Tunnel face surrounding rock grading rapid evaluation method based on image recognition

Abstract

The invention provides a tunnel face surrounding rock grading rapid evaluation method based on image recognition, and belongs to the technical field of tunnel engineering and image recognition. According to the method, a tunnel face surrounding rock data set obtained historically is expanded, and an increased numerical value information vector is converted into image information of tunnel face surrounding rock through a generative AI technology; training the complete image information set through a deep learning technology to obtain a deep learning model; monitoring the grade of the tunnel face surrounding rock in real time according to the deep learning model; according to the method, the image information conforming to the tunnel face surrounding rock development rule is obtained through the data information vectors increased after expansion, and the tunnel face surrounding rock grade predicted through the deep learning model trained by the image information is more accurate; according to the method, the grade information of the surrounding rock of the tunnel face can be accurately and rapidly obtained in real time, real-time monitoring of tunnel construction is facilitated, a corresponding supporting structure is established, and the safety of tunnel construction is ensured.

Description

Technical Field

[0001] The present invention relates to the technical fields of tunnel engineering and image recognition. Specifically, it relates to a method for quickly evaluating the classification of surrounding rock of a tunnel face based on image recognition. Background Art

[0002] Surrounding rock classification refers to dividing an infinite rock mass sequence into a finite number of categories with different stability degrees according to indexes such as the integrity of the rock mass and the strength of the rock, that is, classifying some surrounding rocks with similar stability into one category and dividing all surrounding rocks into several categories; on the basis of the surrounding rock classification, the best construction method and support structure design are given according to the stability degree of each category of surrounding rock.

[0003] Therefore, for the field of tunnel engineering, in order to obtain the most suitable construction method and support structure design; first, it is necessary to be able to determine the quality of the surrounding rock of the tunnel face, that is, to determine the grade of the surrounding rock of the tunnel face. Different grades correspond to different qualities, and corresponding support designs for different qualities can ensure the safe progress of the tunnel project.

[0004] There are various existing methods for obtaining the grade of the surrounding rock of the tunnel face, including: core sampling method, geological acoustic wave detection method, neural network analysis method, on-site measurement of rock mechanics parameters and numerical simulation method. Except for the neural network analysis method, other methods cannot quickly obtain the grade of the surrounding rock of the tunnel face in real time; and the neural network analysis method mainly obtains the grade of the surrounding rock of the tunnel face by predicting the image in real time; however, when the existing neural network technology analyzes image information, only the historical data information set is used as the object of neural network training, and this training method often cannot reflect the change process of all the tunnel faces of the target tunnel.

[0005] Adopting what method to increase the accuracy of the neural network prediction of the grade of the surrounding rock of the tunnel face will be beneficial to improving the construction safety of the tunnel project. Summary of the Invention

[0006] The purpose of the present invention is to overcome the above problems existing in the prior art and greatly improve its technical effect on the basis of the original technology; first, the present invention provides a method for quickly evaluating the classification of surrounding rock of a tunnel face based on image recognition, and the method includes:

[0007] (1) Obtaining all the data information of the surrounding rock of the tunnel face on the terrain that is the same as the target terrain in history; the data information of the surrounding rock of the tunnel face includes: the image information of the surrounding rock of the tunnel face and the numerical information corresponding to the image information; the numerical information includes: the structure and lithology information of the rock stratum corresponding to the image information of the surrounding rock of the tunnel face, the fracture and joint information, the density information of the rock, and the hydrogeological condition information;

[0008] (2) Combine all the numerical information in (1) to form a numerical information vector set, and perform an expansion process on the numerical information vector set to form an expanded numerical information vector set; the expanded numerical information vector set includes: the original numerical information vectors and the added numerical information vectors; among them, the original numerical information vectors refer to the numerical information vectors composed of the numerical information in (1), and the added numerical information vectors refer to the vectors other than the original numerical information vectors added after the expansion process; the method for expanding the numerical information vector set: a) Extract the element information of the numerical information in (1) to form a set corresponding to each element information; b) Observe whether there are obvious breaks in the data in the set corresponding to each element information, and use the equal-distance method or direct insertion method to expand the data for the data with obvious breaks; the obvious breaks include obvious breaks in categorical data and obvious breaks in continuous data; obvious breaks in categorical data mean that the data corresponding to the element is categorical data, and there is a lack of a certain category in the categorical data, and the quantity corresponding to the category is significantly less than that of other categories; obvious breaks in continuous data mean that the distance between two adjacent data sorted in ascending order is significantly greater than the distance between other adjacent data; c) Expand the data information vector set according to the expanded data of each element and the original data of each element, and the expanded part is the added numerical information vector; it is required that the elements in the added numerical information vector contain the expanded data, and at the same time, the added numerical information vector needs to conform to the theoretical structural characteristics of the surrounding rock of the tunnel face and correspond to a classification of the surrounding rock of the tunnel face.

[0009] (3) Use generative AI technology to convert the added numerical information vectors into image information of the surrounding rock of the tunnel face, and obtain the surrounding rock grades corresponding to each data information vector according to the railway tunnel design specifications, that is, obtain the surrounding rock classification corresponding to the image information converted from the numerical information vectors; at the same time, obtain the surrounding rock grades corresponding to each image information according to the numerical information corresponding to the image information in (1), and integrate the image information in (1) and the newly generated image information through generative AI technology together to form a complete image information set.

[0010] (4) Train the complete image information set through deep learning technology to obtain a deep learning model; the training method is: use the image information in the image information set as the input of the deep learning technology, and use the surrounding rock classification corresponding to the image information as the output of the deep learning technology to train the deep learning technology.

[0011] (5) Real-time obtain the image information of the surrounding rock of the tunnel face of the target tunnel, input the obtained image information into the deep learning model, and directly obtain the grade of the surrounding rock of the corresponding tunnel face according to the output result of the deep learning model.

[0012] Specifically, the target terrain includes: The target terrain refers to the terrain that is being excavated and for which the surrounding rock grade of the heading face needs to be determined.

[0013] Specifically, the classified data of obvious faults includes: lithological information and hydrogeological information, and the continuous data of obvious faults includes: fracture information, density information, and hydrogeological information.

[0014] Specifically, the number of the corresponding classification is significantly less than that of other classifications includes: If the number of the corresponding classification is less than half of the minimum number of the smallest classification among other classifications, it is considered that the number of the corresponding classification is significantly less than that of other classifications; The distance between two adjacent data is significantly longer than that between other adjacent data includes: If the distance between two adjacent data is twice the length of the longest data among any other adjacent data, it is considered that the distance between two adjacent data is significantly longer than that between other adjacent data.

[0015] Specifically, the increased numerical information vector needs to conform to the theoretical structural characteristics of the surrounding rock of the tunnel heading face: The surrounding rock of the tunnel heading face formed by natural excavation satisfies the theoretical structural characteristics of the surrounding rock of the tunnel heading face.

[0016] Specifically, the generative AI technologies include: Generative Adversarial Networks (GANs), Variational Autoencoders (VAE), Autoencoders (AE), Long Short-Term Memory Neural Networks (LSTM), Deep Convolutional Generative Models, and Self-Supervised Learning; The increased numerical information vector is transformed into the image information of the surrounding rock of the tunnel heading face through the generative AI technologies.

[0017] Specifically, obtaining the surrounding rock grade corresponding to each data information vector according to the railway tunnel design code includes: Obtaining the BQ value corresponding to each data information vector according to the railway tunnel design code and determining the surrounding rock grade corresponding to each data information vector; Among them, the calculation formula of BQ is:

[0018] BQ = 100 + 3Rc + 250Kv

[0019] Among them, BQ is the basic quality index of the surrounding rock, Rc is the uniaxial saturated compressive strength, Kv refers to the integrity of the rock mass, and the values of Rc and Kv need to be obtained according to the data in the increased numerical information vector;

[0020] Integrating the image information in (1) and the newly generated image information through the generative AI technology together to form a complete image information set includes: The surrounding rock grade of the tunnel heading face is divided into six categories in the complete image information set. Among them, each image information in the complete image information set corresponds to a classification of the surrounding rock of the tunnel heading face.

[0021] Specifically, the training of the complete image information set through deep learning technology includes: Deep learning technology includes: Convolutional Neural Network technology CNN, Deep Residual Network ResNet, Inception network, and Attention Mechanism. The complete image information set is trained through deep learning technology to obtain a deep learning model. The basis for the completion of model training includes: reaching the number of training cycles and the loss function no longer decreasing.

[0022] Specifically, the real-time acquisition of the image information of the surrounding rock of the tunnel face of the target tunnel includes: real-time acquisition of the image information of the surrounding rock of the tunnel face of the target tunnel, and preprocessing the image information, including: denoising processing, normalization, and standardization processing; subsequently, the preprocessed image information is used as the input of deep learning, and the grade of the surrounding rock of the tunnel face corresponding to the image information is determined in real time through the output.

[0023] The beneficial effects of the present invention are:

[0024] The present invention provides a rapid evaluation method for the classification of the surrounding rock of the tunnel face based on image recognition; this method has the following advantages: (1) By expanding the numerical information vector set, an expanded numerical information vector set is formed, and then through the increased data information vectors in the expanded set, image information conforming to the development law of the surrounding rock of the tunnel face is obtained. Finally, the grade of the surrounding rock of the tunnel face predicted by the deep learning model trained with the image information will be more accurate; (2) Through this method, the grade information of the surrounding rock of the tunnel face can be quickly obtained in real time, which is beneficial to the real-time monitoring of tunnel construction, establishing a corresponding support structure, and ensuring the safety of tunnel construction. Description of the Drawings

[0025] Figure 1 : Flowchart of a rapid evaluation method for the classification of the surrounding rock of the tunnel face based on image recognition of the present invention. Detailed Embodiments

[0026] The following detailed description of the specific embodiments of the present invention is provided in conjunction with the accompanying drawings; it should be understood that the specific embodiments given here are only for the purpose of illustrating and explaining the present invention, and should not be used to limit the present invention.

[0027] It should be noted that many specific details are set forth in the following description to facilitate a full understanding of the present invention. However, the present invention may have other embodiments and variations, and therefore, the protection scope of the present invention is not limited by the specific embodiments disclosed below.

[0028] Such as Figure 1As shown in the figure, a flowchart of a rapid evaluation method for surrounding rock classification of tunnel heading faces based on image recognition according to an embodiment of the present invention. The flowchart includes: Step S100, obtaining all the surrounding rock data information of tunnel headings on terrains that are the same as the target terrain in history; the surrounding rock data information of the heading face includes: image information of the surrounding rock of the tunnel heading face and numerical information corresponding to the image information; the numerical information includes: information on the structure and lithology of the rock stratum corresponding to the image information of the surrounding rock of the tunnel heading face, fracture and joint information, density information of the rock, and hydrogeological condition information; Step S101, forming a numerical information vector set from all the numerical information in Step S100, and performing an expansion process on the numerical information vector set to form an expanded numerical information vector set; the expanded numerical information vector set includes: the original numerical information vector and the added numerical information vector; where the original numerical information vector refers to the numerical information vector formed by the numerical information in Step S100, and the added numerical information vector refers to the vector other than the original numerical information vector added through the expansion process; Step S102, converting the added numerical information vector into image information of the surrounding rock of the tunnel heading face through generative AI technology, and obtaining the surrounding rock grade corresponding to each data information vector according to the railway tunnel design specifications, that is, obtaining the surrounding rock classification corresponding to the image information converted from the numerical information vector; at the same time, obtaining the surrounding rock grade corresponding to each image information according to the numerical information corresponding to the image information in Step S100, and integrating the image information in Step S100 and the newly generated image information through generative AI technology together to form a complete image information set; Step S103, training the complete image information set through deep learning technology to obtain a deep learning model; the training method is: using the image information in the image information set as the input of the deep learning technology, and using the surrounding rock classification corresponding to the image information as the output of the deep learning technology to train the deep learning technology; Step S104, obtaining in real time the image information of the surrounding rock of the tunnel heading face of the target tunnel, and inputting the obtained image information into the deep learning model, and directly obtaining the grade of the corresponding tunnel heading face surrounding rock according to the output result of the deep learning model.

[0029] Specifically, the method first obtains all the data information of the surrounding rock of the tunnel face on the terrain that is the same as the target terrain in history; the data information of the surrounding rock of the tunnel face includes the image information of the surrounding rock of the tunnel face and the numerical information corresponding to the image information; all the numerical information is formed into a numerical information vector set, and the numerical information vector set is expanded to form an expanded numerical information vector set; the expanded numerical information vector set is divided into the original numerical information vector and the increased numerical information vector; the increased numerical information vector is converted into the image information of the surrounding rock of the tunnel face through generative AI technology, and the surrounding rock grade corresponding to each data information vector is obtained according to the railway tunnel design specifications; and the surrounding rock grade corresponding to each image information is obtained according to the numerical information corresponding to the image information in history, and the image information in history and the newly generated image information through generative AI technology are integrated together to form a complete image information set; the complete image information set is trained through deep learning technology to obtain a deep learning model; the image information of the surrounding rock of the tunnel face of the target tunnel is obtained in real time, and the obtained image information is input into the deep learning model, and the grade of the surrounding rock of the corresponding tunnel face is quickly obtained according to the output result of the deep learning model.

[0030] Step S100, obtain all the data information of the surrounding rock of the tunnel face on the terrain that is the same as the target terrain in history; the data information of the surrounding rock of the tunnel face includes: the image information of the surrounding rock of the tunnel face and the numerical information corresponding to the image information; the numerical information includes: the structure and lithology information of the rock formation corresponding to the image information of the surrounding rock of the tunnel face, the fracture and joint information, the density information of the rock, and the hydrogeological condition information.

[0031] Specifically, the target terrain refers to the terrain that is being excavated or planned to be excavated, and the terrain for which the surrounding rock grade of the tunnel face needs to be determined.

[0032] Step S101, form all the numerical information in Step S100 into a numerical information vector set, and expand the numerical information vector set to form an expanded numerical information vector set; the expanded numerical information vector set includes: the original numerical information vector and the increased numerical information vector; where the original numerical information vector refers to the numerical information vector formed by the numerical information in Step S100, and the increased numerical information vector refers to the vector other than the original numerical information vector added after the expansion process.

[0033] Specifically, the method for expanding the logarithmic value information vector set: a) Extract the element information of the numerical information in (1) to form a set corresponding to each element information; b) Observe whether there are obvious discontinuities in the data in the set corresponding to each element information, and use the equal-distance method or direct insertion method to expand the data for the data with obvious discontinuities; the obvious discontinuities include obvious discontinuities in categorical data and obvious discontinuities in continuous data; the obvious discontinuity in categorical data means that the data corresponding to the corresponding element is categorical data, and there is a lack of a certain category in the categorical data, and the quantity of the corresponding category is significantly less than that of other categories; the obvious discontinuity in continuous data means that the distance between two adjacent data sorted in ascending order is significantly greater than the distance between other adjacent data; c) Expand the data information vector set according to the expanded data of each element and the original data of each element, and the expanded part is the increased numerical information vector; it is required that the elements in the increased numerical information vector contain the expanded data, and at the same time, the increased numerical information vector needs to conform to the theoretical structural characteristics of the tunnel face surrounding rock and correspond to a classification of the tunnel face surrounding rock.

[0034] In the above embodiment, specifically, the obvious discontinuities in categorical data include: lithology information and hydrogeological information, and the obvious discontinuities in continuous data include: fracture information, density information, and hydrogeological information.

[0035] In the above embodiment, specifically, if the quantity of the corresponding category is less than half of the minimum quantity of other categories, it is considered that the quantity of the corresponding category is significantly less than that of other categories; the distance between two adjacent data being significantly greater than the distance between other adjacent data includes: the distance between two adjacent data is twice the longest data among any other adjacent data, then it is considered that the distance between two adjacent data is significantly greater than the distance between other adjacent data.

[0036] In the above embodiment, specifically, the increased numerical information vector needs to conform to the theoretical structural characteristics of the tunnel face surrounding rock means that the tunnel face surrounding rock formed by natural excavation satisfies the theoretical structural characteristics of the tunnel face surrounding rock.

[0037] Step S102, convert the increased numerical information vector into the image information of the tunnel face surrounding rock through generative AI technology, and obtain the surrounding rock grade corresponding to each data information vector according to the railway tunnel design specification, that is, obtain the surrounding rock classification corresponding to the image information converted from the numerical information vector; at the same time, obtain the surrounding rock grade corresponding to each image information according to the numerical information corresponding to the image information in step S100, and integrate the image information in step S100 and the newly generated image information through generative AI technology together to form a complete image information set.

[0038] Specifically, the generative AI technologies include: Generative Adversarial Networks (GANs), Variational Autoencoders (VAE), Autoencoders (AE), Long Short-Term Memory Neural Networks (LSTM), Deep Convolutional Generative Models, and Self-Supervised Learning; in the present invention, any one of the generative AI technologies is used to convert the added numerical information vector into the image information of the surrounding rock of the tunnel heading face.

[0039] In the above embodiments, specifically, the BQ value corresponding to each data information vector is obtained according to the railway tunnel design specifications, and the surrounding rock grade corresponding to each data information vector is determined; among them, the calculation formula of BQ is:

[0040] BQ = 100 + 3Rc + 250Kv

[0041] Wherein, BQ is the basic quality index of the surrounding rock, Rc is the uniaxial saturated compressive strength, Kv refers to the integrity of the rock mass, and the values of Rc and Kv need to be obtained according to the data in the added numerical information vector;

[0042] Integrating the image information in step S100 and the newly generated image information by the generative AI technology together to form a complete image information set includes: in the complete image information set, the grades of the surrounding rock of the tunnel heading face are divided into six categories, and each image information in the complete image information set corresponds to a classification of the surrounding rock of the tunnel heading face.

[0043] Step S103, training the complete image information set through deep learning technology to obtain a deep learning model; the training method is: using the image information in the image information set as the input of the deep learning technology, and using the surrounding rock classification corresponding to the image information as the output of the deep learning technology to train the deep learning technology.

[0044] Specifically, the deep learning technologies include: Convolutional Neural Network Technology (CNN), Deep Residual Network (ResNet), Inception Network, and Attention Mechanism; training the complete image information set through deep learning technology to obtain a deep learning model; selecting any one of the deep learning technologies to train the complete image information set; the basis for completing the model training includes: reaching the training loop times and the loss function no longer decreasing.

[0045] Step S104, obtaining the image information of the surrounding rock of the tunnel heading face of the target tunnel in real time, and inputting the obtained image information into the deep learning model, and directly obtaining the grade of the surrounding rock of the corresponding tunnel heading face according to the output result of the deep learning model.

[0046] Specifically, the image information of the surrounding rock of the tunnel face of the target tunnel is obtained in real time, and the image information is preprocessed, including: denoising processing, normalization and standardization processing, etc.; subsequently, the preprocessed image information is used as the input of deep learning, and the grade of the surrounding rock of the tunnel face corresponding to the image information is rapidly evaluated in real time, so as to ensure the normal progress of subsequent construction.

[0047] To more clearly explain the above embodiments, the following is illustrated with a specific case: In the Madong Iron Project, the Shuangwendan No. 2 Tunnel has created five most projects in the whole line of the project and in Malaysia, namely, the longest single-hole double-track, the longest drill-and-blast method construction, the longest auxiliary adit, the longest single-hole double-track ballastless track, and the longest reverse slope drainage railway tunnel; the construction organization and arrangement units of the Shuangwendan No. 2 Tunnel are numerous, the tunnel working faces are numerous, and the management span is large. It is necessary to form a systematic tunnel mechanized rapid construction technology to greatly improve the project construction efficiency and reduce the project construction cost on the premise of ensuring the project safety and quality; therefore, it is crucial to determine the grade of the surrounding rock of the tunnel face in real time. The method for rapidly obtaining the analysis and evaluation of the surrounding rock of the tunnel face of the present invention is applied to the Shuangwendan No. 2 Tunnel in the Madong Iron Project; First, all the data information of the surrounding rock of the tunnel face with the same terrain as the Shuangwendan No. 2 Tunnel in history is obtained; the obtained data information of the surrounding rock of the tunnel face is divided into the image information of the surrounding rock of the tunnel face and the numerical information corresponding to the image information; the numerical information is formed into a numerical information vector set, and the numerical information vector set is expanded by using the methods of a), b) and c) in step S101 to form an expanded numerical information vector set; the increased numerical information vectors are converted into the image information of the surrounding rock of the tunnel face through generative AI technology, and the surrounding rock grades corresponding to each data information vector are obtained according to the railway tunnel design specifications; at the same time, the surrounding rock grades corresponding to each image information are obtained according to the numerical information corresponding to the image information in history, and all the image information is integrated together to form a complete image information set; the complete image information set is trained through deep learning technology to obtain a deep learning model; the image information of the surrounding rock of the tunnel face of the Shuangwendan No. 2 Tunnel is obtained in real time, and after the obtained image information is preprocessed, it is input into the deep learning model, and the grade of the surrounding rock of the tunnel face under construction corresponding to the Shuangwendan No. 2 Tunnel is directly and rapidly obtained according to the output result of the deep learning model. The construction personnel carry out corresponding support design according to the obtained surrounding rock grade and the actual situation on site to ensure the construction safety.

[0048] It should be understood that the above embodiments are one or more embodiments of the present invention. Based on the present invention, there are many other embodiments and their deformations; when ordinary technicians in this industry do not make pioneering innovations, the deformations and modifications made through the present invention all fall within the protection scope of the present invention.

Claims

1. A method for rapid assessment of surrounding rock classification of tunnel face based on image recognition, characterized in that: The method comprises: 1) Acquire data information of surrounding rocks of all tunnel faces on the same terrain as the target terrain in history; the data information of surrounding rocks of the tunnel faces includes: image information of surrounding rocks of the tunnel faces and numerical information corresponding to the image information; the numerical information includes: structure and lithology information of rock formations, crack and joint information, rock density information and hydrogeological condition information corresponding to the image information of surrounding rocks of the tunnel faces; 2) All the numerical information in 1) are formed into a numerical information vector set, and the numerical information vector set is expanded to form an expanded numerical information vector set; the expanded numerical information vector set includes: the original numerical information vector and the added numerical information vector; the original numerical information vector refers to the numerical information vector composed of the numerical information in 1), and the added numerical information vector refers to the vector added after the expansion process except the original numerical information vector; the method for expanding the numerical information vector set is as follows: a) extracting each element information of the numerical information in 1) to form a set corresponding to each element information; b) observing whether there is an obvious fault in the data corresponding to each element information, and inserting the data with obvious fault by equidistant method or directly inserting Method for expanding data; the obvious faults include obvious faults of classified data and obvious faults of continuous data; the obvious faults of classified data refer to that the data of the corresponding element is classified data, and a certain classification is missing in the classified data, and the number of corresponding classifications is obviously less than that of other classifications; the obvious faults of continuous data refer to that the distance between two adjacent data sorted in order of size is obviously greater than the distance between other two adjacent data; c) expanding the data information vector set according to the expanded data of each element and the original data of each element, and the expanded part is the increased numerical information vector; the elements in the increased numerical information vector need to contain the expanded data, and at the same time, the increased numerical information vector needs to conform to the theoretical structural characteristics of the surrounding rock of the tunnel face, and correspond to a classification of the surrounding rock of the tunnel face; 3) The added numerical information vectors are converted into image information of the surrounding rock of the tunnel face through generative AI technology, and the surrounding rock grade corresponding to each data information vector is obtained according to the railway tunnel design specification, that is, the surrounding rock classification corresponding to the image information converted from the numerical information vector is obtained; at the same time, the surrounding rock grade corresponding to each image information is obtained according to the numerical information corresponding to the image information in 1), and the image information in 1) and the newly generated image information through generative AI technology are integrated together to form a complete image information set; 4) The complete image information set is trained by deep learning technology to obtain a deep learning model; the training method is: the image information in the image information set is used as the input of the deep learning technology, the surrounding rock classification corresponding to the image information is used as the output of the deep learning technology, and the deep learning technology is trained; 5) Obtain image information of the surrounding rock of the tunnel face of the target tunnel in real time, input the obtained image information into the deep learning model, and directly obtain the grade of the surrounding rock of the corresponding tunnel face according to the output result of the deep learning model.

2. The method for rapid assessment of surrounding rock classification of a tunnel face based on image recognition according to claim 1 is characterized in that: The target terrain includes: the target terrain refers to the terrain under excavation and the terrain where the grade of the surrounding rock of the tunnel face needs to be determined.

3. The method for rapid assessment of surrounding rock classification of a tunnel face based on image recognition according to claim 1 is characterized in that: The obvious faults of the classified data include: lithology information and hydrogeological information, and the obvious faults of the continuous data include: fracture information, density information and hydrogeological information.

4. The method for rapid assessment of surrounding rock classification of a tunnel face based on image recognition according to claim 1 is characterized in that: The number of corresponding categories is obviously less than other categories, including: the number of corresponding categories is less than half of the smallest number of categories in other categories, then the number of corresponding categories is considered to be obviously less than other categories; the distance between two adjacent data is obviously greater than the distance between other two adjacent data, including: the distance between two adjacent data is twice the longest data among any other two adjacent data, then the distance between two adjacent data is considered to be obviously greater than the distance between other two adjacent data.

5. The method for rapid assessment of surrounding rock classification of a tunnel face based on image recognition according to claim 1 is characterized in that: The added numerical information vector must conform to the theoretical structural characteristics of the tunnel face surrounding rock: the tunnel face surrounding rock formed by natural excavation satisfies the theoretical structural characteristics of the tunnel face surrounding rock.

6. The method for rapid assessment of surrounding rock classification of a tunnel face based on image recognition according to claim 1, characterized in that: The generative AI technology includes: generative adversarial networks GANs, variational autoencoders VAE, autoencoders AE, long short-term memory neural networks LSTM, deep convolutional generation models and self-supervised learning; the increased numerical information vectors are converted into image information of the surrounding rock of the tunnel face through the generative AI technology.

7. The method for rapid assessment of surrounding rock classification of a tunnel face based on image recognition according to claim 1, characterized in that: The step of obtaining the surrounding rock grade corresponding to each data information vector according to the railway tunnel design specification includes: obtaining the BQ value corresponding to each data information vector according to the railway tunnel design specification, and determining the surrounding rock grade corresponding to each data information vector; wherein the calculation formula of BQ is: BQ=100+3Rc+250Kv Among them, BQ is the basic quality index of surrounding rock, Rc is the uniaxial saturated compressive strength, Kv refers to the integrity of the rock mass, and the values ​​of Rc and Kv need to be obtained based on the data in the added numerical information vector; The image information in 1) and the newly generated image information through generative AI technology are integrated together to form a complete image information set, including: the complete image information set divides the grade of the surrounding rock of the tunnel face into six categories, wherein each image information in the complete image information set corresponds to a classification of the surrounding rock of the tunnel face.

8. The method for rapid assessment of surrounding rock classification of a tunnel face based on image recognition according to claim 1, characterized in that: The training of the complete image information set by deep learning technology includes: the deep learning technology includes: convolutional neural network technology CNN, deep residual network ResNet, Inception network and attention mechanism AttentionMechanism, the complete image information set is trained by deep learning technology to obtain a deep learning model; the basis for the completion of model training includes: reaching the number of training cycles and the loss function no longer decreasing.

9. The method for rapid assessment of surrounding rock classification of a tunnel face based on image recognition according to claim 1, characterized in that: The real-time acquisition of image information of the surrounding rock of the tunnel face of the target tunnel includes: acquiring image information of the surrounding rock of the tunnel face of the target tunnel in real time, preprocessing the image information, including: denoising, normalization and standardization; then, using the preprocessed image information as input for deep learning, and determining the grade of the surrounding rock of the tunnel face corresponding to the image information in real time through output.

Images