An image-based evaluation method and device for the integrity of a rock mass of a working face

Abstract

The application provides an image-based evaluation method and device for the completeness of a tunnel face rock mass, and belongs to the technical field of tunnel engineering. The embodiment of the application extracts features from the pre-processed tunnel face rock mass image through a joint fissure extraction model, which can more quickly and accurately obtain more joint fissure information of the tunnel face. Meanwhile, by taking different preset positions in the joint fissure feature map as the center of radiation, multiple virtual line layout maps are drawn, which can make the virtual lines more evenly distributed on the joint fissure feature map, increase the line density of each part in the joint fissure feature map, and better reflect the intersection relationship between the virtual lines and the joint fissures. Finally, according to the equivalent area corresponding to each virtual line layout map, the completeness of the tunnel face rock mass is determined, the fine quantitative evaluation of the surrounding rock structure of the tunnel face is realized, and more accurate data reference is provided for tunnel design, construction, operation and maintenance.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of tunnel engineering, in particular to a method and device for evaluating the integrity of a tunnel face based on images. BACKGROUND

[0002] A tunnel face is also known as a working face, which refers to the face that is constantly advancing forward in the process of excavating a tunnel (in coal mining, mining or tunnel engineering). The integrity of rock mass is a basic measure for evaluating the quality and stability of rock mass, which is divided into five levels: complete, relatively complete, poor completeness, relatively broken and broken. The quantitative classification of the integrity of rock mass of a tunnel face is of great significance in tunnel engineering construction. If the quality of the surrounding rock of a tunnel face cannot be scientifically and accurately determined, it will inevitably have a serious impact on tunnel design, construction, operation and maintenance, resulting in waste of construction costs and even casualties.

[0003] At present, the traditional survey of the integrity of rock mass of a tunnel face is generally based on visual observation and engineering experience to infer the development of joint fissures in the tunnel face, manually extract a joint fissure map of the tunnel face, and then evaluate the integrity of rock mass of the tunnel face based on the manually extracted joint fissure map of the tunnel face using the Z-RBI method.

[0004] However, the manually extracted joint fissure map of the tunnel face is low in efficiency and poor in accuracy, and the Z-RBI method takes the center of the tunnel floor as the center of a circle and arranges the measuring lines in each direction from the same starting point, which results in that the measuring lines in the middle and lower parts of the tunnel face are relatively dense, while the measuring lines in the parts of the tunnel face that are closer to the lining structure are relatively sparse, and thus it is difficult to accurately reflect the integrity of rock mass in each part of the tunnel face. SUMMARY

[0005] The present application provides a method and device for evaluating the integrity of rock mass of a tunnel face based on images to solve the problem that the prior art cannot accurately reflect the integrity of rock mass in each part of the tunnel face.

[0006] To solve the above problem, the present application adopts the following technical solutions:

[0007] In a first aspect, the present application provides a method for evaluating the integrity of rock mass of a tunnel face based on images, which comprises:

[0008] inputting a preprocessed tunnel face rock mass image into a pre-trained joint fissure extraction model to output a joint fissure feature map, wherein the joint fissure feature map is used to represent the fissure information of the tunnel face rock mass;

[0009] Based on the joint fissure feature map, a plurality of different virtual survey line layout maps are obtained; wherein, different virtual survey line layout maps are obtained by taking different preset positions in the joint fissure feature map as a radiation center, and preset angles as a layout interval, and drawing virtual survey lines on the joint fissure feature map;

[0010] For each virtual survey line layout map, based on the rock mass blockiness coefficient corresponding to each virtual survey line in the virtual survey line layout map and the preset angle, the equivalent area of the virtual survey line layout map is determined.

[0011] Based on the equivalent areas corresponding to the plurality of virtual survey line layout maps respectively, the rock mass integrity degree of the working face rock mass is determined.

[0012] In an embodiment of the present application, the preset positions include: a vault, a left arch foot, a right arch foot and a working face centroid; wherein, the working face centroid is arranged in a central region of the joint fissure feature map.

[0013] In an embodiment of the present application, for each virtual survey line layout map, based on the rock mass blockiness coefficient corresponding to each virtual survey line in the virtual survey line layout map and the preset angle, the equivalent area of the virtual survey line layout map is determined, comprising:

[0014] For each virtual survey line layout map, the rock mass blockiness coefficient of each virtual survey line in the virtual survey line layout map is calculated.

[0015] Based on the rock mass blockiness coefficients corresponding to adjacent two virtual survey lines and the preset angle, the equivalent area of the measurement region corresponding to the adjacent two virtual survey lines is determined.

[0016] Based on the equivalent area of each measurement region, the equivalent area of the virtual survey line layout map is determined.

[0017] In an embodiment of the present application, based on the rock mass blockiness coefficients corresponding to adjacent two virtual survey lines and the preset angle, the equivalent area of the measurement region corresponding to the adjacent two virtual survey lines is determined, comprising:

[0018] Based on the rock mass blockiness coefficients corresponding to adjacent two virtual survey lines and the preset angle, the triangular region area is calculated by the sine theorem;

[0019] The triangular region area is determined as the equivalent area of the measurement region corresponding to the adjacent two virtual survey lines.

[0020] In an embodiment of the present application, based on the equivalent areas corresponding to the plurality of virtual survey line layout maps respectively, the rock mass integrity degree of the working face rock mass is determined, comprising:

[0021] For each of the virtual survey line layout maps, an initial evaluation index corresponding to the virtual survey line layout map is determined based on a ratio of an equivalent area of the virtual survey line layout map to a preset tunnel face area;

[0022] An average value of the initial evaluation indexes of all the virtual survey line layout maps is determined as a comprehensive quantification index of the rock mass integrity of the tunnel face rock mass;

[0023] Based on a mapping relationship between the comprehensive quantification index of the rock mass integrity and a rock mass integrity type, the rock mass integrity of the tunnel face rock mass is determined.

[0024] In an embodiment of the present application, the pretreated tunnel face rock mass image is input into a pre-trained joint fissure extraction model, and an initial joint fissure feature map is output.

[0025] Based on n different cropping sizes, the tunnel face rock mass image is cropped to obtain n sub-group images, each of which includes a plurality of first sub-images under a corresponding cropping size; n is an integer greater than or equal to 1;

[0026] For any sub-group, the first sub-images in the sub-group are subjected to image equalization processing to obtain second sub-images; the second sub-images are input into a pre-trained joint fissure extraction model to output initial joint fissure feature sub-images; and the initial joint fissure feature sub-images are merged to obtain an initial joint fissure feature map corresponding to the sub-group;

[0027] Based on the initial joint fissure feature maps corresponding to the n sub-groups, the joint fissure feature map is obtained.

[0028] In an embodiment of the present application, the image equalization processing of the first sub-images in the sub-group to obtain the second sub-images includes:

[0029] The first sub-images are subjected to grayscale processing, and the distribution frequency of each original grayscale value in the first sub-images is obtained;

[0030] Based on the distribution frequency of each original grayscale value, a grayscale cumulative distribution frequency is calculated;

[0031] Based on the grayscale cumulative distribution frequency and the grayscale level of the first sub-images, a target grayscale value is obtained;

[0032] Based on the target grayscale value, the second sub-images after image equalization are obtained.

[0033] In an embodiment of the present application, the joint fissure extraction model includes a symmetric encoding path and a decoding path; the second sub-images are input into a pre-trained joint fissure extraction model to output initial joint fissure feature sub-images, including:

[0034] input the second subgraph into the encoding path, and extract feature information of the second subgraph layer by layer from shallow to deep by using a preset number of encoding modules to obtain an encoded feature graph; each encoding module comprises an encoding network and a skip layer network which is connected to the encoding network in a skip layer manner, and the skip layer network is configured to superimpose a feature graph output by a previous encoding module on a feature graph output by the encoding network of a current encoding module, and the superimposed feature graph is the output of the current encoding module;

[0035] input the encoded feature graph into the decoding path, and perform layer-by-layer decoding on the encoded feature graph by using a decoding module corresponding to each encoding module to output the initial joint fissure feature subgraph.

[0036] In a second aspect, based on the same inventive concept, an embodiment of the present application provides a device for evaluating the integrity of a rock mass of a working face based on images, the device comprising:

[0037] a feature extraction module configured to input a preprocessed image of the rock mass of the working face into a pre-trained joint fissure extraction model, and output a joint fissure feature graph, the joint fissure feature graph being configured to represent fissure information of the rock mass of the working face;

[0038] a survey line layout module configured to obtain a plurality of different virtual survey line layout graphs based on the joint fissure feature graph, wherein each different virtual survey line layout graph is obtained by drawing virtual survey lines on the joint fissure feature graph with different preset positions in the joint fissure feature graph as the center of radiation and a preset angle as the layout interval;

[0039] an equivalent area determination module configured to determine, for each virtual survey line layout graph, an equivalent area of the virtual survey line layout graph based on a rock mass blockiness coefficient corresponding to each virtual survey line in the virtual survey line layout graph and the preset angle;

[0040] a rock mass integrity determination module configured to determine the rock mass integrity of the rock mass of the working face based on the equivalent areas corresponding to the plurality of virtual survey line layout graphs.

[0041] In an embodiment of the present application, the preset positions include a vault, a left arch foot, a right arch foot and a working face centroid, wherein the working face centroid is arranged in a central region of the joint fissure feature graph.

[0042] In an embodiment of the present application, the equivalent area determination module comprises:

[0043] a rock mass blockiness coefficient calculation submodule configured to calculate, for each virtual survey line layout graph, a rock mass blockiness coefficient of each virtual survey line in the virtual survey line layout graph;

[0044] The first equivalent area determination submodule is configured to determine, based on the blockiness coefficient of the rock mass corresponding to the two adjacent virtual survey lines and the preset angle, an equivalent area of a measurement region corresponding to the two adjacent virtual survey lines.

[0045] The second equivalent area determination submodule is configured to determine, based on the equivalent area of each measurement region, an equivalent area of the virtual survey line layout diagram.

[0046] In an embodiment of the present application, the first equivalent area determination submodule includes:

[0047] The sine theorem calculation unit is configured to calculate, based on the blockiness coefficient of the rock mass corresponding to the two adjacent virtual survey lines and the preset angle, a triangular region area by using the sine theorem.

[0048] The equivalent area determination unit is configured to determine the triangular region area as the equivalent area of the measurement region corresponding to the two adjacent virtual survey lines.

[0049] In an embodiment of the present application, the rock mass integrity determination module includes:

[0050] The initial evaluation index determination submodule is configured to determine, for each virtual survey line layout diagram, an initial evaluation index corresponding to the virtual survey line layout diagram based on a ratio of the equivalent area of the virtual survey line layout diagram to a preset working face area.

[0051] The rock mass integrity comprehensive quantification index determination submodule is configured to determine, as the rock mass integrity comprehensive quantification index of the working face rock mass, an average value of the initial evaluation indexes of all virtual survey line layout diagrams.

[0052] The rock mass integrity determination submodule is configured to determine the rock mass integrity of the working face rock mass based on a mapping relationship between the rock mass integrity comprehensive quantification index and a rock mass integrity type.

[0053] In an embodiment of the present application, the feature extraction module includes:

[0054] The cropping submodule is configured to crop the working face rock mass image based on n different cropping sizes to obtain n sub-group sets, each of which includes a plurality of first sub-pictures under a corresponding cropping size; n is an integer greater than or equal to 1.

[0055] The initial joint fissure feature map acquisition submodule is configured to, for any sub-group set, perform image equalization processing on the first sub-pictures in the sub-group set to obtain second sub-pictures; input the second sub-pictures into a pre-trained joint fissure extraction model to output an initial joint fissure feature sub-picture; and merge the initial joint fissure feature sub-pictures to obtain an initial joint fissure feature map corresponding to the sub-group set.

[0056] The joint fissure feature map obtaining submodule is configured to obtain the joint fissure feature map based on the initial joint fissure feature maps corresponding to the n subgraph groups.

[0057] In an embodiment of the present application, the image equalization processing submodule comprises:

[0058] The distribution frequency obtaining unit is configured to perform grayscale processing on the first subgraph and obtain a distribution frequency corresponding to each original grayscale value in the first subgraph;

[0059] The cumulative distribution frequency unit is configured to calculate a grayscale cumulative distribution frequency based on the distribution frequency corresponding to each original grayscale value.

[0060] The target grayscale value obtaining unit is configured to obtain a target grayscale value based on the grayscale cumulative distribution frequency and the grayscale level of the first subgraph.

[0061] The second subgraph obtaining unit is configured to obtain the second subgraph after image equalization based on the target grayscale value.

[0062] In an embodiment of the present application, the joint fissure extraction model comprises a symmetric encoding path and a decoding path, and the subgraph output submodule comprises:

[0063] The encoding unit is configured to input the second subgraph into the encoding path, extract feature information of the second subgraph layer by layer from shallow to deep by using a preset number of encoding modules, and obtain an encoded feature map; each encoding module comprises an encoding network and a skip layer network connected to the encoding network in a skip layer manner, the skip layer network is configured to superimpose the feature map output by the previous encoding module on the feature map output by the encoding network of the current encoding module, and the superimposed feature map is the output of the current encoding module.

[0064] The decoding unit is configured to input the encoded feature map into the decoding path, perform layer-by-layer decoding on the encoded feature map by using a decoding module corresponding to each encoding module, and output the initial joint fissure feature subgraph.

[0065] Compared with the prior art, the present application has the following advantages:

[0066] The embodiment of the application provides a kind of based on image's rock mass integrity degree evaluation method of face, by joint fracture extraction model to the rock mass image of face after pre-processing is carried out feature extraction, more quickly, more accurately, more face joint fracture information can be obtained;Meanwhile, with different preset positions in joint fracture feature map as radiation center, a plurality of virtual survey line layout maps can be drawn, so that virtual survey line is more evenly distributed on joint fracture feature map, the survey line density of each part in joint fracture feature map is encrypted, the intersection relationship of virtual survey line and joint fracture can be better reflected, and finally the rock mass integrity degree of face rock mass is determined according to the equivalent area corresponding to each virtual survey line layout map, to realize the fine quantitative evaluation of face surrounding rock structure, and provide more accurate data reference for tunnel design, construction, operation and maintenance. BRIEF DESCRIPTION OF DRAWINGS

[0067] In order to more clearly illustrate the technical solutions in the embodiments of the application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are only some embodiments of the application, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0068] Figure 1 is the virtual survey line layout schematic diagram of Z-RBI method in the prior art.

[0069] Figure 2 is the step flow chart of the rock mass integrity degree evaluation method of face based on image in an embodiment of the application.

[0070] Figure 3 is the initial joint fracture feature map obtained when the clipping size is 32x32 pixels in an embodiment of the application.

[0071] Figure 4 is the initial joint fracture feature map obtained when the clipping size is 256x256 pixels in an embodiment of the application.

[0072] Figure 5 is the first subgraph schematic diagram in an embodiment of the application.

[0073] Figure 6 is the second subgraph schematic diagram in an embodiment of the application.

[0074] Figure 7 is the structure schematic diagram of DeepIntactness model in an embodiment of the application.

[0075] Figure 8 is the original image of face at ZK92+850 of Bimoyuan tunnel in an embodiment of the application.

[0076] Figure 9 is the joint fracture feature map of manual extraction in an embodiment of the present application.

[0077] Figure 10 is the joint fracture feature map of DeepIntactness model extraction in an embodiment of the present application.

[0078] Figure 11 is the virtual survey line layout map corresponding to each of the four radiation centers in an embodiment of the present application.

[0079] Figure 12 is the RBI index rose map of the rock mass integrity of the working face in an embodiment of the present application.

[0080] Figure 13 is the functional module schematic diagram of the image-based working face rock mass integrity evaluation device in an embodiment of the present application.

[0081] The drawings show: 1300, image-based working face rock mass integrity evaluation device; 1301, feature extraction module; 1302, survey line layout module; 1303, equivalent area determination module; 1304, rock mass integrity determination module. DETAILED DESCRIPTION

[0082] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0083] It should be noted that RBI (rock mass blockiness coefficient) is to obtain the actual length of the core pie by means of advanced adit or drilling, and to use the cumulative value of the product of the core sampling rate of five intervals [3cm, 10cm], [10cm, 30cm], [30cm, 50cm], [50cm, 100cm] and [100cm, +∞] and the smaller threshold value of each interval as the rock mass quality measurement standard, which is expressed by the formula as follows:

[0084] RBI = 3C r3 + 10C r10 + 30C r30 + 50C r50 + 100C r100 (1);

[0085] In the formula, C r3 , C r10 , C r30 , Cr50 , C r100 The core recovery rate of the five intervals [3cm, 10cm], [10cm, 30cm], [30cm, 50cm], [50cm, 100cm] and [100cm, +∞] is expressed as a percentage and taken as a weight value; 3, 10, 30, 50, 100 are constants.

[0086] Referring to Table 1, the correspondence between the rock mass integrity and RBI is shown.

[0087] Table 1: Correspondence between rock mass integrity and RBI

[0088] Rock mass integrity RBI Intact >31.39 Relatively intact 14.59～31.4 Relatively broken 4.74～14.59 Broken 0.58～4.74 Extremely broken <0.58

[0089] The rock mass blockiness coefficient RBI is a comprehensive index representing the size of the rock mass blockiness and its structure type, which reflects the size of the blockiness (size) of the rock mass and the mutual combination relationship. The larger the RBI, the better the rock mass integrity. The maximum RBI is 100, which represents that the length of the complete core is greater than 100cm, which is a typical whole block structure.

[0090] Referring to Figure 1 , a virtual survey line layout diagram of the Z-RBI method is shown. The Z-RBI method uses the azimuth angle of the rose diagram to represent the survey line position, and the radius of the rose diagram to represent the corresponding azimuth rock mass structure RBI (rock mass blockiness coefficient) quantization value. The Z-RBI method takes the midpoint of the tunnel floor as the radiation center, and sets the survey line at an interval of 10°, and sets 19 virtual survey lines towards the tunnel contour line. The average RBI of adjacent survey lines is used as the radius to draw each equivalent area circular arc shape, and the sector area enclosed by the equivalent area circular arc shape is used as a measure of the rock mass structure characteristics of the survey area. The sum of the sector areas enclosed by each equivalent area circular arc shape is compared with the area of a semicircle with a radius of 100, and the final Z-RBI value is obtained. Based on the rock mass structure type and Z-RBI correspondence table, the corresponding rock mass integrity of the Z-RBI value can be obtained.

[0091] However, the Z-RBI method takes the midpoint of the tunnel floor as the radiation center, and the survey lines in each direction are laid out with the same starting point, resulting in a higher density of survey lines in the middle and lower parts of the tunnel face, and a lower density of survey lines in the parts of the tunnel face closer to the lining structure, making it difficult to accurately reflect the rock mass integrity of each part of the tunnel face. The RBI index rose diagram can preliminarily reflect the distribution direction of the layered rock mass, but the thickness of the layered rock mass needs to be small, and the trace lines of the layered rock mass intersecting the tunnel face need to be distributed more completely on the entire tunnel face. If the trace line direction is similar to the direction of the survey area where it is mainly distributed, there will be fewer intersections with the survey lines, making it difficult to be detected.

[0092] In view of the defects in the prior art, the present application aims to provide an image-based evaluation method for the integrity of a tunnel face rock mass, which can more quickly and accurately extract more information about the joints and fissures in the tunnel face by using a joint and fissure extraction model to extract features from a preprocessed tunnel face rock mass image; and the method can better reflect the intersection relationship between the virtual lines and the joints and fissures by encrypting the line density of each part in the joint and fissure feature map, so that the calculation results can more objectively reflect the integrity of the tunnel face rock mass and provide more accurate data for tunnel design, construction, operation and maintenance.

[0093] Referring to Figure 2 , an image-based evaluation method for the integrity of a tunnel face rock mass is shown, which can specifically include the following steps:

[0094] S101: input a preprocessed tunnel face rock mass image into a pre-trained joint and fissure extraction model to obtain a joint and fissure feature map.

[0095] It should be noted that due to the limitations of tunnel construction conditions, uneven lighting and large fluctuation amplitude of the tunnel face will cause shadows, and the tunnel face surface may have multiple lithologies, and different lithologies have different light reflection abilities, so there is a large difference in the gray values of different areas of the image, which will interfere with the extraction of joints and fissures. If a high-resolution image is directly input into the joint and fissure extraction model for fissure extraction, the computer hardware device requirements will be high, and the effect will not be ideal.

[0096] In this embodiment, in order to more comprehensively obtain the full-face information of the tunnel face, a high-resolution camera can be used to collect a tunnel face rock mass image with an image size greater than 3000x2000 pixels, and the tunnel face rock mass image can be preprocessed. Specifically, the image can be subjected to image equalization processing and image mean value processing to increase the contrast of the image, so that the extracted joint and fissure feature map can reflect more information about the joints and fissures in the tunnel face, and provide an objective and accurate basis for subsequent quantitative classification of the integrity of the tunnel face rock mass.

[0097] S102: based on the joint and fissure feature map, obtain a plurality of different virtual line layout maps; wherein the different virtual line layout maps are obtained by drawing virtual lines on the joint and fissure feature map with different preset positions in the joint and fissure feature map as the center of radiation and a preset angle as the layout interval.

[0098] In the embodiment, considering that the integrity of the rock mass of the tunnel face at the hole wall has the greatest impact on the tunnel lining, in order to reflect the integrity of the rock mass of the tunnel face at the hole wall and considering the integrity of each part of the tunnel face, different preset positions in the joint fissure feature map are taken as the radiation center to draw a plurality of different virtual survey line layout maps.

[0099] In the embodiment, referring to Figure 5 , four different virtual survey line layout maps can be obtained by taking the four different positions of the tunnel face, i.e., the vault, the left arch foot, the right arch foot, and the center of the tunnel face, as the radiation center and a preset angle of 10°. It should be noted that the present embodiment does not limit the position of the radiation center and the preset angle, which can be set according to actual needs.

[0100] It should be further noted that the center of the tunnel face is arranged in the central region of the joint fissure feature map, and can be arranged on the vertical line of the tunnel floor at a distance of 4R / 3π from the floor, where R is the radius of the rock mass of the tunnel face in the joint fissure feature map. Drawing a plurality of virtual survey line layout maps by taking different preset positions in the joint fissure feature map as the radiation center can make the virtual survey lines more evenly distributed on the joint fissure feature map, increase the survey line density of each part of the joint fissure feature map, better reflect the intersection relationship between the virtual survey lines and the joint fissures, and avoid the situation that when the orientation of the joint fissure is similar to the single direction of the virtual survey line, the intersection between the virtual survey line and the joint fissure is less, which is difficult to be detected, resulting in that the RBI index obtained by calculation is not accurate enough.

[0101] S103: For each virtual survey line layout map, based on the rock mass blockiness coefficient corresponding to each virtual survey line in the virtual survey line layout map and the preset angle, the equivalent area of the virtual survey line layout map is determined.

[0102] In the embodiment, each adjacent two virtual survey lines form a measurement region, and the sum of the equivalent areas of each measurement region is the equivalent area of the virtual survey line layout map; and the equivalent area of each measurement region can be obtained by calculating the rock mass blockiness coefficient corresponding to each virtual survey line in the virtual survey line layout map, i.e., the RBI value.

[0103] In the embodiment, the RBI value corresponding to each virtual survey line can be taken as the equivalent length of the virtual survey line, and the equivalent area of the measurement region formed by the adjacent two virtual survey lines can be calculated according to the RBI values corresponding to the adjacent two virtual survey lines and the preset angle between the two virtual survey lines, and then the equivalent area of the entire virtual survey line layout map is obtained. The equivalent area can represent the integrity of the entire tunnel face rock mass, and the larger the equivalent area, the more complete the tunnel face rock mass.

[0104] Specifically, when calculating the equivalent area of each measurement area, based on the rock mass blockiness coefficient and the preset angle corresponding to the two adjacent virtual measuring lines, the area of a triangular region is calculated through the sine theorem, and the area of the triangular region is determined as the equivalent area of the measurement area. Finally, the equivalent areas of the measurement areas are accumulated, which is the equivalent area of the virtual measuring line layout diagram.

[0105] S104: Determine the rock mass integrity of the working face rock mass based on the equivalent areas corresponding to the plurality of virtual measuring line layout diagrams. r3 In this embodiment, for each virtual measuring line layout diagram, based on the ratio of the equivalent area of the virtual measuring line layout diagram to the preset working face area, the initial evaluation index corresponding to the virtual measuring line layout diagram is determined, that is, different radiation centers will correspond to different initial evaluation indexes.

[0107] It should be noted that since when the complete core length is greater than 100 cm, it is a typical whole block structure. Therefore, the area of a semicircle with a radius of 100 cm can be used as the preset working face area.

[0108] In this embodiment, since there are multiple virtual measuring line layout diagrams, the working face rock mass needs to be comprehensively evaluated through multiple virtual measuring line layout diagrams, therefore, after obtaining the initial evaluation index of each virtual measuring line layout diagram, the average value of the initial evaluation indexes of all virtual measuring line layout diagrams is calculated, and then the rock mass integrity comprehensive quantitative index of the working face rock mass is obtained, which is represented by I-RBI. The I-RBI value is the average value of the initial evaluation indexes under the condition of each measuring line radiation center.

[0109] In this embodiment, after obtaining the I-RBI of the working face rock mass, the rock mass integrity of the working face rock mass can be determined based on the mapping relationship between the rock mass integrity comprehensive quantitative index and the rock mass integrity type.

[0110] It should be noted that when the rock mass integrity index K v The following relationship exists between RBI and K

[0111] K v = 0.187RBI 0.403 (2).

[0112] Where K v is the rock mass integrity index, which is an index in the specification for measuring the degree of fragmentation of rock mass, and is defined as the square of the ratio of the elastic longitudinal wave velocity of rock mass to the elastic longitudinal wave velocity of rock.

[0113] The following relationship exists between I-RBI and RBI:

[0114]

[0115] On the basis of the correlation relationship between K v and RBI, combined with the correlation relationship between RBI and I-RBI, a conversion expression of the rock mass integrity index Kv and the I-RBI value can be obtained:

[0116] K v = 1.196 (I_RBI) 0.201 (4);

[0117] In the present embodiment, according to the K v correspondence relationship of rock mass integrity degree (see Table 2) proposed by the Engineering Rock Mass Classification Standard (GB50218-2014), the rock mass integrity degree type and RBI value correspondence relationship (see Table 1), and the formulas (1)-(3), the rock mass integrity degree type and the comprehensive quantitative index I-RBI value of the rock mass integrity degree can be obtained (see Table 3).

[0118] Table 2 Rock mass integrity degree and K v correspondence table

[0119] Rock mass integrity Kv Intact >0.75 Relatively intact 0.55～0.75 Relatively broken 0.35～0.55 Broken 0.15～0.35 Extremely broken <0.15

[0120] Table 3 Rock mass integrity degree and I-RBI correspondence table

[0121] Rock mass integrity I-RBI Intact >9.82 x 10 -2 ]]> Relatively intact 2.10 x 10 -2 ~ 9.82 x 10 -2 ]]> Relatively broken 2.22 x 10 -2 ~ 2.10 x 10 -2 ]]> Broken 3.27 x 10 -5 ~ 2.22 x 10 -3 ]]> Extremely broken <3.27 x 10 -5 ]]>

[0122] In the present embodiment, after the I-RBI is calculated, the corresponding rock mass integrity degree of the tunnel face can be automatically matched by looking up the table.

[0123] In the present embodiment, the pre-processed tunnel face rock mass image is subjected to feature extraction by the joint fissure extraction model, so that more tunnel face joint fissure information can be obtained more quickly and accurately. Meanwhile, by taking different preset positions in the joint fissure feature map as the radiation center, a plurality of virtual line layout maps are drawn, so that the virtual lines are more evenly distributed on the joint fissure feature map, the line density of each part in the joint fissure feature map is encrypted, the intersection relationship between the virtual lines and the joint fissures can be better reflected, and finally the rock mass integrity degree of the tunnel face rock mass is determined according to the equivalent area corresponding to each virtual line layout map, so as to realize the fine quantitative evaluation of the tunnel face surrounding rock structure and provide more accurate data reference for tunnel design, construction, operation and maintenance.

[0124] In a feasible implementation, to make the calculated I-RBI more accurately reflect the true degree of intactness of the rock mass of the tunnel face, a modified joint fracture extraction model (hereinafter referred to as the DeepIntactness model) is used to extract the joint fractures of the rock mass of the tunnel face.

[0125] It should be noted that, although the traditional joint fracture extraction method based on deep learning can greatly improve the recognition efficiency compared with manual extraction, it is difficult to achieve the ideal extraction effect due to the complexity of the tunnel construction environment and the characteristics of the rock mass of the tunnel face. The difficulties are specifically embodied in the following aspects: (1) The tunnel construction situation is complex, and the image of the rock mass of the tunnel face is often disturbed by factors such as dust, uneven illumination, and shadow obstruction, making it difficult to extract joint fractures; (2) Unlike the cracks on the concrete surface, which have a relatively uniform background, the rock mass of the tunnel face has a complex structure, and the joint fractures on the tunnel face are dense, which requires a high crack detection algorithm, and the color change of different rock masses will greatly hinder the extraction of joint fractures; (3) The tunnel face has a large area and rich joint fractures, which requires a large receptive field while paying attention to the pixel-scale details.

[0126] The inventors of the present application found that when extracting features from the image of the rock mass of the tunnel face through the joint fracture extraction model, more detailed information can be extracted by cropping the image of the rock mass of the tunnel face into a smaller sub-image and then extracting features from the sub-image. However, this extraction method also has the problem that the smaller the cropping size, the more discontinuous the joint fracture extraction result. Referring to Figure 3 and Figure 4 respectively show the initial joint fracture feature map obtained when the cropping size is 32x32 pixels and the initial joint fracture feature map obtained when the cropping size is 256x256 pixels, based on Figure 3 and Figure 4 It can be seen that when the cropping size is set to 32x32 pixels, more detailed information can be extracted, but the main joint fractures are not continuous; when the cropping size is set to 256x256 pixels, the main joint fractures are continuous, but some detailed information is missing.

[0127] To solve the above problems and balance the detailed richness of the joint fracture extraction and the continuity of the main joint fractures to achieve better joint fracture feature extraction effect, in the present embodiment, the image of the rock mass of the tunnel face is cropped according to different cropping sizes, and the initial joint fracture feature maps under different cropping sizes are extracted, and then the initial joint fracture feature maps extracted under different cropping sizes are superimposed to obtain the final required joint fracture feature map. Specifically, S101 can include the following sub-steps:

[0128] S101-1: Based on n different cropping sizes, the rock mass image of the working face is cropped to obtain n sub-group images, each sub-group image including a plurality of first sub-images under the corresponding cropping size; n is an integer greater than or equal to 1.

[0129] In the embodiment, the cropping size can be set to 5, which are 32x32 pixels, 64x64 pixels, 128x128 pixels, 256x256 pixels, and 512x256 pixels. Based on the 5 cropping sizes, the rock mass image of the working face is cropped to obtain 5 sub-group images, which can be sequentially set as the first sub-group image, the second sub-group image, the third sub-group image, the fourth sub-group image, and the fifth sub-group image from small to large, wherein the first sub-group image is composed of first sub-images with a size of 32x32 pixels; the second sub-group image is composed of first sub-images with a size of 64x64 pixels; the third sub-group image is composed of first sub-images with a size of 128x128 pixels; the fourth sub-group image is composed of first sub-images with a size of 128x128 pixels; and the fifth sub-group image is composed of first sub-images with a size of 512x256 pixels.

[0130] It should be noted that the number of types of cropping sizes and the specific size of each cropping size can be set according to actual efficiency requirements and computer hardware level, and the present embodiment does not make specific limitations.

[0131] S101-2: For any sub-group image, the first sub-image in the sub-group image is subjected to image equalization processing to obtain a second sub-image; the second sub-image is input into a pre-trained joint fissure extraction model to output an initial joint fissure feature sub-image; and the initial joint fissure feature sub-images are merged to obtain an initial joint fissure feature map corresponding to the sub-group image.

[0132] In the embodiment, compared with directly performing image equalization on the original image, the processing manner of the present embodiment focuses on local pictures, and uses the principle that the factors affecting the gray value of local pictures are similar to perform targeted image enhancement on the first sub-images under each cropping size, which is conducive to highlighting the joint fissure information of the working face, so that the DeepIntactness model can extract more detailed features of the joint fissure, and thus improve the extraction quality.

[0133] S101-3: Based on the initial joint fissure feature maps corresponding to the n sub-group images, a joint fissure feature map is obtained.

[0134] In the embodiment, the initial joint fissure feature maps extracted under each cropping size are added in the corresponding pixel positions to obtain the final joint fissure feature map.

[0135] In the embodiment, with reference to Figure 10, shows the joint fissure feature map extracted by the DeepIntactness model, which is the final effect map obtained based on the initial joint fissure feature maps corresponding to the five sub-group respectively. By comparing Figure 3 、 Figure 4 and Figure 10 , it can be seen that by pixel superposition on the initial joint fissure feature maps corresponding to the five sub-group respectively, more extraction lines reflecting the subtle joint fissure can be extracted while the continuity of the main joint fissure is ensured, thereby realizing better joint fissure feature extraction effect and more comprehensively reflecting the actual condition of the working face.

[0136] In a feasible implementation, the step of performing image equalization processing on the first sub-graph in the sub-group in S101-2 to obtain the second sub-graph can specifically include the following sub-steps:

[0137] S101-2-1: Perform gray processing on the first sub-graph, and obtain the distribution frequency corresponding to each original gray value in the first sub-graph.

[0138] In this embodiment, the original first sub-graph is first processed by gray processing, such as quantizing the gray value with continuous black-gray-white change into 256 gray levels, and the gray value ranges from 0 to 255, indicating that the brightness ranges from deep to shallow, and the color in the image ranges from black to white; after obtaining the initial gray value of each pixel point in the first sub-graph, the distribution frequency corresponding to each original gray value in the first sub-graph can be obtained, which is the gray histogram of the first sub-graph.

[0139] In an example, referring to Figure 5 , a first sub-graph schematic diagram in the image equalization processing process is shown, assuming that the gray level range of the first sub-graph is [0, 9], and it can be seen that the total number of pixels of the first sub-graph is N=5*5=25, and the gray histogram n k of the first sub-graph corresponding to the first sub-graph is [3, 2, 4, 4, 1, 1, 4, 1, 2, 3]. The distribution frequency p r (k) of each original gray value corresponding to each original gray value is n k / N=[3 / 25, 2 / 25, 4 / 25, 4 / 25, 1 / 25, 1 / 25, 4 / 25, 1 / 25, 2 / 25, 3 / 25].

[0140] S101-2-2: Calculate the gray cumulative distribution frequency based on the distribution frequency corresponding to each original gray value.

[0141] Continuing to refer to the above example, the gray cumulative distribution frequency

[0142] S101-2-3: obtaining a target gray value based on the gray cumulative distribution frequency and the gray level of the first subgraph.

[0143] In the embodiment, the normalized s k is multiplied by L-1 and rounded to make the gray level of the image after equalization consistent with the first sub Figure 1 graph before normalization, and a target gray value of the image after equalization is obtained, where L is the gray level of the first subgraph.

[0144] For example, s0=3 / 25*(10-1)=1.08, and after rounding, its value is 1, that is, the gray level 0 in the original first subgraph corresponds to the gray level 1 after equalization, i.e. 0→1; s1=5 / 25*(10-1)=1.8, and after rounding, its value is 2, i.e. 1→2; and so on, until s9=9, and after rounding, its value is 9, i.e. 9→9.

[0145] S101-2-4: obtaining the second subgraph after image equalization based on the target gray value.

[0146] In the embodiment, with reference to Figure 6 , based on the mapping relationship above, the second subgraph after image equalization can be obtained.

[0147] It should be noted that image equalization can effectively enhance the contrast of the image, for example Figure 5 In the original image, the number of pixels with gray values of 4, 5, and 7 is 1, so in Figure 6 , these three pixels are merged into adjacent gray values, and because there are three gray values merged, three empty spaces appear after equalization, and the original adjacent gray values are expanded by these empty spaces. For example, 5 and 6 are adjacent, and after equalization, they become 5 and 7 adjacent, thus expanding the contrast.

[0148] In the embodiment, by performing histogram equalization on the first subgraph under each cropping size, a second subgraph with stronger contrast can be obtained, and inputting the second subgraph into the trained extraction model can make the extraction result reflect more information of the joint fissure of the subgraph, and be closer to the real situation.

[0149] In a feasible embodiment, to make the joint fissure extraction model, i.e. the DeepIntactness model, better extract the initial joint fissure feature subgraph, the structure of the DeepIntactness model is optimized.

[0150] Specifically, the DeepIntactness model can be constructed based on a DeepCrack network. It should be noted that the DeepCrack network is essentially a UNet network, but unlike the UNet network, the DeepCrack network uses skip-layer fusion to connect corresponding layers in the encoding path and the decoding path, but only obtains one loss value. The DeepCrack network combines the deep supervision technique (Deep Supervision) and supervises the loss values at each scale (Stage), and adds the loss values at each scale to obtain the total loss value.

[0151] To improve the accuracy of the network, the network depth and the number of training times are often required to increase, but due to the small number of fracture labeling maps of the rock mass of the working face, the requirement of massive data training may cause overfitting problem. To solve this problem, the DeepIntactness model integrates the design concept of the residual network ResNet based on the DeepCrack network, and uses the shortcut connection method to connect the feature maps output by the previous scale at each scale in the DeepIntactness encoding path. After dimensionality increasing and pooling processing by a 1x1 convolution kernel, the feature maps are elementally superimposed with the feature maps processed by the backbone network, further optimizing the encoding network structure.

[0152] Referring to Figure 7 , a structural schematic diagram of the DeepIntactness model is shown, including a symmetric encoding path and a decoding path, wherein the encoding path includes five scale encoding modules, the decoding path includes decoding modules corresponding to the encoding modules one by one, and the encoding modules and the decoding modules at each scale are connected by skip fusion, and the loss values at each scale are added to obtain the total loss value.

[0153] In the embodiment, based on the DeepIntactness model optimized in structure, the step of inputting the second subgraph into the pre-trained joint fracture extraction model in S101-2 to output an initial joint fracture feature subgraph can specifically include the following substeps:

[0154] S101-2-5: input the second subgraph into the encoding path, and use a preset number of encoding modules to extract the feature information of the second subgraph layer by layer from shallow to deep to obtain an encoding feature map; wherein each encoding module includes an encoding network and a shortcut network connected with the encoding network by a shortcut connection, the shortcut network is used to superimpose the feature map output by the previous encoding module on the feature map output by the encoding network of the current encoding module, and the superimposed feature map is the output of the current encoding module.

[0155] S101-2-6: input the encoded feature map into the decoding path, and decode the encoded feature map layer by layer using the decoding module corresponding to the encoding module to output the initial joint fissure feature subgraph.

[0156] In the embodiment, the element of the shallow layer feature map output by the skip layer network and the deep layer feature map output by the corresponding encoding module is superimposed, so that the data before processing is superimposed on the information after processing, which can effectively avoid the loss of many details after obtaining deep features.

[0157] Next, to verify the feasibility of the DeepIntactness model, the following experiments will be conducted:

[0158] Taking the image of the working face at ZK92+850 of the Bimo Garden Tunnel of the Leshan-Xichang Expressway as an example, the working face rock mass integrity quantification system is studied, and the original image collected is as shown in Figure 8

[0159] The collected digital image of the working face at ZK92+850 has a pixel of 2304 pixels*1728 pixels, which is divided into 324 first subgraphs (pictures less than 128 pixels are supplemented with white) according to the size of 128 pixels*128 pixels. The obtained 32 first subgraphs are subjected to gray processing and histogram equalization respectively, and the second subgraph after equalization is obtained. The second subgraph after equalization is input into the trained DeepIntactness model, and the corresponding working face trace extraction graph of each subgraph is predicted. The subgraph extraction graph is rearranged and combined to restore and clip into a size of 2304 pixels*1728 pixels.

[0160] Referring to Figure 9 and Figure 10 , the manually extracted joint fissure feature graph and the joint fissure feature graph extracted by the DeepIntactness model are shown respectively. The manually extracted joint fissure feature graph only contains the main joint fissure, while the working face joint fissure result automatically extracted by the DeepIntactness model contains extraction lines reflecting the fine joint fissure, which can comprehensively reflect the actual situation of the working face.

[0161] Referring to Figure 11 , a rectangular coordinate system is established with the intersection point of the tunnel bottom plate intersection line and the left wall as the origin, and the virtual measuring line layout graphs corresponding to the four preset positions of the vault, the working face center, the left arch foot and the right arch foot are shown respectively.

[0162] ​The sampling frequency of the virtual line segment formed by the virtual line of each radio center being cut by the trace of the structural plane is calculated using formula (1) in five intervals of [3cm, 10cm], [10cm, 30cm], [30cm, 50cm], [50cm, 100cm] and [100cm, +∞], and the RBI value of each line is calculated, and the calculation results are shown in Table 4. The RBI index rose diagram of the rock mass integrity of the working face is shown in Figure 12 .

[0163] Table 4 RBI calculation table of the rock mass of the working face at ZK92+850

[0164]

[0165]

[0166]

[0167]

[0168]

[0169] It should be noted that, Figure 12 Fig. (a) is the RBI rose diagram of the virtual line drawn with the vault as the radio center, Fig. (b) is the RBI rose diagram of the virtual line drawn with the center of the working face as the radio center, Fig. (c) is the RBI rose diagram of the virtual line drawn with the left springing as the radio center, and Fig. (d) is the RBI rose diagram of the virtual line drawn with the right springing as the radio center.

[0170] From Figure 12 As can be seen from Fig. (a), taking the vault as the starting point, the RBI value in the direction of 90° is the smallest, and the ratio of the length of the virtual line in this direction is the smallest, so when reinforcing the vault, it is considered to reinforce the surrounding rock in the direction of 90°. And the overall RBI value of the right part of the working face is larger, which proves that the right part of the working face has better integrity. As can be seen from Fig. (b), in the directions of 90°-100° and 270°-280°, the RBI values are both low, which also proves that rock anchoring support in the direction perpendicular to the vault is needed. By combining each line in the range of 0°-170° with each virtual line in the range of 180°-350° according to whether they are collinear, it can be found that the maximum RBI value appears in the range of 10°(190°)-30°(210°), and the minimum RBI value appears in the range of 90°(270°)-110°(290°), with an included angle of about 90°, which proves that the rock mass in the range of 10°(190°)-30°(210°) is stratified, and the main distribution part is in the 10°(190°)-30°(210°) range of the zone with the center of the working face as the center.

[0171] The I-RBI calculation results are shown in Table 5:

[0172] Table 5 ZK92+850 face rock mass I-RBI calculation table

[0173]

[0174]

[0175]

[0176] As shown in Table 5 above, when the face center is taken as the radiation center, the I-RBI value is 0.0154, which belongs to relatively broken; when the face vault is taken as the radiation center, the I-RBI value is 0.0154, which belongs to relatively broken; when the left and right arch feet of the face are taken as the radiation center, the I-RBI values are 0.0172 and 0.0153 respectively, which belong to relatively broken. The final I-RBI value is the average value of the I-RBI values calculated for each measuring line center:

[0177]

[0178] According to Table 3, it can be found that the completeness degree type of the face at ZK92+850 of the Laxi highway Bimoyuan tunnel mileage stake is relatively broken. It can be known from the face original drawing that the grading result obtained in the embodiment is consistent with the grading method result advocated by the specification, which proves the reliability of the conclusion.

[0179] In the embodiment, in combination with image processing technology, histogram equalization is proposed to process the rock mass image of the tunnel face in different sizes. The rock mass image of the tunnel face is cropped in different sizes, and histogram equalization is performed respectively, which can avoid the problem that the joint fissure of the tunnel face is difficult to accurately extract due to uneven illumination, color change of lithology, etc. The second subgraph after histogram equalization processing extracts the joint fissure condition. The extraction effect is different due to different cropping sizes. The small size subgraph can extract fine joint fissures, and the large size subgraph can ensure the continuity of the significant joint fissure. Therefore, the fusion of the extraction results of each scale subgraph can take into account the fine and macro joint fissure conditions of the tunnel face, and ensure the comprehensiveness of the joint fissure extraction. At the same time, in combination with the coding-decoding neural network and the residual neural network design concept, the DeepIntactness model is proposed to realize the automatic extraction function of the tunnel face joint fissure. The DeepIntactness model uses the deep supervision technique on the basis of the U-shaped structure, supervises the different loss values of the network when extracting shallow and deep features, so that the model can extract features of different depths, and ensure the accuracy of the model in extracting joint fissures. Finally, since the Z-RBI method only considers the center of the tunnel floor as a virtual radiation point of the measuring line, the measuring line of the edge area of the tunnel face which has the greatest impact on the stability of the lining structure is sparse, and it is difficult to accurately reflect the rock mass structure information of the edge area of the tunnel face. The embodiment combines the automatically extracted joint fissure map of the tunnel face, and proposes the I-RBI algorithm, taking the four parts of the tunnel face most likely to appear instability conditions, i.e., the vault, the centroid of the tunnel face, and the left and right arch feet as the virtual radiation centers of the measuring line, to increase the density of the measuring line of different potential instability parts, realize the fine quantitative evaluation of the surrounding rock structure of the tunnel face, and obtain the overall integrity of the rock mass of the tunnel face, the main distribution direction and distribution position of the layered structure rock mass, and the rock mass structure category through the I-RBI values and the RBI rose diagram obtained under multiple radiation centers, to provide targeted reinforcement measures.

[0180] In a second aspect, based on the same inventive concept, referring to Figure 13 , a kind of based on image's tunnel face rock mass integrity evaluation device 1300 provided by the embodiment of the application is shown, and the based on image's tunnel face rock mass integrity evaluation device 1300 includes:

[0181] The feature extraction module 1301 is used to input the preprocessed tunnel face rock mass image into the pre-trained joint fissure extraction model, and output the joint fissure feature map, which is used to represent the fissure information of the tunnel face rock mass.

[0182] The line layout module 1302 is configured to obtain a plurality of different virtual line layout maps based on the joint fissure feature map; wherein the different virtual line layout maps are obtained by drawing virtual lines on the joint fissure feature map with different preset positions in the joint fissure feature map as the radiation centers and a preset angle as the layout interval.

[0183] The equivalent area determination module 1303 is configured to determine, for each virtual line layout map, an equivalent area of the virtual line layout map based on the rock mass blockiness coefficients corresponding to each virtual line in the virtual line layout map and the preset angle.

[0184] The rock mass integrity determination module 1304 is configured to determine the rock mass integrity of the working face rock mass based on the equivalent areas corresponding to the plurality of virtual line layout maps.

[0185] In an implementable embodiment, the preset positions include: a vault, a left arch foot, a right arch foot and a working face centroid; wherein the working face centroid is arranged in a central region of the joint fissure feature map.

[0186] In an implementable embodiment, the equivalent area determination module 1303 comprises:

[0187] The rock mass blockiness coefficient calculation submodule is configured to calculate, for each virtual line layout map, a rock mass blockiness coefficient of each virtual line in the virtual line layout map.

[0188] The first equivalent area determination submodule is configured to determine, based on the rock mass blockiness coefficients corresponding to two adjacent virtual lines and the preset angle, an equivalent area of a measurement region corresponding to the two adjacent virtual lines.

[0189] The second equivalent area determination submodule is configured to determine, based on the equivalent areas of each measurement region, the equivalent area of the virtual line layout map.

[0190] In an implementable embodiment, the first equivalent area determination submodule comprises:

[0191] The sine theorem calculation unit is configured to calculate, based on the rock mass blockiness coefficients corresponding to two adjacent virtual lines and the preset angle, a triangular region area by the sine theorem.

[0192] The equivalent area determination unit is configured to determine the triangular region area as the equivalent area of the measurement region corresponding to the two adjacent virtual lines.

[0193] In an implementable embodiment, the rock mass integrity determination module 1304 comprises:

[0194] The initial evaluation index determination submodule is configured to determine, for each virtual survey line layout, an initial evaluation index corresponding to the virtual survey line layout based on a ratio of an equivalent area of the virtual survey line layout to a preset working face area.

[0195] The rock mass integrity comprehensive quantification index determination submodule is configured to determine, as a rock mass integrity comprehensive quantification index of the working face rock mass, an average value of the initial evaluation indexes of all the virtual survey line layouts.

[0196] The rock mass integrity determination submodule is configured to determine, based on a mapping relationship between the rock mass integrity comprehensive quantification index and a rock mass integrity type, a rock mass integrity of the working face rock mass.

[0197] In a feasible implementation, the feature extraction module 1301 includes:

[0198] The cropping submodule is configured to crop the working face rock mass image based on n different cropping sizes to obtain n sub-group sets, each of which includes a plurality of first sub-pictures under a corresponding cropping size; n is an integer greater than or equal to 1.

[0199] The initial joint fissure feature map acquisition submodule is configured to, for any sub-group set, perform image equalization processing on the first sub-pictures in the sub-group set to obtain second sub-pictures; input the second sub-pictures into a pre-trained joint fissure extraction model to output an initial joint fissure feature sub-picture; and merge the initial joint fissure feature sub-pictures to obtain an initial joint fissure feature map corresponding to the sub-group set.

[0200] The joint fissure feature map acquisition submodule is configured to obtain a joint fissure feature map based on the initial joint fissure feature maps corresponding to the n sub-group sets.

[0201] In a feasible implementation, the image equalization processing submodule includes:

[0202] The distribution frequency acquisition unit is configured to perform grayscale processing on the first sub-picture and acquire a distribution frequency corresponding to each original grayscale value in the first sub-picture.

[0203] The cumulative distribution frequency unit is configured to calculate a grayscale cumulative distribution frequency based on the distribution frequency corresponding to each original grayscale value.

[0204] The target grayscale value acquisition unit is configured to obtain a target grayscale value based on the grayscale cumulative distribution frequency and a grayscale level of the first sub-picture.

[0205] The second sub-picture acquisition unit is configured to obtain a second sub-picture after image equalization based on the target grayscale value.

[0206] In one possible implementation, the joint fracture extraction model comprises a symmetrical encoding path and a decoding path; a subgraph output submodule comprising:

[0207] An encoding unit is configured to input the second subgraph into the encoding path, and extract feature information of the second subgraph layer by layer from shallow to deep by using a preset number of encoding modules, to obtain an encoded feature map; each encoding module comprises an encoding network and a skip layer network connected to the encoding network in a skip layer manner, and the skip layer network is configured to superimpose the feature map output by the previous encoding module on the feature map output by the encoding network of the current encoding module, and the superimposed feature map is the output of the current encoding module;

[0208] A decoding unit is configured to input the encoded feature map into the decoding path, and perform layer-by-layer decoding on the encoded feature map by using a decoding module corresponding to each encoding module, to output an initial joint fracture feature subgraph.

[0209] It should be noted that the specific implementation of the image-based rock mass integrity evaluation device 1300 of the working face in the embodiment is described with reference to the specific implementation of the image-based rock mass integrity evaluation method of the first aspect of the application, which will not be described here.

[0210] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, device, or computer program product. Therefore, the embodiments of the present application can be in the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the embodiments of the present application can be in the form of a computer program product implemented on one or more computer usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer usable program code.

[0211] The embodiments of the present application are described with reference to the flowcharts and / or block diagrams according to the method, terminal device (system), and computer program product of the embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of the flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing terminal device to produce a machine, so that the instructions executed by the computer or other programmable data processing terminal device produce a device that implements the functions specified in the flowcharts and / or block diagrams. Figure 1 The device that implements the functions specified in one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0212] These computer program instructions can also be stored in a computer- readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the flow Figure 1 one or more flow or block Figure 1 one or more flow or block

[0213] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the flow Figure 1 one or more flow or block Figure 1 one or more flow or block

[0214] Although preferred embodiments of the present application have been described, those skilled in the art will be able to make additional modifications and variations to these embodiments without departing from the scope of the present application. Accordingly, the appended claims are intended to encompass all such modifications and variations as falling within the scope of the present application.

[0215] Finally, it should be noted that the terms "first" and "second" and the like are used merely to distinguish one entity or action from another, without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variation thereof are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus. An element proceeded by "comprises... a" does not, without more constraints, exclude the existence of additional identical elements in the process, method, article, or apparatus that comprises the element.

[0216] The above provides a kind of based on image's face of rock mass integrity degree evaluation method and device of the present application, has carried out detailed introduction, the principle and implementation mode of the present application are described in this paper with specific example, the above example is only for helping understanding the method of the present application and its core idea;For the general technical personnel of the field, according to the idea of the present application, there will be changes in specific implementation mode and application range, as described above, the content of the specification should not be understood as the limitation of the present application.

Claims

1. An image-based evaluation method for the integrity of a rock mass at a tunnel face, characterized in that The method comprises: inputting the preprocessed tunnel face rock mass image into a pre-trained joint fracture extraction model to output a joint fracture feature map, the joint fracture feature map being used to represent the fracture information of the tunnel face rock mass, comprising: based on n different cropping sizes, cropping the tunnel face rock mass image to obtain n subgraph groups, each subgraph group comprising a plurality of first subgraphs under the corresponding cropping size; n is an integer greater than or equal to 1; for any subgraph group, performing image equalization processing on the first subgraphs in the subgraph group to obtain second subgraphs; inputting the second subgraphs into a pre-trained joint fracture extraction model to output initial joint fracture feature subgraphs; merging the initial joint fracture feature subgraphs to obtain the initial joint fracture feature map corresponding to the subgraph group; based on the initial joint fracture feature maps corresponding to the n subgraph groups, obtaining the joint fracture feature map; based on the joint fracture feature map, obtaining a plurality of different virtual line layout maps; wherein a different virtual line layout map is obtained by drawing virtual lines on the joint fracture feature map with different preset positions in the joint fracture feature map as the radiation center and a preset angle as the layout interval; for each virtual line layout map, determining the equivalent area of the virtual line layout map based on the rock mass blockiness coefficient corresponding to each virtual line in the virtual line layout map and the preset angle; based on the equivalent areas corresponding to the plurality of virtual line layout maps, determining the rock mass integrity of the tunnel face rock mass; wherein the image equalization processing on the first subgraphs in the subgraph group to obtain second subgraphs comprises: performing grayscale processing on the first subgraphs and obtaining the distribution frequency of each original grayscale value in the first subgraphs; based on the distribution frequency of each original grayscale value, calculating the grayscale cumulative distribution frequency; based on the grayscale cumulative distribution frequency and the grayscale level of the first subgraphs, obtaining a target grayscale value; based on the target grayscale value, obtaining the second subgraphs after image equalization; the joint fracture extraction model comprises a symmetrical encoding path and a decoding path; inputting the second subgraphs into a pre-trained joint fracture extraction model to output initial joint fracture feature subgraphs comprises: inputting the second subgraphs into the encoding path and using a preset number of encoding modules to extract the feature information of the second subgraphs layer by layer from shallow to deep to obtain an encoding feature map; wherein each encoding module comprises an encoding network and a skip layer network connected to the encoding network, the skip layer network being used to superimpose the feature map output by the previous encoding module on the feature map output by the encoding network of the current encoding module, and the superimposed feature map being the output of the current encoding module; inputting the encoding feature map into the decoding path and using a decoding module corresponding to each encoding module to decode the encoding feature map layer by layer to output the initial joint fracture feature subgraphs.

2. The method of claim 1, wherein, the preset positions include: vault, left arch foot, right arch foot and tunnel face center; wherein the tunnel face center is arranged in the center region of the joint fracture feature map.

3. The method of claim 1, wherein, The equivalent area of each virtual survey line layout is determined based on the rock mass blockiness coefficient corresponding to each virtual survey line in the virtual survey line layout and the preset angle. The rock mass blockiness coefficient of each virtual survey line in the virtual survey line layout is calculated. The equivalent area of a measurement region corresponding to two adjacent virtual survey lines is determined based on the rock mass blockiness coefficients corresponding to the two adjacent virtual survey lines and the preset angle. The equivalent area of each measurement region is used to determine the equivalent area of the virtual survey line layout.

4. The method of claim 3, wherein, The equivalent area of a measurement region corresponding to two adjacent virtual survey lines is determined based on the rock mass blockiness coefficients corresponding to the two adjacent virtual survey lines and the preset angle. The equivalent area of a measurement region corresponding to two adjacent virtual survey lines is determined based on the rock mass blockiness coefficients corresponding to the two adjacent virtual survey lines and the preset angle. The equivalent area of a measurement region corresponding to two adjacent virtual survey lines is determined based on the rock mass blockiness coefficients corresponding to the two adjacent virtual survey lines and the preset angle.

5. The method of claim 1, wherein, The equivalent area of each virtual survey line layout is determined based on the rock mass blockiness coefficient corresponding to each virtual survey line in the virtual survey line layout and the preset angle. The initial evaluation index corresponding to each virtual survey line layout is determined based on the ratio of the equivalent area of the virtual survey line layout to the preset tunnel face area. The average value of the initial evaluation indexes of all virtual survey line layouts is determined as the rock mass integrity comprehensive quantitative index of the tunnel face rock mass. The rock mass integrity of the tunnel face rock mass is determined based on the mapping relationship between the rock mass integrity comprehensive quantitative index and the rock mass integrity type.

6. An image-based evaluation device for the completeness of a rock mass at a tunnel face, characterized by The device comprises: The feature extraction module is configured to input the preprocessed tunnel face rock mass image into a pre-trained joint fissure extraction model to output a joint fissure feature map, wherein the joint fissure feature map is used to represent the fissure information of the tunnel face rock mass, and the joint fissure feature map comprises: The tunnel face rock mass image is cropped based on n different cropping sizes to obtain n sub-group images, wherein each sub-group image comprises a plurality of first sub-images corresponding to the cropping size; n is an integer greater than or equal to 1; For any sub-group image, the first sub-image in the sub-group image is subjected to image equalization processing to obtain a second sub-image; the second sub-image is input into the pre-trained joint fissure extraction model to output an initial joint fissure feature sub-image; and the initial joint fissure feature sub-images are merged to obtain an initial joint fissure feature map corresponding to the sub-group image; The joint fissure feature map is obtained based on the initial joint fissure feature maps corresponding to the n sub-group images; The survey line layout module is configured to obtain a plurality of different virtual survey line layouts based on the joint fissure feature map, wherein different virtual survey line layouts are obtained by drawing virtual survey lines on the joint fissure feature map with different preset positions in the joint fissure feature map as the center and a preset angle as the layout interval; The equivalent area determination module is configured to determine the equivalent area of each virtual survey line layout based on the rock mass blockiness coefficient corresponding to each virtual survey line in the virtual survey line layout and the preset angle. The rock mass integrity degree determination module is configured to determine the rock mass integrity degree of the working face rock mass based on the equivalent areas corresponding to the multiple virtual line layout maps. The first subgraph in the subgraph group is subjected to image equalization processing to obtain a second subgraph, including: The first subgraph is subjected to gray scale processing, and a distribution frequency corresponding to each original gray scale value in the first subgraph is obtained; Based on the distribution frequency corresponding to each original gray scale value, a gray scale cumulative distribution frequency is calculated; Based on the gray scale cumulative distribution frequency and the gray scale level of the first subgraph, a target gray scale value is obtained; Based on the target gray scale value, the second subgraph subjected to image equalization is obtained; The joint fissure extraction model includes a symmetrical encoding path and a decoding path; the second subgraph is input into the pre-trained joint fissure extraction model to output an initial joint fissure feature subgraph, including: The second subgraph is input into the encoding path, and a preset number of encoding modules are used to extract feature information of the second subgraph from shallow to deep layer by layer to obtain an encoding feature map; each encoding module includes an encoding network and a skip layer network connected to the encoding network in a skip layer manner, and the skip layer network is configured to superimpose a feature map output by a previous encoding module on a feature map output by the encoding network of a current encoding module, and the superimposed feature map is the output of the current encoding module; The encoding feature map is input into the decoding path, and each decoding module corresponding to the encoding module is used to decode the encoding feature map layer by layer to output the initial joint fissure feature subgraph.

7. The apparatus of claim 6, wherein, The preset positions include a vault, a left arch foot, a right arch foot and a working face center; the working face center is arranged in a central region of the joint fissure feature map.

Images