A method and electronic device for automatically identifying key geological parameters of a tunnel face

Abstract

The present invention discloses a method for automatically identifying geological key parameters of a tunnel face and an electronic device. The method includes the following steps: constructing a tunnel face classification and segmentation model; the tunnel face classification and segmentation model classifies and segments the tunnel face image to obtain classification results and segmentation result diagrams respectively; calculating the number of structural surface groups and the structural surface spacing of the joint traces according to the segmentation result diagram, obtaining the face development degree result according to the number of structural surface groups and the structural surface spacing, and the face development degree result is used to judge the integrity of the face structure surface; and optimizing measures are proposed for tunnel construction according to the attribute category of the face and the integrity of the face structure surface. The method of the present application extracts the key parameters of the tunnel face to grasp the stability of the tunnel face, realizes real-time monitoring and early warning of the stability of the tunnel face, which can not only improve the monitoring efficiency, but also reduce the cost of manpower and material resources.

Description

Technical Field

[0001] The present invention relates to the field of tunnel construction information technology, and in particular to an automatic identification method for key geological parameters of a tunnel face and electronic equipment. Background Art

[0002] In railway or highway tunnels, blasting is the main excavation method. The tunnel face area formed after blasting is prone to instability due to its high deformation rate, which leads to tunnel collapse. Therefore, it is necessary to extract key parameters of the tunnel face area to understand the stability of the tunnel face in real time.

[0003] Before tunnel excavation and blasting, the stability of the tunnel face needs to be evaluated. The conventional method is to first take photos of the tunnel face with a digital camera. The geological survey personnel will judge the key parameters of the tunnel face based on the photos taken and combined with experience, such as rock hardness, rock integrity, groundwater and other surrounding rock geological information, to form a geological sketch of the tunnel face. However, the problem with this method is that the degree of automation and monitoring efficiency are low, the evaluation differences are large, and the cost is high.

[0004] Therefore, how to extract the key parameters of the tunnel face in order to grasp the stability of the tunnel face, realize real-time monitoring and early warning of the tunnel face stability, improve monitoring efficiency, and reduce manpower and material costs is an important issue in current research. Summary of the invention

[0005] The main technical problem solved by the present invention is to provide a method and electronic equipment for automatically identifying key geological parameters of a tunnel face, so as to solve the problem of how to extract the key parameters of the tunnel face so as to grasp the stability of the tunnel face, realize real-time monitoring and early warning of the stability of the tunnel face, improve monitoring efficiency, and reduce manpower and material costs.

[0006] In order to solve the above technical problems, a technical solution adopted by the present invention is to provide a method for automatically identifying key geological parameters of a tunnel face, comprising the steps of:

[0007] A tunnel face classification and segmentation model is constructed; the tunnel face classification and segmentation model classifies and segments the tunnel face image, and obtains classification results and segmentation result maps respectively. The classification results are used to identify the attribute categories of the face; the number of structural surface groups and the structural surface spacing of the joint traces are calculated according to the segmentation result map, and the face development degree results are obtained according to the number of structural surface groups and the structural surface spacing. The face development degree results are used to judge the integrity of the face structure surface; optimization measures for tunnel construction are proposed according to the attribute category of the face and the integrity of the face structure surface.

[0008] In some embodiments, the tunnel face classification and segmentation model includes: a backbone network, a trace segmentation network, and an attribute classification network, wherein the trace segmentation network includes a fusion unit and a detection unit.

[0009] The input end of the backbone network is used to input the tunnel face image. The backbone network is used to extract features from the tunnel face image to obtain a feature map, and then input the feature map into the trace segmentation network and the attribute classification network respectively.

[0010] The attribute classification network receives the feature map, performs feature calculations on the feature map, and outputs the classification results.

[0011] The fusion unit of the trace segmentation network receives the feature map and performs feature fusion on the feature map to obtain a fusion map.

[0012] The detection unit of the trace segmentation network is used to receive the fusion image, detect the fusion image, and output the segmentation result image.

[0013] In some embodiments, the tunnel face classification and segmentation model classifies and segments the tunnel face image to obtain a classification result and a segmentation result map, respectively. The classification result is used to identify the attribute category of the face. The method also includes: performing data enhancement processing on the tunnel face image with a small number of samples in the attribute category to obtain an expanded image set, and inputting the expanded image set and the tunnel face image into the tunnel face classification and segmentation model.

[0014] In some embodiments, the number of structural surface groups and the structural surface spacing of the joint trace are calculated according to the segmentation result graph, and the method is:

[0015] The trace attributes of the joint traces are calculated according to the segmentation result diagram, and the number of structural surface groups and the distance between structural surfaces are calculated according to the trace attributes.

[0016] In some embodiments, the method for calculating the trace attributes of the joint traces according to the segmentation result graph is:

[0017] Preprocess the segmentation result map; solve and record the connected domain according to the preprocessed segmentation result map; extract the trace skeleton according to the connected domain to obtain the skeleton map; perform line segment fitting on the skeleton map to calculate the length and angle of the joint trace, thereby obtaining the trace property of the joint trace.

[0018] In some embodiments, the number of structure face groups and the structure face spacing are calculated based on the trace attributes by:

[0019] Cluster the angles of joint traces to obtain clustering results, and group the structural surfaces according to the clustering results to obtain the number of structural surface groups; calculate the normal slope of the joint traces; select the reference survey line; calculate the average distance of the joint traces intercepted by the reference survey line to obtain the structural surface spacing.

[0020] In some embodiments, the total loss function of the tunnel face classification and segmentation model is calculated as follows:

[0021] ;

[0022] in, represents the classification loss function, represents the bounding box loss function, represents the segmentation loss function, , , , are the weight coefficients corresponding to different loss functions, Represents the loss function corresponding to different attribute categories.

[0023] In some embodiments, the classification loss function is calculated as:

[0024] ;

[0025] in, represents the number of samples, represents the number of categories, Indicates The samples belong to The true category label of the class attribute, Indicates The samples belong to The predicted probability of the class attribute.

[0026] In some embodiments, the bounding box loss function is calculated as:

[0027] ;

[0028] ;

[0029] ;

[0030] in, represents the complete intersection-over-union loss function, represents the distribution focusing loss function, Represents a bounding box and bounding box The center point distance, Represents a bounding box and bounding box The diagonal length of the minimum bounding box of Represents a bounding box and bounding box The aspect ratio similarity of for The impact factor, represents the center point of the prediction box, represents the center point of the real frame, represents the regression label, and To be closest The two values ​​of express The probability of passing sotfmax is, express The probability of passing sotfmax.

[0031] In order to solve the above technical problems, the present application also provides an electronic device, including a processor and a memory, wherein the memory stores a computer program that can be executed by the processor, and when the computer program is executed by the processor, it implements any of the above-mentioned methods for automatically identifying key geological parameters of a tunnel face.

[0032] Beneficial effects: The present invention discloses a method and electronic equipment for automatically identifying geological key parameters of a tunnel face, the method comprising the steps of: constructing a tunnel face classification and segmentation model; the tunnel face classification and segmentation model classifies and segments the tunnel face image, and obtains classification results and segmentation result diagrams respectively; the number of structural surface groups and the structural surface spacing of the joint traces are calculated according to the segmentation result diagram, and the face development degree results are obtained according to the number of structural surface groups and the structural surface spacing, and the face development degree results are used to judge the integrity of the face structure surface; optimization measures are proposed for tunnel construction according to the attribute category of the face and the integrity of the face structure surface. The method of the present application extracts the key parameters of the tunnel face to grasp the stability of the tunnel face, and realizes real-time monitoring and early warning of the stability of the tunnel face, which can not only improve the monitoring efficiency, but also reduce the cost of manpower and material resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 It is a flow chart of an embodiment of a method for automatically identifying key geological parameters of a tunnel face according to the present invention;

[0034] Figure 2 is a structural schematic diagram of a tunnel face classification and segmentation model according to the present invention;

[0035] Figure 3 The images of tunnel faces with different weathering degrees are given;

[0036] Figure 4 The tunnel face images with different underground water discharge conditions are given;

[0037] Figure 5 An example of a mask image obtained from a segmentation result image is given;

[0038] Figure 6 Yes Figure 5 The image obtained after binarization processing;

[0039] Figure 7 is based on Figure 6 The obtained skeleton graph;

[0040] Figure 8 It is a schematic structural diagram of an electronic device according to the present invention. DETAILED DESCRIPTION

[0041] In order to facilitate the understanding of the present invention, the present invention is described in more detail below in conjunction with the accompanying drawings and specific embodiments. Preferred embodiments of the present invention are provided in the accompanying drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described in this specification. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present invention more thorough and comprehensive.

[0042] It should be noted that, unless otherwise defined, all technical and scientific terms used in this specification have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used in this specification includes any and all combinations of one or more related listed items.

[0043] It should be noted that the "tunnel face" referred to in this application refers to the excavation working face in underground engineering or mining engineering. Joint traces refer to the cracks in the rock without obvious relative displacement on both sides of the crack surface (as opposed to faults with obvious displacement), which can also be called traces of geological fractures.

[0044] Figure 1 A flow chart showing an embodiment of a method for automatically identifying key geological parameters of a tunnel face of the present invention includes the following steps:

[0045] Step S1, constructing a tunnel face classification and segmentation model.

[0046] Step S2, the tunnel face classification and segmentation model classifies and segments the tunnel face image to obtain a classification result and a segmentation result map, respectively. The classification result is used to identify the attribute category of the tunnel face.

[0047] Step S3, calculating the number of structural surface groups and the distance between structural surfaces of the joint traces according to the segmentation result diagram, and obtaining the tunnel face development degree result according to the number of structural surface groups and the distance between structural surfaces, and the tunnel face development degree result is used to judge the integrity of the tunnel face structural surface.

[0048] Step S4, proposing optimization measures for tunnel construction according to the attribute category of the tunnel face and the integrity of the tunnel face structure surface.

[0049] The present application extracts the key parameters of the tunnel face through a tunnel face classification and segmentation model, wherein the key parameters include the attribute category of the face and the number of structural face groups and the structural face spacing of the joint trace. The stability of the face can be grasped through the above key parameters, and real-time monitoring and early warning of the stability of the tunnel face can be achieved, thereby improving monitoring efficiency and reducing manpower and material costs.

[0050] In some embodiments, in combination Figure 2 The tunnel face classification and segmentation model includes: a backbone network 1, a trace segmentation network 2 and an attribute classification network 3, wherein the trace segmentation network 2 includes a fusion unit 21 and a detection unit 22.

[0051] The input end of the backbone network 1 is used to input the tunnel face image. The backbone network 1 is used to extract features from the tunnel face image to obtain a feature map, and input the feature map into the trace segmentation network 2 and the attribute classification network 3 respectively.

[0052] The attribute classification network 3 receives the feature map, performs feature calculation on the feature map, and outputs the classification result.

[0053] The fusion unit 21 of the trace segmentation network 2 receives the feature map and performs feature fusion on the feature map to obtain a fusion map.

[0054] The detection unit 22 of the trace segmentation network 2 is used to receive the fusion image, detect the fusion image, and output a segmentation result image.

[0055] Furthermore, the attribute categories of the tunnel face include rock hardness, structural surface integrity, and underground water status. Rock hardness includes rock lithology and weathering degree. The weathering degree is mainly divided into five categories: unweathered, slightly weathered, weakly weathered, strongly weathered or fully weathered, and other (unidentifiable).

[0056] Figure 3 Four tunnel face images with different weathering degrees are given. Figure 3 Figure A in the figure indicates that it is not weathered, the rock structure has not changed, and it is fresh and has no discoloration; Figure 3 Figure B shows slight weathering, with slight changes in rock structure and slight discoloration; Figure 3 Figure C shows weak weathering, partial destruction of rock structure, and obvious discoloration; Figure 3 Figure D shows strong or complete weathering, with most of the rock structure destroyed and the rock color and mineralization obvious.

[0057] Furthermore, underground water outflow states include four types of water outflow states: dry, wet or dripping, rain or linear water outflow, and gushing water outflow.

[0058] Figure 4 Three different tunnel face images with underground water discharge are given, and two tunnel face images are given for each underground water discharge state. Figure 4 E1 and E2 in the figure are both images of tunnel faces with gushing water outflow; F1 is an image of a tunnel face with rain-like water outflow, and F2 is an image of a tunnel face with linear water outflow; G1 is an image of a wet or dripping tunnel face, and G2 is an image of a dry tunnel face.

[0059] In some embodiments, the total loss function of the tunnel face classification and segmentation model is calculated as follows:

[0060] ;

[0061] in, represents the classification loss function, represents the bounding box loss function, represents the segmentation loss function, , , , are the weight coefficients corresponding to different loss functions, Represents the loss function corresponding to different attribute categories.

[0062] In some embodiments, the classification loss function is calculated as:

[0063] ;

[0064] in, represents the number of samples, represents the number of categories, Indicates The samples belong to The true category label of the class attribute, Indicates The samples belong to The predicted probability of the class attribute.

[0065] In some embodiments, the bounding box loss function includes a complete intersection-over-union (CIOU) loss function and a distribution focal loss (DFL) loss function.

[0066] In some embodiments, the bounding box loss function is calculated as:

[0067] ;

[0068] ;

[0069] ;

[0070] in, represents the complete intersection-over-union loss function, represents the distribution focusing loss function, Represents a bounding box and bounding box The center point distance, Represents a bounding box and bounding box The diagonal length of the minimum bounding box of Represents a bounding box and bounding box The aspect ratio similarity of for The impact factor, represents the center point of the prediction box, represents the center point of the real frame, represents the regression label, and To be closest The two values ​​of express The probability of passing sotfmax is, express The probability of passing sotfmax.

[0071] In some embodiments, the tunnel face classification and segmentation model classifies and segments the tunnel face image to obtain a classification result and a segmentation result map, respectively. The classification result is used to identify the attribute category of the face. The method also includes: performing data enhancement processing on the tunnel face image with a small number of samples in the attribute category to obtain an expanded image set, and inputting the expanded image set and the tunnel face image into the tunnel face classification and segmentation model.

[0072] In some embodiments, data enhancement processing includes processing such as rotation, translation, scaling, adding noise, adding blur, etc., to improve the situation where the number of samples of a certain type in the tunnel face image is small. For example, under normal circumstances, rain or linear water outflow and gushing water outflow are rare. Data enhancement is performed on the tunnel face images containing these two types of groundwater outflow states, respectively, which can increase the number of samples of this type, thereby improving the generalization ability of the tunnel face classification and segmentation model.

[0073] In some embodiments, the number of structural surface groups and the spacing between structural surfaces of the joint trace are calculated according to the segmentation result graph, and the method is: calculating the trace attributes of the joint trace according to the segmentation result graph, and calculating the number of structural surface groups and the spacing between structural surfaces according to the trace attributes.

[0074] In some embodiments, the method for calculating the trace attributes of the joint traces according to the segmentation result graph is:

[0075] Preprocess the segmentation result map; solve and record the connected domain according to the preprocessed segmentation result map; extract the trace skeleton according to the connected domain to obtain the skeleton map; perform line segment fitting on the skeleton map to calculate the length and angle of the joint trace, thereby obtaining the trace property of the joint trace.

[0076] Furthermore, the above-mentioned preprocessing refers to binarizing the segmentation result image. Specifically, thresholding is an image segmentation method, which can convert a grayscale image into a binary image by setting the grayscale of pixels greater than a certain critical grayscale value to the grayscale maximum value and setting the grayscale of pixels less than the critical grayscale value to the grayscale minimum value, thereby achieving binarization.

[0077] Specifically, Figure 5 The mask image of the segmentation result image is given. Figure 6 For the general Figure 5 The image after binarization.

[0078] Furthermore, in this application, the segmentation result image is binarized by:

[0079] ;

[0080] in, Represents the segmentation result Figure 2 The pixel value after quantization, Represents the grayscale value before binarization, represents the first threshold, Indicates the second threshold.

[0081] Furthermore, the connected domain is solved and recorded based on the preprocessed segmentation result map. A connected domain generally refers to an image region (Region, Blob) composed of foreground pixels with the same pixel value and adjacent positions in the image. Connected domains can be solved through connected domain analysis. Specifically, connected domain analysis (Connected Component Analysis, Connected Component Labeling) refers to finding and recording each connected domain in the preprocessed segmentation result map. Each connected domain has a corresponding number, and each connected domain number is independent.

[0082] Combination Figure 7 , Figure 7 Given the basis Figure 6 Specifically, according to the connected domain obtained by the above solution, the preprocessed segmentation result image (hereinafter referred to as "image") is corroded, and the image after the corroded operation is opened, and the image after the opening operation is subtracted from the corroded image to obtain a single extraction result. , repeat the above operation until all pixels in the image are 0. The skeleton image can be obtained by taking the union. Specifically, the erosion operation is to place a template of a set size at the center of the pixel point of the image. When the template moves on the image, if the image below can be seen through the template, then the pixel value at the center point is retained, otherwise it is removed. Specifically, the opening operation refers to first performing an erosion operation on the image and then performing an expansion operation, which can smooth the boundaries of the joint traces.

[0083] Specifically, the calculation method of the skeleton graph is:

[0084] ;

[0085] in, Indicates a single extraction result.

[0086] Furthermore, the skeleton graph is fitted with line segments to calculate the length and angle of the joint trace, thereby obtaining the trace properties of the joint trace. Further explanation: the trace properties of the joint trace are mainly the length and angle of the joint trace.

[0087] Specifically, the line segment fitting is performed on the skeleton graph by:

[0088] ;

[0089] in, and They represent the slope and intercept of the fitted line segment respectively.

[0090] Further, solve and The fitting error of all points on the skeleton graph is minimized. The calculation method of the fitting error is:

[0091] ;

[0092] in, Represents the vertical intercept of a point on the skeleton graph.

[0093] Furthermore, the angle of the joint trace is calculated based on the slope of the fitted line segment. , the calculation method is:

[0094] .

[0095] Furthermore, the length of the joint trace can be obtained according to the pixel count of the shortest path between the head and tail ends of the joint trace in the skeleton graph. Exemplarily, the Dijkstra algorithm can be used for calculation, which is a prior art and will not be described in detail.

[0096] In some embodiments, the number of structural surface groups and the spacing between structural surfaces are calculated based on the trace attributes, and the method is as follows: clustering the angles of the joint traces to obtain clustering results, and grouping the structural surfaces according to the clustering results to obtain the number of structural surface groups; calculating the normal slope of the joint traces; selecting a reference survey line; calculating the average distance at which the joint traces are truncated by the reference survey line, and obtaining the spacing between structural surfaces.

[0097] In some embodiments, the angles of the joint traces are clustered, and the calculation method is:

[0098] ;

[0099] ;

[0100] in, Indicates Angle data, represents the angle value of the cluster center, Represents the error of the entire set of data, The clustering result is obtained when the value is the smallest.

[0101] In some embodiments, the normal slope of the joint trace is calculated by taking the structural surface with the largest number of joint traces as the main trace direction and finding the average angle of the joint trace at this time. , the average angle Rotating 90 degrees gives the normal slope of the joint trace. .

[0102] In some embodiments, a reference survey line is selected. Specifically, the survey line is calculated as follows:

[0103] ;

[0104] in, represents the normal slope, Represents the intercept of the survey line.

[0105] Furthermore, starting from the leftmost joint trace of the skeleton diagram, traversing to the rightmost joint trace, the survey line passing through the largest number of joint traces is taken as the reference trace.

[0106] Furthermore, the development degree of the tunnel face can be obtained according to the number of structural face groups and the spacing between structural faces. The development degree of the tunnel face includes five levels of development: undeveloped, relatively developed, developed, very developed, and disordered species development. Furthermore, the corresponding structural type of the tunnel face can be determined according to the tunnel face image. Combining the tunnel face development degree results and the corresponding structural type, the integrity of the tunnel face structural surface can be determined. The integrity of the structural surface includes five levels of integrity: complete, relatively complete, relatively broken, broken, and extremely broken.

[0107] Furthermore, based on key parameters such as the attribute category of the tunnel face and the integrity of the tunnel face structure, optimization measures are proposed for tunnel construction. For example, the optimization measures may be to reinforce or repair the tunnel face, or to improve the construction technology and method of the tunnel.

[0108] Combination Figure 8 The present application also provides an electronic device 30, including a processor 32 and a memory 31, wherein the memory 31 stores a computer program that can be executed by the processor, and when the computer program is executed by the processor 32, any of the above-mentioned methods for automatically identifying key geological parameters of a tunnel face is implemented.

[0109] Specifically, the memory 31 is used to store computer programs or data. The memory 31 can be, but is not limited to, a random access memory (RAM), a read only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), etc.

[0110] The processor 32 is used to read / write data or computer programs stored in the memory 31 and execute corresponding functions. For example, when the computer program stored in the memory 31 is executed by the processor 32, a method for automatically identifying key geological parameters of a tunnel face provided in an embodiment of the present application can be implemented.

[0111] In summary, the present invention discloses a method and electronic device for automatically identifying geological key parameters of a tunnel face, the method comprising the steps of: constructing a tunnel face classification and segmentation model; the tunnel face classification and segmentation model classifies and segments the tunnel face image, and obtains classification results and segmentation result diagrams respectively; the number of structural surface groups and the structural surface spacing of the joint traces are calculated according to the segmentation result diagram, and the face development degree results are obtained according to the number of structural surface groups and the structural surface spacing, and the face development degree results are used to judge the integrity of the face structure surface; optimization measures are proposed for tunnel construction according to the attribute category of the face and the integrity of the face structure surface. The method of the present application extracts the key parameters of the tunnel face to grasp the stability of the tunnel face, and realizes real-time monitoring and early warning of the stability of the tunnel face, which can not only improve the monitoring efficiency, but also reduce the cost of manpower and material resources.

[0112] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structural transformations made using the contents of the present invention's specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for automatically identifying key geological parameters of a tunnel face, characterized in that: Includes steps: Construct a tunnel face classification and segmentation model; The tunnel face classification and segmentation model classifies and segments the tunnel face image to obtain a classification result and a segmentation result map, respectively, and the classification result is used to identify the attribute category of the tunnel face; Calculating the number of structural surface groups and the distance between structural surfaces of the joint traces according to the segmentation result diagram, obtaining the tunnel face development degree result according to the number of structural surface groups and the distance between structural surfaces, and the tunnel face development degree result is used to judge the integrity of the tunnel face structural surface; Propose optimization measures for tunnel construction according to the attribute category of the tunnel face and the completeness of the structural surface of the tunnel face; The method for calculating the number of structural surface groups and the distance between structural surfaces of the joint trace line according to the segmentation result graph is as follows: calculating the trace attributes of the joint trace line according to the segmentation result graph, and calculating the number of structural surface groups and the distance between structural surfaces according to the trace attributes; The method for calculating the number of structural surface groups and the distance between structural surfaces according to the trace attributes is: clustering the angles of the joint traces to obtain clustering results, grouping the structural surfaces according to the clustering results to obtain the number of structural surface groups; calculating the normal slope of the joint traces; Select reference survey line; Calculating the average distance of the joint trace being intercepted by the reference survey line to obtain the structural plane spacing; Among them, the calculation method for clustering the angles of joint traces is: ; ; Indicates Angle data, represents the angle value of the cluster center, Represents the error of the entire set of data, The smallest clustering result is obtained; The method for calculating the normal slope of the joint trace is: taking the structural surface with the largest number of joint traces as the main trace direction, and calculating the average angle of the joint trace at this time , the average angle Rotating 90 degrees, the normal slope of the joint trace can be obtained ; The calculation method of the reference survey line is: ;in, represents the normal slope, Represents the intercept of the survey line.

2. The method for automatically identifying key geological parameters of a tunnel face according to claim 1, characterized in that: The tunnel face classification and segmentation model includes: a backbone network, a trace segmentation network and an attribute classification network, and the trace segmentation network includes a fusion unit and a detection unit; The input end of the backbone network is used to input the tunnel face image, and the backbone network is used to extract features from the tunnel face image to obtain a feature map, and input the feature map into the trace segmentation network and the attribute classification network respectively; The attribute classification network receives the feature map, performs feature calculation on the feature map and outputs the classification result; The fusion unit of the trace segmentation network receives the feature map and performs feature fusion on the feature map to obtain a fusion map; The detection unit of the trace segmentation network is used to receive the fusion image, detect the fusion image, and output the segmentation result image.

3. The method for automatically identifying key geological parameters of a tunnel face according to claim 2, characterized in that: The tunnel face classification and segmentation model classifies and segments the tunnel face image to obtain a classification result and a segmentation result map, respectively. The classification result is used to identify the attribute category of the tunnel face, and further includes: Data enhancement processing is performed on the tunnel face images with a small number of samples in the attribute category to obtain an expanded image set, and the expanded image set and the tunnel face images are input into the tunnel face classification and segmentation model.

4. The method for automatically identifying key geological parameters of a tunnel face according to claim 1, characterized in that: The method for calculating the trace attributes of the joint traces according to the segmentation result graph is: Preprocessing the segmentation result image; Solve and record connected domains according to the preprocessed segmentation result graph; Extracting a trace skeleton according to the connected domain to obtain a skeleton graph; Line segment fitting is performed on the skeleton graph to calculate the length and angle of the joint trace, thereby obtaining the trace properties of the joint trace.

5. The method for automatically identifying key geological parameters of a tunnel face according to claim 2, characterized in that: The total loss function of the tunnel face classification and segmentation model is calculated as follows: ； in, represents the classification loss function, represents the bounding box loss function, represents the segmentation loss function, , , , are the weight coefficients corresponding to different loss functions, Represents the loss function corresponding to different attribute categories.

6. The method for automatically identifying key geological parameters of a tunnel face according to claim 5, characterized in that: The calculation method of the classification loss function is: ； in, represents the number of samples, represents the number of categories, Indicates The samples belong to The true category label of the class attribute, Indicates The samples belong to The predicted probability of the class attribute.

7. The method for automatically identifying key geological parameters of a tunnel face according to claim 6, characterized in that: The calculation method of the bounding box loss function is: ； ； ； in, represents the complete intersection-over-union loss function, represents the distribution focusing loss function, Represents a bounding box and bounding box The center point distance, Represents a bounding box and bounding box The diagonal length of the minimum bounding box of Represents a bounding box and bounding box The aspect ratio similarity of for The impact factor, represents the center point of the prediction box, represents the center point of the real frame, represents the regression label, and To be closest The two values ​​of express The probability of passing sotfmax is, express The probability of passing sotfmax.

8. An electronic device, characterized in that: It comprises a processor and a memory, wherein the memory stores a computer program that can be executed by the processor, and when the computer program is executed by the processor, the method for automatically identifying key geological parameters of a tunnel face as claimed in any one of claims 1 to 7 is implemented.

Images