmentation processing is performed on any image to implement area-based classification, and the area is classified based on the original block, the first feature block, and the second feature block of the same area and based on the three channel classification model. The classification value of the region is corrected to improve the classification accuracy, so that the rock region is found, the background region is peeled off, and the accuracy of identifying the loose rock is improved. The roughness of the rock and the condition of joints and cracks are characterized by the roughness condition of the surface of the rock area, and it is determined whether the rock is a loose rock according to the average surface roughness of the rock area, based on the fact that the larger the roughness is, the more easily the rock is eroded, the looser the rock is, and the more easily the rock collapses.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to the technical field of image processing, and in particular to a method for identifying loose rock masses on slopes. [Background technology]

[0002] Slope stability is a key issue in fields such as civil engineering, mining engineering, and transportation infrastructure. The presence of loose rock can lead to slope instability, which can cause geological disasters such as landslides and pose a threat to the safety of people's lives and property. Traditional methods for detecting rock on slopes primarily rely on manual on-site surveys, which are time-consuming, labor-intensive, and inefficient, making it difficult to quickly assess large areas. Currently, loose rock is identified in rock by collecting images of the target area. When performing image recognition for loose rock on rock slopes, the collected images often contain a lot of non-target background information due to the overgrown vegetation surrounding the target area. This background information hinders accurate identification of loose rock, reducing the accuracy of the identification. Summary of the Invention [Problem to be solved by the invention]

[0003] In response to the above-mentioned problems in the prior art, the present invention provides a method for identifying loose rock masses on slopes, which solves the problem in the prior art that the accuracy of identifying loose rock masses is low. [Means for solving the problem]

[0004] In order to achieve the above object of the invention, the present invention adopts the following technical solution: A method for identifying loose rock mass on a slope, comprising: S1, collecting images of a target area to obtain an original image; S2. Constructing a first detail feature image and a second detail feature image based on color values ​​in the original image; S3: performing cutting on the original image, the first detailed feature image, and the second detailed feature image, respectively, to obtain a plurality of original blocks corresponding to the original image, a plurality of first feature blocks corresponding to the first detailed feature image, and a plurality of second feature blocks corresponding to the second detailed feature image; S4, according to the original block, the first feature block and the second feature block of the same area, according to a three-channel classification model, perform classification on this area to obtain a classification value of this area; S5, performing a correction process on the classification value of each area to find the rock area; S6: calculating an average surface roughness of the rock region based on the first contour distribution unevenness, the second contour distribution unevenness, and the third contour distribution unevenness of each pixel point on the rock region; and when the average surface roughness is greater than a roughness threshold, determining the rock region as a loose rock body.

[0005] The beneficial effects of the present invention are as follows: after collecting an original image, the present invention constructs a first detailed feature image and a second detailed feature image based on the original image, prominently showing the image detailed features, and then performs a cutting process on both to realize area-specific classification; based on the original block, first feature block, and second feature block of the same area, the area is classified based on a three-channel classification model to obtain a classification value for the area; and then performs a correction process to improve classification accuracy, find the rock area, remove the background area, and improve the accuracy of identifying loose rock bodies; and further determine whether the rock is loose rock body based on the average surface roughness of the rock area. The present invention characterizes the roughness and joint and crack conditions of the rock based on the roughness condition of the rock area surface; the greater the roughness, the more likely the rock body is eroded, the looser the rock, and the more likely it is to crumble.

[0006] Furthermore, the S2 is S21, a substep of centering each pixel point in the original image; S22, a sub-step of calculating a first color feature value and a second color feature value based on the center neighborhood color value, respectively; S23, the sub-step of using the first color feature value as a new color value of the pixel point at the center to construct a first detail feature image; S24: The sub-step of setting the second color feature value as a new color value of the pixel point at the center to construct a second detailed feature image.

[0007] The beneficial effects of the above further invention are as follows: the present invention calculates a first color feature value and a second color feature value based on the neighboring color values ​​of each center, constructs a first detailed feature image by using the first color feature value as a new color value of the center, constructs a second detailed feature image by using the second color feature value as a new color value of the center, characterizes the color distribution situation at the center by the first color feature value and the second color feature value, and constructs a new feature image using these two distribution situations, thereby prominently showing the imaging features of rocks on the image.

[0008] JPEG2026012618000002.jpg42170

[0009] The beneficial effects of the above further invention are as follows: The present invention takes color values ​​at four positions, (x, y+1), (x, y-1), (x+1, y), and (x-1, y), which are in a neighborhood range at the center, and reflects the color distribution status of the four directions at this center based on the difference in color values ​​between the positions (x, y+1) and (x, y-1) and the difference in color values ​​between the positions (x+1, y) and (x-1, y).

[0010] JPEG2026012618000003.jpg44170

[0011] The beneficial effects of the above further invention are as follows: The present invention takes color values ​​at four positions, (x+1, y+1), (x-1, y-1), (x-1, y+1), and (x+1, y-1), which are in a neighborhood range at the center, and reflects the color distribution status of four other directions at this center based on the difference in color values ​​between the positions (x+1, y+1) and (x-1, y-1) and the difference in color values ​​between the positions (x-1, y+1) and (x+1, y-1).

[0012] Furthermore, the three-channel classification model in S4 includes a first channel processing unit, a second channel processing unit, a third channel processing unit, a first multiplier M1, a second multiplier M2, a first Maxpool layer, a first Avgpool layer, a second Maxpool layer, a second Avgpool layer, a first adder A1, a second adder A2, a first Concat layer, and a fully connected layer; an input terminal of the first channel processing unit is used to input a first feature block, an input terminal of the second channel processing unit is used to input an original block, and an input terminal of the third channel processing unit is used to input a second feature block; The first input terminal of the first multiplier M1 is connected to the output terminal of the first channel processing unit, the second input terminal thereof is connected to the output terminal of the second channel processing unit, and the output terminal thereof is respectively connected to the input terminal of the first Maxpool layer and the input terminal of the first Avgpool layer. The first input terminal of the second multiplier M2 is connected to the output terminal of the second channel processing unit, and the second input terminal thereof is connected to the output terminal of the third channel processing unit, and the output terminal thereof is respectively connected to the input terminal of the second Maxpool layer and the input terminal of the second Avgpool layer. The first input terminal of the first adder A1 is connected to the output terminal of the first Maxpool layer, and its second input terminal is connected to the output terminal of the first Avgpool layer. The first input terminal of the second adder A2 is connected to the output terminal of the second Maxpool layer, and its second input terminal is connected to the output terminal of the second Avgpool layer. The input terminals of the first Concat layer are respectively connected to the output terminals of the first adder A1 and the second adder A2, and their output terminals are connected to the input terminal of the fully connected layer, and the output terminal of the fully connected layer is the output terminal of the three-channel classification model.

[0013] The beneficial effects of the above further invention are as follows: the present invention employs three channel processing units to process the first feature block, the original block, and the second feature block, respectively; employs the output feature of the first channel processing unit to enhance the output feature of the second channel processing unit; employs the output feature of the third channel processing unit to enhance the output feature of the second channel processing unit, thereby fusing the distribution feature with the original image feature, prominently showing the feature situation of this area and making the feature prominent; further employs Maxpool layer and Avgpool layer to extract features, thereby extracting prominent features and global features, and performs classification by combining the prominent features and global features, thereby improving the classification accuracy.

[0014] Furthermore, the first channel processing unit, the second channel processing unit, and the third channel processing unit have the same structure, and each includes a first Convolution block, a second Convolution block, a third Convolution block, a fourth Convolution block, a fifth Convolution block, a sixth Convolution block, a second Concat layer, a third Concat layer, and a fourth Concat layer; the input terminal of the first Conv block is the input terminal of the first channel processing unit, the second channel processing unit, or the third channel processing unit, and its output terminal is respectively connected to the input terminal of the second Conv block and the input terminal of the third Conv block; the input terminal of the second ConcAt layer is respectively connected to the output terminal of the second Conv block and the output terminal of the third Conv block, and its output terminal is connected to the input terminal of the fourth Conv block; the input terminal of the third ConcAt layer is respectively connected to the output terminal of the second Conv block and the output terminal of the third Conv block, and its output terminal is connected to the input terminal of the fifth Conv block; the input terminal of the fourth ConcAt layer is respectively connected to the output terminal of the fourth Conv block and the output terminal of the fifth Conv block, and its output terminal is connected to the input terminal of the sixth Conv block, and the output terminal of the sixth Conv block is the output terminal of the first channel processing unit, the second channel processing unit, or the third channel processing unit; The size of the convolution kernel of the second Convolution Block and the fourth Convolution Block is 3*3; The size of the convolution kernel between the third Conv block and the fifth Conv block is 5*5.

[0015] The beneficial effects of the above further invention are as follows: The present invention extracts image features with different scales by constructing two side paths in the channel processing unit, using different convolution kernels, and sharing the output features of the second Conv block and the third Conv block.

[0016] Furthermore, the S5 S51, a substep of making each area a reference area; JPEG2026012618000004.jpg18170S53, when the distance is greater than the distance threshold, a substep of correcting the classification value of the reference area based on the classification values ​​of the surrounding areas of the reference area; S54: When the classification value after correction or the classification value not involved in the correction is greater than the classification threshold, mark the corresponding area as a candidate area, and compose the rock region with a plurality of candidate areas.

[0017] The beneficial effects of the above further invention are as follows: the present invention calculates the distance between the classification values ​​of the reference area and the surrounding area, and when the distance is greater than the distance threshold, it indicates that the difference between the classification values ​​of the reference area and the surrounding area is large, so the classification value of the surrounding area is adopted to correct the classification value of the reference area, and the rock area is selected based on the corrected classification value and the classification value not involved in the correction.

[0018] JPEG2026012618000005.jpg19170

[0019] The beneficial effects of the above further invention are as follows: the present invention corrects the classification value of the reference area based on the average value of the classification values ​​of multiple surrounding areas, and the greater the distance, the greater the degree of correction for the classification value of the reference area, thereby accurately extracting rock areas and removing background areas.

[0020] Furthermore, S6 S61, a sub-step of calculating a first contour distribution non-uniformity, a second contour distribution non-uniformity and a third contour distribution non-uniformity of each pixel point on each area of ​​the rock region; S62, a sub-step of calculating rock roughness of each area of ​​the rock region based on the first contour distribution unevenness, the second contour distribution unevenness and the third contour distribution unevenness of each pixel point on each area of ​​the rock region; S63, a substep of calculating the average surface roughness based on the rock roughness of each area of ​​the rock region; S64, when the surface average roughness is greater than the roughness threshold, the rock area is determined to be a loose rock body.

[0021] JPEG2026012618000006.jpg55170

[0022] The beneficial effects of the above further invention are as follows: the present invention first calculates the contour distribution unevenness of each pixel point on each area of ​​the rock region, characterizes the degree of roughness at each pixel point by the contour distribution unevenness of the pixel point, further calculates the rock roughness of each area of ​​the rock region, and averages the rock roughness of each area on the rock region to obtain the average surface roughness of the rock region; when the average surface roughness of the rock region is large, it indicates that the weathering or erosion of the rock is serious, and the rock is more likely to crumble; when the average surface roughness of the rock region is small, the rock surface is smooth, the rock surface has been polished for a long time, and the rock is harder.

[0023] JPEG2026012618000007.jpg39170

[0024] The beneficial effects of the above further invention are as follows: the present invention sums up the three contour distribution unevennesses of each pixel point on the rock area of ​​the original image, the first detailed feature image and the second detailed feature image to reflect the contour distribution situation at this pixel point, and sums up all pixel points to reflect the rock roughness of each area. [Brief explanation of the drawings]

[0025] [Figure 1] 1 is a flowchart of a method for identifying loose rock masses on slopes. [Figure 2] FIG. 1 is a structural schematic diagram of a three-channel classification model. [Figure 3] FIG. 2 is a structural schematic diagram of a first channel processing unit, a second channel processing unit and a third channel processing unit; DETAILED DESCRIPTION OF THE INVENTION

[0026] As shown in FIG. 1, a method for identifying slope loose rock masses, comprising: S1, an image acquisition step of acquiring an image of a target area to obtain an original image; S2: A feature image construction step of constructing a first detail feature image and a second detail feature image based on the color values ​​of the original image; S3: an image cutting step of performing cutting on the original image, the first detailed feature image, and the second detailed feature image, respectively, to obtain a plurality of original blocks corresponding to the original image, a plurality of first feature blocks corresponding to the first detailed feature image, and a plurality of second feature blocks corresponding to the second detailed feature image; S4, an area classification step, which is based on the original block, the first feature block and the second feature block of the same area and performs classification on the area based on a three-channel classification model to obtain a classification value for the area; S5, a classification value correction step of performing a correction process on the classification value of each area to find a rock area; S6: a loose rock body identification step of calculating an average surface roughness of the rock area based on the first contour distribution unevenness, the second contour distribution unevenness, and the third contour distribution unevenness of each pixel point on the rock area, and classifying the rock area as a loose rock body when the average surface roughness is greater than a roughness threshold.

[0027] In this embodiment, step S3 of performing the same cutting method on the original image, the first detailed feature image, and the second detailed feature image is specifically as follows: cutting is performed on the original image, the first detailed feature image, and the second detailed feature image, that is, cutting the original image into a plurality of areas, each of which is an original block, cutting the first detailed feature image into a plurality of areas, each of which is a first feature block, and cutting the second detailed feature image into a plurality of areas, each of which is a second feature block.

[0028] Since the image is two-dimensional, by performing K-1 horizontal cuts and K-1 vertical cuts on the original image, the first detailed feature image and the second detailed feature image, K*K original blocks, K*K first feature blocks and K*K second feature blocks are obtained, where K is a positive integer greater than or equal to 3.

[0029] The S2 is S21, a substep of centering each pixel point in the original image; S22, a sub-step of calculating a first color feature value and a second color feature value based on the center neighborhood color value, respectively; S23, the sub-step of using the first color feature value as a new color value of the pixel point at the center to construct a first detail feature image; S24: The sub-step of setting the second color feature value as a new color value of the pixel point at the center to construct a second detailed feature image.

[0030] The present invention calculates a first color feature value and a second color feature value based on the neighboring color values ​​of each center, constructs a first detailed feature image by using the first color feature value as the new color value of the center, and constructs a second detailed feature image by using the second color feature value as the new color value of the center.The color distribution situation at the center is characterized by the first color feature value and the second color feature value, and a new feature image is constructed using these two distribution situations, thereby clearly showing the imaging characteristics of the rock on the image.

[0031] JPEG2026012618000008.jpg42170

[0032] The present invention takes color values ​​at four positions, (x, y+1), (x, y-1), (x+1, y), and (x-1, y), which are in the vicinity of the center, and reflects the color distribution situation in the four directions at this center based on the difference in color values ​​between positions (x, y+1) and (x, y-1) and the difference in color values ​​between positions (x+1, y) and (x-1, y).

[0033] JPEG2026012618000009.jpg42170

[0034] The present invention takes color values ​​at four positions in the vicinity of the center, namely (x+1, y+1), (x-1, y-1), (x-1, y+1), and (x+1, y-1), and reflects the color distribution situation in four other directions at this center based on the difference in color values ​​between positions (x+1, y+1) and (x-1, y-1) and the difference in color values ​​between positions (x-1, y+1) and (x+1, y-1).

[0035] As shown in FIG. 2, the three-channel classification model in S4 includes a first channel processing unit, a second channel processing unit, a third channel processing unit, a first multiplier M1, a second multiplier M2, a first Maxpool layer, a first Avgpool layer, a second Maxpool layer, a second Avgpool layer, a first adder A1, a second adder A2, a first Concat layer, and a fully connected layer; an input terminal of the first channel processing unit is used to input a first feature block, an input terminal of the second channel processing unit is used to input an original block, and an input terminal of the third channel processing unit is used to input a second feature block; The first input terminal of the first multiplier M1 is connected to the output terminal of the first channel processing unit, the second input terminal thereof is connected to the output terminal of the second channel processing unit, and the output terminal thereof is respectively connected to the input terminal of the first Maxpool layer and the input terminal of the first Avgpool layer. The first input terminal of the second multiplier M2 is connected to the output terminal of the second channel processing unit, and the second input terminal thereof is connected to the output terminal of the third channel processing unit, and the output terminal thereof is respectively connected to the input terminal of the second Maxpool layer and the input terminal of the second Avgpool layer. The first input terminal of the first adder A1 is connected to the output terminal of the first Maxpool layer, and its second input terminal is connected to the output terminal of the first Avgpool layer. The first input terminal of the second adder A2 is connected to the output terminal of the second Maxpool layer, and its second input terminal is connected to the output terminal of the second Avgpool layer. The input terminals of the first Concat layer are respectively connected to the output terminals of the first adder A1 and the second adder A2, and their output terminals are connected to the input terminal of the fully connected layer, and the output terminal of the fully connected layer is the output terminal of the three-channel classification model.

[0036] The present invention employs three channel processing units to process the first feature block, the original block, and the second feature block, respectively; employs the output features of the first channel processing unit to enhance the output features of the second channel processing unit; employs the output features of the third channel processing unit to enhance the output features of the second channel processing unit, thereby fusing the distribution features with the original image features, highlighting the feature situation of this area and making the features prominent; and further employs Maxpool layer and Avgpool layer to extract features, thereby extracting prominent features and global features, and classifying the prominent features and global features together to improve the classification accuracy.

[0037] As shown in FIG. 3 , the first channel processing unit, the second channel processing unit, and the third channel processing unit have the same structure, and each includes a first Convolution block, a second Convolution block, a third Convolution block, a fourth Convolution block, a fifth Convolution block, a sixth Convolution block, a second Concat layer, a third Concat layer, and a fourth Concat layer; the input terminal of the first Conv block is the input terminal of the first channel processing unit, the second channel processing unit, or the third channel processing unit, and its output terminal is respectively connected to the input terminal of the second Conv block and the input terminal of the third Conv block; the input terminal of the second ConcAt layer is respectively connected to the output terminal of the second Conv block and the output terminal of the third Conv block, and its output terminal is connected to the input terminal of the fourth Conv block; the input terminal of the third ConcAt layer is respectively connected to the output terminal of the second Conv block and the output terminal of the third Conv block, and its output terminal is connected to the input terminal of the fifth Conv block; the input terminal of the fourth ConcAt layer is respectively connected to the output terminal of the fourth Conv block and the output terminal of the fifth Conv block, and its output terminal is connected to the input terminal of the sixth Conv block, and the output terminal of the sixth Conv block is the output terminal of the first channel processing unit, the second channel processing unit, or the third channel processing unit; The size of the convolution kernel of the second Convolution Block and the fourth Convolution Block is 3*3; The size of the convolution kernel between the third Conv block and the fifth Conv block is 5*5.

[0038] In this embodiment, the size of the convolution kernel of the first Conv block and the sixth Conv block is 1*1.

[0039] The present invention constructs two side paths in the channel processing unit, uses different convolution kernels, and shares the output features of the second Conv block with the output features of the third Conv block to extract image features with different scales.

[0040] In this embodiment, each Conv block includes a convolution layer, a ReLU layer, and a BN layer.

[0041] The step S5 includes the following sub-steps.

[0042] S51, each area is set as a reference area.

[0043] JPEG2026012618000010.jpg18170 In the present invention, since the image is cut in the horizontal and vertical directions in step S3, there are multiple surrounding areas for each area.

[0044] S53, when the distance is greater than the distance threshold, correct the classification value of the reference area based on the classification values ​​of the surrounding areas of the reference area.

[0045] S54: When the classification value after correction or the classification value not involved in the correction is greater than the classification threshold, the corresponding area is marked as a candidate area, and a plurality of candidate areas constitute a rock region.

[0046] In the present invention, the surrounding area is the area in contact with the reference area.

[0047] In this embodiment, the distance threshold is a threshold set for each distance, and is used to find a reference area having a large difference in classification value from the surrounding area.

[0048] The present invention calculates the distance between the classification values ​​of the reference area and the surrounding area, and when the distance is greater than the distance threshold, it indicates that the difference between the classification values ​​of the reference area and the surrounding area is large, so the classification value of the surrounding area is used to correct the classification value of the reference area, and rock areas are selected based on the corrected classification value and the classification value not involved in the correction.

[0049] In the present invention, since the image is cut in S3, there are multiple areas on the original image, multiple areas on the first detailed feature image, and multiple areas on the second detailed feature image. Because the cutting method is the same, after adopting a three-channel classification model to classify the same areas on the three images, the classification value of each area on the original image, the classification value of each area on the first detailed feature image, and the classification value of each area on the second detailed feature image are obtained, and the classification values ​​of the same areas in the three images are the same.

[0050] In this embodiment, if the classification value for the same area has been corrected, the corrected classification value is used in step S54 to compare with the classification threshold, and the classification value before correction is no longer used for comparison with the classification threshold.

[0051] In this embodiment, the classification threshold is a threshold set for each classification value and is used to find areas where the classification value is large.

[0052] JPEG2026012618000011.jpg19170

[0053] The present invention corrects the classification value of the reference area based on the average value of the classification values ​​of multiple surrounding areas, and the greater the distance, the greater the degree of correction to the classification value of the reference area, thereby accurately extracting rock areas and removing background areas.

[0054] The step S6 is S61, a sub-step of calculating a first contour distribution non-uniformity, a second contour distribution non-uniformity and a third contour distribution non-uniformity of each pixel point on each area of ​​the rock region; S62, a sub-step of calculating rock roughness of each area of ​​the rock region based on the first contour distribution unevenness, the second contour distribution unevenness and the third contour distribution unevenness of each pixel point on each area of ​​the rock region; S63, a substep of calculating the average surface roughness based on the rock roughness of each area of ​​the rock region; S64, when the surface average roughness is greater than the roughness threshold, the rock area is determined to be a loose rock body.

[0055] In this embodiment, the roughness threshold is a threshold set for the average surface roughness.

[0056] JPEG2026012618000012.jpg55170

[0057] The present invention first calculates the contour distribution unevenness of each pixel point on each area on the rock region, and characterizes the degree of roughness at each pixel point by the contour distribution unevenness of the pixel point. Then, it calculates the rock roughness of each area on the rock region and takes the average value of the rock roughness of each area on the rock region to obtain the average surface roughness of the rock region. When the average surface roughness of the rock region is large, it indicates that the weathering or erosion of the rock is serious, and the rock is more likely to crumble. When the average surface roughness of the rock region is small, the rock surface is smooth, the rock surface has been polished for a long time, and the rock is harder.

[0058] JPEG2026012618000013.jpg39170

[0059] The present invention adds up the three contour distribution unevennesses of each pixel point on the rock area of ​​the original image, the first detailed feature image, and the second detailed feature image to reflect the contour distribution situation at this pixel point, and adds up all pixel points to reflect the rock roughness of each area.

[0060] In this embodiment, all thresholds are specifically set based on experiments or experiences.

[0061] After collecting the original image, the present invention constructs a first detailed feature image and a second detailed feature image based on the original image, highlighting the image detail features, and then performs a cutting process to realize area-specific classification. Based on the original block, first feature block, and second feature block of the same area, the area is classified based on a three-channel classification model to obtain a classification value for the area, and then performs a correction process to improve the classification accuracy, find the rock area, remove the background area, and improve the accuracy of identifying loose rock. Furthermore, the average surface roughness of the rock area is used to determine whether the rock is loose rock. The present invention characterizes the roughness and joint and crack conditions of the rock based on the roughness of the rock area surface, and the greater the roughness, the more likely the rock is eroded, loose, and more likely to collapse.

Claims

1. A method for identifying loose rock mass on a slope, comprising: S1, collecting images of a target area to obtain an original image; S2. Constructing a first detail feature image and a second detail feature image based on color values ​​in the original image; S3: performing cutting on the original image, the first detailed feature image, and the second detailed feature image, respectively, to obtain a plurality of original blocks corresponding to the original image, a plurality of first feature blocks corresponding to the first detailed feature image, and a plurality of second feature blocks corresponding to the second detailed feature image; S4. According to the original block, the first feature block and the second feature block of the same area, perform classification on the area based on a three-channel classification model to obtain a classification value of the area; S5: A step of performing a correction process on the classification value of each area to find a rock area; S6: calculating an average surface roughness of the rock area based on the first contour distribution unevenness, the second contour distribution unevenness, and the third contour distribution unevenness of each pixel point on the rock area; and when the average surface roughness is greater than a roughness threshold, determining the rock area as a loose rock body; The S2 is S21, a sub-step of centering each pixel point in the original image; S22, a sub-step of calculating a first color feature value and a second color feature value based on the center neighborhood color value; S23: constructing a first detailed feature image by using the first color feature value as a new color value of the pixel point at the center; S24: constructing a second detailed feature image by using the second color feature value as a new color value of the pixel point at the center; The three-channel classification model in S4 includes a first channel processing unit, a second channel processing unit, a third channel processing unit, a first multiplier M1, a second multiplier M2, a first Maxpool layer, a first Avgpool layer, a second Maxpool layer, a second Avgpool layer, a first adder A1, a second adder A2, a first Concat layer, and a fully connected layer; an input terminal of the first channel processing unit is used to input a first feature block, an input terminal of the second channel processing unit is used to input an original block, and an input terminal of the third channel processing unit is used to input a second feature block; The first input terminal of the first multiplier M1 is connected to the output terminal of the first channel processing unit, the second input terminal of the first multiplier M1 is connected to the output terminal of the second channel processing unit, and the output terminals of the first multiplier M1 are respectively connected to the input terminal of the first Maxpool layer and the input terminal of the first Avgpool layer; the first input terminal of the second multiplier M2 is connected to the output terminal of the second channel processing unit, and the second input terminal of the second multiplier M2 is connected to the output terminal of the third channel processing unit, and the output terminals of the second Maxpool layer and the input terminal of the second Avgpool layer; A first input terminal of the second adder A1 is connected to the output terminal of the first Maxpool layer, and its second input terminal is connected to the output terminal of the first Avgpool layer; a first input terminal of the second adder A2 is connected to the output terminal of the second Maxpool layer, and its second input terminal is connected to the output terminal of the second Avgpool layer; input terminals of the first Concat layer are respectively connected to the output terminal of the first adder A1 and the output terminal of the second adder A2, and their output terminals are connected to the input terminal of the fully connected layer, and the output terminal of the fully connected layer is the output terminal of the three-channel classification model.

2. the first channel processing unit, the second channel processing unit, and the third channel processing unit have the same structure, and each includes a first Conv block, a second Conv block, a third Conv block, a fourth Conv block, a fifth Conv block, a sixth Conv block, a second Concat layer, a third Concat layer, and a fourth Concat layer; the input terminal of the first Conv block is the input terminal of the first channel processing unit, the second channel processing unit, or the third channel processing unit, and its output terminal is respectively connected to the input terminal of the second Conv block and the input terminal of the third Conv block; the input terminal of the second Concat layer is respectively connected to the output terminal of the second Conv block and the output terminal of the third Conv block, and its output terminal is connected to the input terminal of the fourth Conv block; the input terminal of the third Concat layer is respectively connected to the output terminal of the second Conv block and the output terminal of the third Conv block, and its output terminal is connected to the input terminal of the fifth Conv block; the input terminal of the fourth Concat layer is respectively connected to the output terminal of the fourth Conv block and the output terminal of the fifth Conv block, and its output terminal is connected to the input terminal of the sixth Conv block, and the output terminal of the sixth Conv block is the output terminal of the first channel processing unit, the second channel processing unit, or the third channel processing unit; The size of the convolution kernel of the second Conv block and the fourth Conv block is 3*3; The method for identifying slope loose rock masses according to claim 1, wherein the size of the convolution kernel of the third Conv block and the fifth Conv block is 5*5.

3. The step S5 is S51, a sub-step of setting each area as a reference area; S53, when the distance is greater than the distance threshold, correcting the classification value of the reference area based on the classification values ​​of the surrounding areas of the reference area; S54: When the classification value after correction or the classification value not involved in the correction is greater than the classification threshold, mark the corresponding area as a candidate area, and compose a rock region from a plurality of candidate areas.

4.

5. The step S6 is S61: a sub-step of calculating a first contour distribution non-uniformity, a second contour distribution non-uniformity and a third contour distribution non-uniformity of each pixel point on each area of ​​the rock region; S62, a sub-step of calculating rock roughness of each area of ​​the rock region based on the first contour distribution unevenness, the second contour distribution unevenness, and the third contour distribution unevenness of each pixel point on each area of ​​the rock region; S63, a sub-step of calculating an average surface roughness based on the rock roughness of each area of ​​the rock region; 2. The method for identifying loose rock mass on a slope according to claim 1, further comprising the sub-step of: S64, when the average surface roughness is greater than the roughness threshold, classifying the rock region as a loose rock mass.

6.

7.

Images