Tunnel face unmanned geological recording method and system for tunnel intelligent construction

Abstract

The application belongs to the technical field of tunnel engineering construction, and provides a tunnel face unmanned geological logging method and system for intelligent tunnel construction. The method comprises the following steps: acquiring on-site tunnel surrounding rock image data, classifying the tunnel surrounding rock based on a deep migration learning model; processing on-site test data of the rebound value, wave speed and density of the rock based on a strength prediction neural network model matched with the rock type; extracting a proposal frame from the on-site tunnel surrounding rock image data, and judging whether the object in the proposal frame is a structural plane; when the object in the proposal frame is determined to be a structural plane, extracting a crack proposal frame in the structural plane, judging the combination degree of the structural plane, and performing image processing and grouping statistics on the cracks in the crack proposal frame to calculate a rock mass integrity coefficient; and collecting the tunnel surrounding rock type, the uniaxial compressive strength of the rock and the rock mass integrity coefficient to obtain tunnel face geological logging information.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of tunnel engineering construction, and particularly relates to a tunnel face unmanned geological recording method and system for tunnel intelligent construction. BACKGROUND

[0002] The statements in this section merely provide background information related to the present application and do not necessarily constitute prior art.

[0003] The conventional tunnel face geological recording method is to make a prediction with reference to the survey and design data, observe and record the exposed geological body at the tunnel face, and make a reliable inference on the lithology change of the rock mass in front of the tunnel face based on the geological change law of the surrounding rock. During the tunnel face construction, the geological conditions of the rock mass in front of the tunnel face can be grasped based on the tunnel face information, which can provide an important basis for evaluating the stability of the surrounding rock, determining the tunnel support scheme and construction technology.

[0004] At present, the geological survey personnel generally use traditional tools such as a tape measure, a geological compass and a vernier caliper to measure the geological information on site during the tunnel construction. In the process of engineering geological survey, the traditional geological recording has low efficiency and poor comprehensive effect, especially in the case of complex tunnel geological conditions and accelerated construction progress, it is difficult to achieve the ideal effect. For example, the geological information collection takes a long time, which cannot meet the needs of rapid construction of the project; the measurement and analysis accuracy depends on the experience of the measurement personnel, which cannot guarantee the accuracy and reliability; it is difficult to collect rock samples; the measurement personnel faces the danger of falling and collapse of the tunnel face, etc. Therefore, the construction personnel cannot fully extract, analyze and utilize the tunnel face information by using the manual geological sketching technology, and the updating speed of the manual geological information drawing is slow, which cannot achieve timely and effective feedback for subsequent construction operations.

[0005] Photogrammetry technology is widely used in the field of tunnel face, which can effectively obtain geological information under the condition of poor tunnel construction site environment, and make up for the long time of manual collection, low accuracy and other shortcomings. However, the inventors find that in the result analysis process, manual assistance is still needed to judge the rock type and characteristics, and to fill in the geological recording chart, and it is difficult to determine the position and time of the tunnel face image shooting in a short time, and then obtain high-quality tunnel face images. SUMMARY

[0006] In order to solve the technical problems in the background art, the present application provides a tunnel face unmanned geological recording method and system for tunnel intelligent construction.

[0007] In order to achieve the above purpose, the present application adopts the following technical solutions:

[0008] The first aspect of the present application provides a tunnel face unmanned geological recording method for tunnel intelligent construction.

[0009] A tunnel face unmanned geological logging method for tunnel intelligent construction, comprising:

[0010] Obtaining on-site tunnel surrounding rock image data, and classifying the tunnel surrounding rock based on a deep transfer learning model;

[0011] Processing on-site test data of rebound value, wave speed and density of the rock based on a strength prediction neural network model matched with the rock type;

[0012] Extracting a proposal frame from the on-site tunnel surrounding rock image data, and judging whether the object in the proposal frame is a structural plane; when judging that the object in the proposal frame is a structural plane, extracting a crack proposal frame in the structural plane, judging the combination degree of the structural plane, and performing image processing and grouping statistics on the cracks in the crack proposal frame to calculate a rock mass integrity coefficient;

[0013] Summarizing the tunnel surrounding rock type, the uniaxial compressive strength of the rock and the rock mass integrity coefficient to obtain tunnel face geological logging information.

[0014] The second aspect of the present application provides a tunnel face unmanned geological logging system for tunnel intelligent construction.

[0015] A tunnel face unmanned geological logging system for tunnel intelligent construction, comprising:

[0016] A lithology classification module for obtaining on-site tunnel surrounding rock image data, and classifying the tunnel surrounding rock based on a deep transfer learning model;

[0017] A strength prediction module for processing on-site test data of rebound value, wave speed and density of the rock based on a strength prediction neural network model matched with the rock type;

[0018] An integrity analysis module for extracting a proposal frame from the on-site tunnel surrounding rock image data, and judging whether the object in the proposal frame is a structural plane; when judging that the object in the proposal frame is a structural plane, extracting a crack proposal frame in the structural plane, judging the combination degree of the structural plane, and performing image processing and grouping statistics on the cracks in the crack proposal frame to calculate a rock mass integrity coefficient;

[0019] A geological logging module for summarizing the tunnel surrounding rock type, the uniaxial compressive strength of the rock and the rock mass integrity coefficient to obtain tunnel face geological logging information.

[0020] The third aspect of the present application provides a computer readable storage medium.

[0021] A computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the tunnel-oriented intelligent construction facing unmanned geological logging method as described above.

[0022] A fourth aspect of the present application provides an electronic device.

[0023] An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor implements the steps of the tunnel-oriented intelligent construction facing unmanned geological logging method as described above when executing the program.

[0024] Compared with the prior art, the present application has the following beneficial effects:

[0025] (1) The present application applies image recognition and deep transfer learning technology to the identification and classification of tunnel surrounding rock lithology and the identification and extraction of fissures, replacing traditional manual identification, which can reduce the subjective misjudgment of construction technical personnel on lithology, improve the accuracy of the identification of surrounding rock lithology and structure surface occurrence, and save time and effort.

[0026] (2) The present application comprehensively considers the relationship between the three factors of rock rebound value, longitudinal wave velocity and rock density and rock strength, establishes a rock strength prediction model under multi-factor evaluation, and realizes rapid and accurate evaluation of rock strength.

[0027] The advantages of the additional aspects of the present application will be partially given in the following description, partially become obvious from the following description, or be understood through the practice of the present application. BRIEF DESCRIPTION OF DRAWINGS

[0028] The drawings accompanying the specification of the present application serve to provide a further understanding of the present application, and the illustrative embodiments of the present application and their descriptions serve to explain the present application, and do not constitute an improper limitation on the present application.

[0029] Figure 1 It is an InceptionNet-V3 rock classification model structure diagram.

[0030] Figure 2 It is a coupling strength model confidence analysis flowchart.

[0031] Figure 3 It is a Faster R-CNN basic structure schematic diagram.

[0032] Figure 4 It is a proposal box acquisition flowchart.

[0033] Figure 5 It is a structure surface combination degree acquisition flowchart.

[0034] Figure 6A flowchart for rock mass integrity analysis. DETAILED DESCRIPTION

[0035] The application will be further described below in connection with the drawings and examples.

[0036] It should be noted that the following detailed description is illustrative only, and is intended to provide further description of the application. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0037] It is also to be understood that the terminology used herein is for the purpose of describing particular embodiments only, and is not intended to be limiting of example embodiments in accordance with the present application. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and / or "comprising," when used in this specification, specify the presence of stated features, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, steps, operations, elements, components, and / or groups thereof.

[0038] Example One

[0039] The embodiment is a face-to-face tunnel intelligent construction of the faceless geological recording method. First, the method uses image recognition technology and deep migration learning book to classify the on-site rock, then gets the rebound value, wave speed and density of the rock through on-site test, and inputs the corresponding rock type strength prediction neural network model to get the uniaxial compressive strength of the rock. Second, the Faster R-CNN model in the target detection field is used to extract the crack proposal frame in the face picture, judge the combination degree of the structure surface, and carry out image processing and grouping statistics on the cracks in the proposal frame, realize intelligent face information geological recording.

[0040] Now combined with specific implementation steps are described.

[0041] A face-to-face tunnel intelligent construction of the faceless geological recording method, which specifically comprises:

[0042] Obtain the image data of the on-site tunnel surrounding rock, and classify the tunnel surrounding rock based on the deep migration learning model;

[0043] Process the rebound value, wave speed and density of the rock on-site test data based on the strength prediction neural network model matched with the rock type;

[0044] Extract the proposal frame from the on-site tunnel surrounding rock image data, and judge whether the object in the proposal frame is a structure surface. When the object in the proposal frame is a structure surface, extract the crack proposal frame in the structure surface, judge the combination degree of the structure surface, and carry out image processing and grouping statistics on the cracks in the crack proposal frame to calculate the rock mass integrity coefficient;

[0045] The uniaxial compressive strength of the rock and the rock mass integrity coefficient are summarized to obtain the geological logging information of the tunnel face.

[0046] It should be noted that the step of obtaining the uniaxial compressive strength of the rock and the step of calculating the rock mass integrity coefficient can be performed simultaneously, or one of the steps can be performed before the other, which does not affect the final geological logging information of the tunnel face obtained by the present application.

[0047] In the specific implementation process, the process of obtaining the on-site tunnel surrounding rock image data and classifying the tunnel surrounding rock based on the deep transfer learning model includes:

[0048] A large number of tunnel rock photos are obtained by using a high-definition digital camera, and images of known rock types are collected as a prediction set to perform transfer learning by taking InceptionNet-V3 as a pre-training model to establish a lithology identification model.

[0049] All convolutional layers before the last convolutional layer of the InceptionNet-V3 framework are frozen, a fully connected layer and a SoftMax layer are set as a classification layer to establish a classification result confidence, a feature vector of the rock picture is extracted and trained, and the photos of the on-site rock are identified and analyzed to realize intelligent classification of the rock.

[0050] Specifically:

[0051] (1a) 1850 rock photos of the Hubei HuBei expressway furnace red mountain tunnel and the Jinan big west ring Qinglong mountain tunnel are obtained by using a high-definition digital camera, 1500 of which are taken as a training set and 350 are taken as verification pictures to establish a training classic image identification model.

[0052] (1b) Three lithologies of granite, sandstone and limestone are preset, color and texture of the rock are compared and analyzed for the training set images, and the collected images are pre-classified and analyzed.

[0053] (1c) For example, transfer learning is performed by taking InceptionNet-V3 as a pre-training model, and then the original image data set is transmitted into the pre-training model.

[0054] (1d) All convolutional layers before the last convolutional layer of the InceptionNet-V3 framework are frozen, and a fully connected layer and a SoftMax layer are set as a classification layer at the end of the model framework, as shown in Figure 1 .

[0055] (1e) The SoftMax function performs exponentiation on the input value, i.e. the output value of the fully connected layer, to obtain the confidence of the classification result after sparse processing.

[0056] (1f) The pictures of typical granite, sandstone, limestone and other different types of rocks collected are sent into the partially frozen InceptionNet-V3 model to extract feature vectors, and the convolutional layers and classification layers that are not frozen are retrained to make the model capable of identifying the types of rocks by using the pictures of rocks taken on the construction site.

[0057] (1g) The model structure and parameters on the InceptionNet-V3 model large data training set are used for tunnel internal rock identification to achieve higher accuracy.

[0058] The process of obtaining the uniaxial compressive strength of the rock includes:

[0059] The confidence maximum and the second largest two labels are used, and the BP neural network technology is used for comprehensive analysis of the measured rebound value, longitudinal wave velocity and rock density.

[0060] The confidence maximum and the second largest values in the rock type image recognition result are doc1 and doc2 (degree of confidence), and the corresponding rock labels are rock1 and rock2, respectively. The strength ranges of the two rocks are range1 and range2, respectively. The results obtained by bringing the measured rebound value, longitudinal wave velocity and density into the rock strength prediction model corresponding to rock1 and rock2 are σ1 and σ2, respectively.

[0061] The following takes σ1 and σ2 as examples for comparison with 60 and 20, respectively:

[0062] If doc1≥60 and doc2≤20, the strength is σ1. Otherwise, if σ1 belongs to range1, the result is σ1; if σ1 does not belong to range1 and σ2 belongs to range2, the result is σ2; if σ1 does not belong to range1 and σ2 does not belong to range2, the on-site rock photo is re-taken for prediction or artificial assistance for correction and identification, and the rock strength is obtained.

[0063] The preset rock1 and rock2 are granite and sandstone, respectively, and do not satisfy doc1≥60 and doc≤20, and the rock strength is predicted. The test data uses homogeneous granite and moderately weathered heterogeneous sandstone collected in the Three Gorges Reservoir Area of Chongqing, and there are 10 test pieces of each type of rock. The data is shown in Table 1.

[0064] Table 1 Rock-related data

[0065]

[0066] It should be noted that the threshold values of σ1 and σ2 can be set according to actual conditions, which will not be described in detail here.

[0067] The strength prediction model based on the BP neural network structure is a three-layer fully connected network. The data is normalized before input. The model parameters are randomly initialized with a normal distribution with a standard deviation of 0.1 and a mean of 0. The ReLU activation function is used. The batch size is 3. The mean square error is used as the loss function. The learning rate is 0.0001. The process is repeated 200 times, as shown in Figure 2 .

[0068] A set of data is selected from the 20 data as a test group. The remaining 19 sets of data are placed in two BP neural network structures according to the rock type for learning. The strength prediction models of the two rock types are obtained. Then the test group data is predicted by the process, and the relative error is calculated according to the following formula.

[0069] Relative error = × 100% (2)

[0070] Where: σ is the predicted value; σ0 is the test strength; the unit is MPa.

[0071] The process of determining whether the object in the proposal box is a structural surface includes:

[0072] Proposal box generation:

[0073] The loss function is used to generate the proposal box part (region proposal network) to frame all the non-background objects in the picture. The classification loss is calculated to evaluate the ability of the model to distinguish objects from backgrounds. The cross-entropy error The loss function is used to determine whether the proposal box is a true detection box. The regression loss is calculated to evaluate the ability of the proposal box to fit the true detection box after shifting.

[0074] Loss function (1)

[0075] Wherein, wherein is the number of anchor boxes; The loss function is used to determine whether the proposal box is a true detection box. The regression loss is calculated to evaluate the ability of the proposal box to fit the true detection box after shifting. Among them is the total number of positive and negative samples selected when training the proposal box regression; is the proportion coefficient between the two components of the loss function, which is related to the number of anchor boxes and the total number of positive and negative samples selected when training the proposal box regression; p* is 1 when there is an object and 0 when there is no object; respectively represent the offset of the predicted proposal box and the true detection box to the anchor box, both of which are in vector form.

[0076] Identify the structural surface in the frame:

[0077] Training stage: the input data is the original picture and the position, size and label of the real detection box. The features of the original picture are extracted by convolution and pooling, and the machine learning operation is performed by the comprehensive use of the full connection layer features. The result is compared with the label information in the loss function, and the deviation of the machine learning from the expected value is output. According to the numerical value of the degree, the parameters of the machine learning are optimized by back propagation, so that the structure plane in the picture can be more close to the expected recognition.

[0078] Test stage: input the picture into the model, and output the result containing the proposal box and the label after the model processing. Compare it with the expected detection box and label to evaluate the detection ability of the model; judge the occurrence information of the structure plane in the proposal box by recognizing the structure plane in the box.

[0079] Input the tunnel face image, and input the whole picture into the convolution layer for feature extraction to obtain the feature map of the tunnel face image. The feature map of the tunnel face photo uses the method of transfer learning, and uses the VGGNet16 model without the last full connection layer and SoftMax layer to extract.

[0080] Many anchors are generated on the feature map by using the region proposal network. These anchors correspond to anchor boxes in the original image. The objects in the box are judged to be or not to be structure planes by the full connection layer and SoftMax function.

[0081] The process of integrity analysis is:

[0082] The region proposal network in Faster R-CNN is used to locate the proposal box of each structure plane of the tunnel face. The structure of Faster R-CNN is as shown in Figure 3

[0083] This embodiment introduces Faster R-CNN technology, which solves the problem that the area of the tunnel face is too large, the features such as illumination, shape, color and size of different parts are quite different, and the traditional method is difficult to identify and extract the structure plane.

[0084] The Faster R-CNN framework is selected to identify and extract the structure plane in the tunnel face image, and the combination degree of the structure plane in each proposal box is obtained.

[0085] Table 2 Classification table of structure plane combination degree

[0086]

[0087] ​The image processing is performed on each proposal box picture to obtain the morphology of the structural surface, which is approximated as a straight line. Then, the number of groups of main structural surfaces and the development degree index such as the interval are obtained through statistical analysis, and the rock mass integrity coefficient is calculated.

[0088] The calculation method is to measure the interval of each group of structural surfaces in the rock mass, and calculate the number of fractures per unit volume of rock mass by taking the average value.

[0089] The calculation formula is:

[0090]

[0091] In the formula: K V is the rock mass integrity coefficient; J v is the volume of the rock mass joint number (strip / m 3 ); n is the number of groups of structural surfaces in the statistical area; S i is the number of structural surfaces per meter long along the normal direction of the first i group of structural surfaces; S 0 is the number of non-grouped joints per cubic meter of rock mass.

[0092] The position of the anchor frame is corrected using the bounding box regression (BBR) algorithm to be closer to the real structural surface detection frame. The two branches finally form multiple structural surface proposal frames with more accurate positioning, as shown in Figure 4 . Using the Faster R-CNN technology, the feature map corresponding to the structural surface proposal frame obtained in the previous step is used to determine which combination degree the structural surface in each proposal frame belongs to through the full connection layer and SoftMax function, and the combination degree with the largest proportion is taken as the combination degree of the main structural surface of the whole working face. The steps are shown in Figure 5 .

[0093] The corresponding gray scale image is generated for each proposal frame, and binary processing is performed to separate the structural surface from the working face picture. The binary transformation method is shown in formula (3), and the optimal threshold value is adaptively determined based on the maximum inter-class variance segmentation method (also known as the Otsu method) , n ( x , y ) is the output value after threshold transformation, m ( x , y ) is the image input value.

[0094] (3)

[0095] The image is sequentially subjected to closing and opening operations to bridge the breakpoints of the structural surfaces and remove impurities. The closing operation involves dilation followed by erosion, which fills small cracks; the opening operation involves erosion followed by dilation, which removes isolated points. The structural surfaces, abstracted as line segments, are obtained using Hough transform and a point counter.

[0096] Three 5m×2m test frames were randomly selected, and the number of groups of main structural surfaces that were abstracted into straight lines and their average distance (relative distance) were counted. The absolute distance was obtained based on the shooting distance and the camera zoom ratio.

[0097] The calculation process is as follows Figure 6 As shown, the overall structure of this tunnel face is generally considered to have a moderate degree of bonding, with two groups and an average spacing of 0.5 meters. Kv The value was 0.65, which is only 0.05 different from the 0.7 calculated according to the design documents.

[0098] Example 2

[0099] This embodiment provides an unmanned geological logging system for tunnel face construction, which includes:

[0100] The lithology classification module is used to acquire on-site tunnel surrounding rock image data and classify the tunnel surrounding rock based on a deep transfer learning model.

[0101] The strength prediction module is used to process field test data of rock rebound value, wave velocity and density based on the strength prediction neural network model matched to rock type.

[0102] The integrity analysis module is used to extract proposal frames from the on-site tunnel surrounding rock image data and determine whether the objects in the proposal frames are structural surfaces. When the objects in the proposal frames are determined to be structural surfaces, the crack proposal frames in the structural surfaces are extracted, the degree of bonding of the structural surfaces is determined, and the cracks in the crack proposal frames are processed and grouped statistically to calculate the rock mass integrity coefficient.

[0103] The geological logging module is used to summarize the tunnel surrounding rock type, the uniaxial compressive strength of the rock, and the rock mass integrity coefficient to obtain the geological logging information of the tunnel face.

[0104] It should be noted that each module in this embodiment corresponds one-to-one with each step in Embodiment 1, and their specific implementation processes are the same, so they will not be repeated here.

[0105] Example 3

[0106] This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in the unmanned geological logging method for tunnel face as described above for intelligent tunnel construction.

[0107] Embodiment Four

[0108] The embodiment provides an electronic device, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps in the tunnel-oriented intelligent construction faceless geological recording method as described above when executing the program.

[0109] The present application is described with reference to flowcharts and / or block diagrams of the methods, devices (systems) and computer program products 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 processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate a means for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 The functions specified in one flow or multiple flows and / or blocks Figure 1 The means for implementing the functions specified in one flow or multiple flows and / or blocks.

[0110] The above only describes the preferred embodiments of the present application and is not used to limit the present application. The present application can have various modifications and changes for those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A faceless geological logging method for tunnel intelligent construction, characterized in that, The method comprises the following steps: acquiring tunnel surrounding rock image data, and classifying the tunnel surrounding rock based on a deep transfer learning model; processing field test data of rock rebound value, wave velocity and density based on a strength prediction neural network model matched with the rock type; extracting a proposal frame from the field tunnel surrounding rock image data, and judging whether the object in the proposal frame is a structural plane; when the object in the proposal frame is determined to be a structural plane, extracting a crack proposal frame in the structural plane, judging the combination degree of the structural plane, and performing image processing and grouping statistics on the cracks in the crack proposal frame to calculate a rock mass integrity coefficient; summarizing the tunnel surrounding rock type, the uniaxial compressive strength of the rock and the rock mass integrity coefficient to obtain face geological logging information; the deep transfer learning model performs transfer learning by taking InceptionNet as a pre-training model; the calculation process of the rock mass integrity coefficient is as follows: the spacing of each group of structural planes in the rock mass is measured, and the average value is used to calculate the number of cracks per unit volume of rock mass to obtain the rock mass integrity coefficient.

2. The tunnel-oriented, intelligence-built, face-unmanned, geological-logging method of claim 1, wherein, The strength prediction neural network model is based on a BP neural network structure.

3. The tunnel-oriented, intelligent, constructed face, unmanned, geological logging method of claim 1, wherein, In the process of judging whether the object in the proposal frame is a structural plane, the ability of the classification loss evaluation model to distinguish the object from the background is evaluated by calculating the cross-entropy error, and the loss function of whether the proposal frame is a real detection frame is calculated and judged.

4. A faceless geological logging system for tunneling intelligent construction, characterized in that, The method comprises the following steps: a lithology classification module for acquiring field tunnel surrounding rock image data and classifying the tunnel surrounding rock based on a deep transfer learning model; a strength prediction module for processing field test data of rock rebound value, wave velocity and density based on a strength prediction neural network model matched with the rock type; an integrity analysis module for extracting a proposal frame from the field tunnel surrounding rock image data, and judging whether the object in the proposal frame is a structural plane; when the object in the proposal frame is determined to be a structural plane, extracting a crack proposal frame in the structural plane, judging the combination degree of the structural plane, and performing image processing and grouping statistics on the cracks in the crack proposal frame to calculate a rock mass integrity coefficient; a geological logging module for summarizing the tunnel surrounding rock type, the uniaxial compressive strength of the rock and the rock mass integrity coefficient to obtain face geological logging information; the deep transfer learning model performs transfer learning by taking InceptionNet as a pre-training model; the calculation process of the rock mass integrity coefficient is as follows: the spacing of each group of structural planes in the rock mass is measured, and the average value is used to calculate the number of cracks per unit volume of rock mass to obtain the rock mass integrity coefficient.

5. The tunnel-oriented, intelligent, built face, unmanned, geological logging method of claim 4, wherein, The strength prediction neural network model is based on a BP neural network structure.

6. The tunnel-oriented, intelligent, constructed face, unmanned, geological logging method of claim 4, wherein, In the process of judging whether the object in the proposal frame is a structural plane, the ability of the classification loss evaluation model to distinguish the object from the background is evaluated by calculating the cross-entropy error, and the loss function of whether the proposal frame is a real detection frame is calculated and judged.

7. A computer readable storage medium having stored thereon a computer program, which, when executed by a processor, implements the steps of the face unmanned geological logging method for tunnel intelligent construction according to any one of claims 1-3.

8. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, the processor implementing the steps of the method for the unmanned geological logging of a tunnel face oriented to intelligent tunnel construction according to any one of claims 1 to 3 when executing the program.

Images