Image recognition-based full-automatic identification method and device for weathering degree of surrounding rock

Abstract

The present application relates to a kind of based on image recognition's weathered degree of surrounding rock full-automatic identification method and device, the identification method includes the following steps: obtaining the image of the to-be-identified face of hand, the to-be-identified face of hand image is preprocessed and light correction, obtain the image after processing;The image after processing is based on the image segmentation of weathered surrounding rock feature, obtain segmentation image;Based on the pre-constructed surrounding rock weathering degree determination index system, corresponding index value is obtained from the image after processing and segmentation image;Through the quantitative evaluation model of surrounding rock weathering degree obtained by training and the corresponding index value after normalization processing, the quantitative determination result of surrounding rock weathering degree is obtained;Wherein, the quantitative evaluation model of surrounding rock weathering degree is the model based on multiple linear regression method and the surrounding rock weathering degree determination index system is constructed.Compared with prior art, the present application has the advantages of high discrimination accuracy, quantitative determination can be realized and the like.

Description

Technical Field

[0001] The present invention relates to the technical field of tunnel construction, and in particular to a method and device for fully automatically identifying the weathering degree of surrounding rocks based on image recognition. Background Art

[0002] The degree of rock weathering significantly affects the engineering properties of the surrounding rock. The "Highway Tunnel Design Code" and the "Railway Tunnel Design Code" both use the weathering coefficient (i.e., the ratio of the compressive strength of the weathered surrounding rock to the compressive strength of the unweathered surrounding rock) as a quantitative evaluation indicator of the degree of surrounding rock weathering, as shown in Table 1. However, in actual engineering, it is difficult to drill and sample the exposed tunnel face after each excavation and conduct indoor testing.

[0003] Table 1 Classification of surrounding rock weathering degree in Highway Tunnel Design Code

[0004]

[0005] Currently, in tunnel engineering, the degree of weathering of surrounding rocks is often classified using qualitative evaluation methods, primarily based on comprehensive analysis of indicators such as rock color, secondary minerals, development of joints and weathering fissures, degree of integrity, and changes in the rock's physical, mechanical, and hydraulic properties. Due to differences in surrounding rock types and hydrogeological environments, the weathering effects and post-weathering morphological changes on the surrounding rocks vary greatly, making it difficult to establish a unified classification standard. The following problems exist in actual engineering practice:

[0006] (1) The method that relies on the visual identification of geological engineers is highly subjective and easily affected by on-site lighting and humidity. Different geological engineers often have different judgments on the surrounding rock conditions of the same tunnel face. This subjective judgment methodology is not scientific.

[0007] (2) To address the above subjective errors, survey and construction units often form a multi-person inspection team, with multiple geological experts conducting multiple assessments and repeated verifications on the same tunnel face to obtain more accurate division results. However, this increases the workload of the identification work, reduces work efficiency, and results in a huge waste of manpower and funds.

[0008] (3) Based on the quantitative judgment of the specification, indoor compressive strength test or on-site wave velocity ratio test can be used. However, it is difficult to find similar fresh rocks as a control group in the field engineering environment. This quantitative evaluation method is almost not feasible from the implementation perspective.

[0009] During actual tunnel excavation, quantitative evaluation methods such as field sampling and laboratory testing are inefficient and labor-intensive. Consequently, the determination of surrounding rock weathering severity relies on the subjective judgment of geologists based on field rock characteristics. Due to the complex rock conditions surrounding the tunnel face, even engineers with extensive experience in geology struggle to accurately classify the degree of weathering. Furthermore, due to the vague and subjective criteria, different geologists often produce different results on the same tunnel face.

[0010] To address the problems of multiple indicators, vague judgment criteria, and inaccurate judgment results in traditional judgment methods, some researchers have attempted to use deep learning to analyze them, such as patent application CN201910768866.X. However, there are still obvious defects as follows:

[0011] (1) Slow training speed. One problem with neural networks is the training speed. In order to obtain better fitting results, it is often necessary to increase the network layers and the number of neurons in each layer. However, this will lead to a large increase in the parameters to be optimized in the network, resulting in excessive computational effort. Training will take a long time, seriously affecting the model iteration speed. On the other hand, due to the influence of hardware performance, simpler neural network training must be adopted, which in turn reduces the accuracy of the model. Finally, neural networks often need to collect a large number of valid samples to train a model with strong generalization ability, otherwise it is easy to cause the model to overfit.

[0012] (2) Poor interpretability. For the constructed multi-layer neural network, the input and output of the intermediate neurons are not regular, and the feature knowledge extracted by each layer of the network is difficult to understand. The multi-layer neural network can only be used as a black box. The training results are often unclear. On the one hand, it is difficult to make further adjustments and optimizations based on the training model. On the other hand, it is impossible to study the relationship between the surrounding rock indicators and further quantify the indicators. Finally, the black box system trained based on the neural network is difficult for on-site personnel to understand and apply. Summary of the Invention

[0013] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and provide a highly accurate fully automatic identification method and device for the degree of weathering of surrounding rocks based on image recognition, so as to realize quantitative judgment of the degree of weathering of surrounding rocks.

[0014] The purpose of the present invention can be achieved by the following technical solutions:

[0015] A method for automatically identifying the weathering degree of surrounding rocks based on image recognition comprises the following steps:

[0016] Acquire a tunnel face image to be identified, perform preprocessing and illumination correction on the tunnel face image to be identified, and obtain a processed image;

[0017] performing image segmentation on the processed image based on weathered surrounding rock features to obtain a segmented image;

[0018] Based on a pre-established index system for determining the degree of weathering of surrounding rocks, corresponding index values ​​are obtained from the processed image and the segmented image;

[0019] The quantitative evaluation model of surrounding rock weathering degree obtained through training and the corresponding index value after normalization are used to obtain the quantitative judgment result of surrounding rock weathering degree;

[0020] The quantitative evaluation model for the degree of weathering of surrounding rocks is a model constructed based on a multivariate linear regression method and an index system for determining the degree of weathering of surrounding rocks.

[0021] Furthermore, the surrounding rock weathering degree determination index system is constructed based on parameter correlation analysis, including primary indicators and secondary indicators. The primary indicators include color, surrounding rock integrity and face texture. The secondary indicators of color include the first-order moment, second-order moment and third-order moment of the a component and b component in the Lab color space. The secondary indicators of surrounding rock integrity include the number and length of joints and cracks per unit area. The secondary indicators of face texture include texture contrast, correlation, entropy, smoothness and second-order moment.

[0022] Furthermore, the quantitative evaluation model of surrounding rock weathering degree is obtained by training through the following steps:

[0023] Constructing a tunnel face image database, wherein tunnel face image samples in the database are labeled with image tags, wherein the image tags include lithology and weathering degree, and the tunnel face image samples are images obtained after preprocessing and illumination correction of the collected images;

[0024] performing image segmentation on the tunnel face image sample based on weathered surrounding rock features;

[0025] Obtaining the corresponding index value of each sample based on the surrounding rock weathering degree determination index system, and obtaining the model input and output data set of each sample after normalization processing;

[0026] The samples are randomly divided into a training set and a test set, and the quantitative evaluation model of the weathering degree of the surrounding rock is obtained through training.

[0027] Furthermore, the preprocessing includes image denoising and image size unification.

[0028] Furthermore, the image denoising process includes Gaussian low-pass filtering and median filtering.

[0029] Furthermore, the illumination correction is specifically as follows:

[0030] Convert the preprocessed image from RGB color space to Lab color space;

[0031] Separate the Lab color space into three independent channel components, forming L component map, a component map and b component map, where L is the brightness component, a and b are color components and are independent of light brightness, and perform brightness correction on the L component map;

[0032] Extract texture features from the original RGB image to form a feature grayscale image;

[0033] The L component image, a component image, b component image and feature grayscale image are fused.

[0034] Furthermore, the tunnel face images in the database cover various weathering degrees including unweathered, slightly weathered, moderately weathered, strongly weathered, and fully weathered.

[0035] Furthermore, when training the quantitative evaluation model for the degree of surrounding rock weathering, a variety of kernel functions are used to construct an SVM classifier, and the classification accuracy is compared using the training set and the test set to select the optimal kernel function; and a grid partitioning method is used to search for the global optimal hyperparameters to construct an SVM classifier that meets the optimal degree of weathering of the surrounding rock of the tunnel face.

[0036] Furthermore, the quantitative evaluation model of the weathering degree of surrounding rocks is expressed as:

[0037]

[0038] in:

[0039]

[0040]

[0041]

[0042] β1=A1+A2+A3+A4+A5+A6

[0043] β2=A7+A8

[0044] β3=A9+A 10 +A 11 +A 12 +A 13

[0045] Where: c - color index, obtained by combining the six secondary indicators;

[0046] w——surrounding rock integrity index, obtained by combining the two secondary indicators;

[0047] t——Tunnel face texture index, which is obtained by combining the five secondary indicators;

[0048] β1——weight corresponding to color index;

[0049] β2——weight corresponding to surrounding rock integrity index;

[0050] β3——weight corresponding to the face texture index;

[0051] n i ——The index of the corresponding indicator, i = 1, 2, ..., 13;

[0052] x i ——Normalized indicator value.

[0053] The present invention also provides an electronic identification device, comprising:

[0054] one or more processors;

[0055] Memory; and

[0056] One or more programs stored in the memory, the one or more programs comprising instructions for executing the fully automatic recognition method as described above.

[0057] Compared with the existing technology, the present invention constructs a determination index with clear physical meaning and capable of characterizing the degree of weathering of the surrounding rock. Based on this, the corresponding relationship between "surrounding rock characteristics-weathering degree" and the quantitative index value are established, ultimately achieving quantitative determination of the degree of weathering of the surrounding rock, which has the following beneficial effects:

[0058] (1) The present invention first performs image denoising and light intensity correction on the collected tunnel face image materials to eliminate the influence of the complex environment of the construction site on the judgment accuracy to the greatest extent.

[0059] (2) The present invention extracts the color, integrity, and texture features of the tunnel face image, lists a large number of indicators, and then performs parameter correlation analysis on each indicator. Indicators that have no effect on the results are eliminated, and indicators with strong correlation are screened. Finally, the judgment indicators that can characterize the degree of weathering of the surrounding rock are determined, which solves the problem of fuzzy judgment indicators in the traditional judgment process and greatly reduces the model training time.

[0060] (3) The present invention extracts relevant characteristic values ​​based on a large amount of face material and adopts multivariate linear fitting to obtain the expression of the combination of multiple characteristic indicators and the weight of each indicator while meeting the accuracy requirements, thereby accurately realizing the quantitative description of the surrounding rock weathering indicators.

[0061] (4) The present invention adopts traditional machine learning methods. Support vector machines and polynomial fitting methods have good interpretability. Each algorithm gives an optimization goal and can be represented geometrically. It can construct a specific relationship expression between input and output based on a large number of case studies, summarize and quantify the value range of indicators, and quickly realize the quantitative determination of the degree of weathering of the surrounding rock. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 It is a schematic diagram of the process of the present invention;

[0063] Figure 2 This is the image fusion effect diagram;

[0064] Figure 3 Schematic diagram of the model training process of the present invention. DETAILED DESCRIPTION

[0065] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0066] Explanation of terms

[0067] Tunnel face: Tunnel face, also known as tunnel face, is a term used in tunnel construction. It refers to the working face that continuously advances during tunnel excavation (coal mining, mining or tunnel engineering).

[0068] Weathering degree of surrounding rock: It refers to the degree of damage to the rock mass caused by weathering, which includes the degree of disintegration and change of the rock mass and the depth of weathering. It is generally divided into five degrees: unweathered, slightly weathered, moderately weathered, strongly weathered, and fully weathered.

[0069] Example 1

[0070] The present invention provides a fully automatic identification method for the degree of surrounding rock weathering based on image recognition. On-site construction personnel use mobile phones to take photos of the latest tunnel face exposed by excavation. Based on the surrounding rock weathering degree judgment indicators in relevant specifications, a feature index system for digital images of the degree of surrounding rock weathering is established. The tunnel face collected images are subjected to image preprocessing, image segmentation, and image-related feature extraction. The degree of surrounding rock weathering is comprehensively judged based on the extracted features. The individual indicators are quantified based on a large amount of data, and an evaluation model is established to ultimately achieve automated judgment of the degree of surrounding rock weathering at the tunnel face.

[0071] like Figure 1 As shown, the fully automatic identification method of the present invention includes the following steps:

[0072] Step S1, obtaining a tunnel face image to be identified, performing preprocessing and illumination correction on the tunnel face image to be identified, and obtaining a processed image;

[0073] Step S2, performing image segmentation on the processed image based on the characteristics of the weathered surrounding rock to obtain a segmented image;

[0074] Step S3, based on a pre-established surrounding rock weathering degree determination index system, obtaining corresponding index values ​​from the processed image and the segmented image;

[0075] Step S4, obtaining a quantitative determination result of the degree of weathering of the surrounding rock by using the trained quantitative evaluation model of the degree of weathering of the surrounding rock and the corresponding index value after normalization, wherein the quantitative evaluation model of the degree of weathering of the surrounding rock is a model constructed based on the multivariate linear regression method and the indicator system for determining the degree of weathering of the surrounding rock.

[0076] (1) Parameter index correlation analysis and key index determination

[0077] In engineering, the degree of weathering of surrounding rocks is mainly determined by analyzing multiple dimensions such as "color, secondary minerals, joints, development of weathering fissures, integrity, hardness, hydraulic properties, freshness, and surface smoothness."

[0078] However, many of these indicators are correlated. Changes in surrounding rock color primarily result from mineral alteration and the formation of new secondary minerals. From a machine vision perspective, these two indicators can be summarized as a single color indicator. Similarly, the development and integrity of joints, weathering fissures, and fractures can be summarized as indicators of surrounding rock integrity. Rock freshness and surface smoothness or roughness can be summarized as tunnel face texture characteristics. Thus, three primary indicators reflecting weathered surrounding rock are obtained: color, surrounding rock integrity, and tunnel face texture.

[0079] Based on machine vision technology:

[0080] 1) For color feature description, nine secondary indicators, including the first-order moment, second-order moment, and third-order moment of L, a, and b components, can be used for characterization;

[0081] 2) The description of surrounding rock integrity characteristics is mainly based on the joints and weathering cracks. Two secondary indicators can be used: the number of joints and cracks per unit area and the total length of joints and cracks per unit area.

[0082] 3) For the description of the texture characteristics of the palm face, five secondary indicators including texture contrast, correlation, entropy, smoothness and second-order moment can be used for representation.

[0083] The correlation index (Pearson correlation coefficient) was calculated to study the correlation between the above indicators and between the indicators and results. The calculation formula is as follows:

[0084]

[0085] Where: x i ,y i ——The i-th group of data for simultaneous measurement of x and y quantities;

[0086] ——The mean of the corresponding x-value and y-value data;

[0087] n——total number of statistical groups;

[0088] S(x), S(y)——standard deviation of the corresponding x and y data.

[0089] According to the results of the correlation study, the three secondary indicators of the first-order moment, second-order moment, and third-order moment of the image brightness component L that have no effect on the judgment result are eliminated, and the remaining 13 secondary indicators are finally retained as the judgment indicators of the weathering degree of the surrounding rock, as shown in Table 2.

[0090] Table 2 Determination index of surrounding rock weathering degree

[0091]

[0092] (2) Quantitative evaluation model of surrounding rock weathering degree

[0093] The quantitative evaluation model for the degree of weathering of surrounding rocks used in this method is a model constructed based on the multivariate linear regression method and the surrounding rock weathering degree determination index system, and is obtained through the following training steps:

[0094] 11) Constructing a tunnel face image database, wherein tunnel face image samples in the database are labeled with image tags, including lithology and weathering degree, and the tunnel face image samples are images obtained after preprocessing and illumination correction of the collected images.

[0095] In this embodiment, captured images were taken by engineers using mobile phones at a distance directly in front of the tunnel face. Fill lighting was used during the capture process to ensure that the tunnel face image was free of shadows. The captured images were designed to cover as many weathering levels as possible, including unweathered, slightly weathered, moderately weathered, strongly weathered, and fully weathered. Each photo was labeled with "lithology + weathering degree."

[0096] Image preprocessing primarily involves image denoising and image resizing. Image noise can be categorized as Gaussian and impulse noise due to the harsh tunnel construction environment, including dust, light source interference, and interference from the camera itself. First, Gaussian low-pass filtering is used to eliminate Gaussian noise, followed by median filtering to eliminate impulse noise. Finally, the filtered tunnel face image undergoes pixel transformation, image resizing, and format standardization.

[0097] In this embodiment, the illumination correction specifically includes:

[0098] (101) Color space transformation

[0099] The RGB color data of the pre-processed image directly captured by the CMOS camera is transformed through a two-step matrix transformation. First, the RGB color space is converted to the CIEL XYZ color space:

[0100]

[0101] Then convert the CIEL XYZ color space to the LAB color space:

[0102]

[0103]

[0104]

[0105] (102) Illumination correction based on single-scale Retinex enhancement algorithm

[0106] Using the Retinex algorithm, the image is defined as the product of the reflected light image and the incident light image, which can be expressed as follows:

[0107] S(x,y)=R(x,y)·L(x,y) (5)

[0108] Where S(x,y) is the original image, R(x,y) is the reflected light image, which contains the essential properties of the image, and L(x,y) is the incident light image, which contains the dynamic range of the image. Based on the above theory, the image illumination intensity can be enhanced and corrected by adjusting or eliminating the incident light image. Expressed in logarithmic form:

[0109] log(R(x,y))=log(S(x,y))-log(L(x,y)) (6)

[0110] By performing convolution calculation on the Gaussian function and the original image, the incident light image can be obtained. The calculation method is as follows:

[0111] L(x,y)=F(x,y)*S(x,y) (7)

[0112] Where F(x,y) is the Gaussian surround function, defined as:

[0113]

[0114] Where c represents the Gaussian surround scale, λ is a constant such that the integral of F(x,y) is 1. Substituting equation (7) into equation (6), we can obtain the image correction algorithm:

[0115] r(x,y)=log(R(x,y))=log(S(x,y))-log[F(x,y)*S(x,y)] (9)

[0116] (103) Feature extraction and image fusion

[0117] The canny edge detection operator is used to extract the image texture information from the original RGB image and generate a grayscale image. In order to avoid the halo, color distortion and detail loss caused by traditional algorithm processing, the fusion idea is adopted to fuse the corrected L component image, the separated a and b component images and the texture feature grayscale image to generate a light intensity correction image that retains details and colors, such as Figure 2 shown.

[0118] After the above processing, the pre-processed image is subjected to light brightness correction in the Lab color space and fused with the image texture features to ensure that the image brightness is uniform while retaining details and without color distortion, which is more in line with the human eye perception.

[0119] 12) Performing image segmentation on the tunnel face image sample based on weathered surrounding rock features.

[0120] Traditional color image segmentation, based on color gradient changes, is less suitable for weathered surrounding rock. To accurately segment joints and fissures, extract their number and statistically analyze their lengths, and target surrounding rock characteristics (discoloration and fragmentation), a comprehensive segmentation approach based on traditional color image segmentation methods and surrounding rock joint characteristics is employed. Traditional color image region segmentation methods based on color differences are modified by combining the joint, fissure, and texture characteristics of the tunnel face to achieve an image region segmentation that better reflects the characteristics of weathered surrounding rock, facilitating subsequent feature extraction.

[0121] 13) Obtain the corresponding index value of each sample based on the surrounding rock weathering degree determination index system, and obtain the model input and output data set of each sample after normalization.

[0122] For multi-index data, normalization is used to process various types of data into dimensionless values ​​within a unified range. The normalization mapping is as follows:

[0123]

[0124] Where: x min ——The minimum value min(x) in the original data set;

[0125] x max ——The maximum value max(x) in the original data set;

[0126] x——original data;

[0127] y – normalized data.

[0128] The surrounding rock samples are marked according to the degree of weathering, where "unweathered" is marked as "1", "slightly weathered" is marked as "2", "moderately weathered" is marked as "3", "strongly weathered" is marked as "4", and "fully weathered" is marked as "5". The input and output data sets can be expressed as:

[0129] {x1,x2,x3,...,x 13 ,y}

[0130] Above: x1 - the calculated value of the first-order moment of the normalized a component;

[0131] x2——the normalized second-order moment calculation value of component a;

[0132] x3——the calculated value of the third-order moment of the normalized a component;

[0133] x4——the calculated value of the first-order moment of the normalized b component;

[0134] x5——normalized calculated value of the second-order moment of the b component;

[0135] x6——normalized calculated value of the third-order moment of the b component;

[0136] x7——the calculated value of the number of joints and cracks per unit area after normalization;

[0137] x8——the calculated value of the total length of joints and cracks per unit area after normalization;

[0138] x9——normalized texture contrast calculation value;

[0139] x 10 ——Normalized texture correlation calculation value;

[0140] x 11 ——Normalized texture entropy calculation value;

[0141] x 12 ——The normalized texture stability calculation value;

[0142] x 13 ——The normalized calculated value of the texture second-order moment;

[0143] y——image label value, which is the output value of the training model, and the values ​​are 1, 2, 3, 4, and 5.

[0144] 14) Randomly grouping the samples into training sets and test sets, and training to obtain the quantitative evaluation model for the weathering degree of the surrounding rock.

[0145] Image samples were divided into training and test sets in a ratio of 7 to 3. The 13 characteristic parameters and label parameters that characterize the degree of tunnel face surrounding rock weathering were used as input parameters. Several commonly used kernel functions were used to construct a support vector machine (SVM) classifier. Classification accuracy was compared between the training and test sets to select the optimal kernel function. A grid partitioning method was used to search for the globally optimal values ​​of the hyperparameters C and γ, ultimately constructing an SVM classifier that optimally matches the degree of tunnel face surrounding rock weathering. Based on a large number of image feature parameters, a polynomial fitting algorithm was employed to fit the relationship between the feature quantities and the output variables to a certain degree of accuracy. The combined relationship between each primary indicator and its subordinate secondary indicators was trained, along with the weights of each primary indicator, resulting in a quantitative expression for the degree of surrounding rock weathering. Furthermore, the polynomial expression was used to determine the calculation methods and corresponding weights for the three indicators (color, surrounding rock integrity, and tunnel face texture), and the parameter value ranges were summarized and divided. Finally, a quantitative evaluation model for the degree of surrounding rock weathering was established, enabling quantitative evaluation of the degree of surrounding rock weathering.

[0146] In this embodiment, the multivariate linear regression method is used to establish the model, and the highest degree n in the polynomial and the training accuracy threshold ε are set. When the deviation between the model output result and the label value is within the allowable accuracy ε, the model training is completed. The model is verified using a test set. If the trained model has low prediction accuracy on the test set, the highest degree n of the polynomial is readjusted until a model with good performance on both the training set and the test set is trained. The model training process is as follows: Figure 3 shown.

[0147] In this embodiment, the final constructed model is as follows:

[0148]

[0149] in:

[0150]

[0151]

[0152]

[0153] β1=A1+A2+A3+A4+A5+A6

[0154] β2=A7+A8

[0155] β3=A9+A 10 +A 11 +A 12 +A 13

[0156] Where: c is the color index parameter, which is obtained by combining the six secondary indicators;

[0157] w——surrounding rock integrity index, obtained by combining the two secondary indicators;

[0158] t——Tannel face texture index, which is obtained by combining the five subordinate secondary indicators.

[0159] β1——weight corresponding to color index;

[0160] β2——weight corresponding to surrounding rock integrity index;

[0161] β3——weight corresponding to the face texture index;

[0162] n i ——Index of the corresponding indicator.

[0163] Based on the existing tunnel face samples, the above method is used for processing, the model fitting accuracy is set to 50%, and the index of each indicator is 1 (n i =1), a corresponding surrounding rock weathering degree index determination model can be established, as shown in Table 3.

[0164] in:

[0165] (1) Color index: It represents the area and degree of alteration of the surrounding rock at the tunnel face and has actual physical significance. The larger the index value, the larger the weathering and alteration area of ​​the tunnel face and the deeper the weathering degree.

[0166] (2) The surrounding rock integrity index characterizes the development of joints and fissures on the tunnel face and has practical physical significance. The larger the index value, the more broken the surrounding rock on the tunnel face and the deeper the weathering degree.

[0167] (3) The tunnel face texture index characterizes the roughness of the surrounding rock surface and has actual physical significance. The larger the index value, the rougher the tunnel face and the deeper the weathering degree.

[0168] (4) The value ranges of each judgment indicator in Table 3 are obtained based on the current sample training and summary. When the sample size increases, the value range can be revised until the sample size reaches a certain level and the value range tends to be stable.

[0169] Table 3. Determination model of surrounding rock weathering degree index

[0170]

[0171] Note: In order to simplify the calculation, the following processing measures are taken in this table:

[0172] (1) The model accuracy is set to 50%, but the weathering index F takes a continuous value;

[0173] (2) Set the index n of each secondary indicator i is 1, the indicator combination becomes a linear combination;

[0174] (3) weight the color index, surrounding rock integrity index, and tunnel face texture index so that the sum of the weights is 1;

[0175] (4) Under the above processing conditions, a calculation result with a value interval length of 1 is generated.

[0176] If the above method is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, and other media that can store program code.

[0177] Example 2

[0178] This embodiment provides an electronic identification device, including one or more processors, a memory, and one or more programs stored in the memory, wherein the one or more programs include instructions for executing the fully automatic identification method described in Example 1.

[0179] The above describes in detail the preferred embodiments of the present invention. It should be understood that those skilled in the art can make numerous modifications and variations based on the concepts of the present invention without inventive effort. Therefore, any technical solutions that can be derived by those skilled in the art through logical analysis, reasoning, or limited experimentation based on the concepts of the present invention and the prior art should be within the scope of protection defined by the claims.

Claims

1. A fully automatic identification method for the weathering degree of surrounding rocks based on image recognition, characterized in that: The following steps are involved: Acquire a tunnel face image to be identified, perform preprocessing and illumination correction on the tunnel face image to be identified, and obtain a processed image; performing image segmentation on the processed image based on weathered surrounding rock features to obtain a segmented image; Based on a pre-established index system for determining the degree of weathering of surrounding rocks, corresponding index values ​​are obtained from the processed image and the segmented image; The quantitative evaluation model of surrounding rock weathering degree obtained through training and the corresponding index value after normalization are used to obtain the quantitative judgment result of surrounding rock weathering degree; The quantitative evaluation model for the degree of weathering of surrounding rocks is a model constructed based on a multivariate linear regression method and the indicator system for determining the degree of weathering of surrounding rocks. The surrounding rock weathering degree determination index system is constructed based on parameter correlation analysis and includes primary and secondary indicators. The primary indicators include color, surrounding rock integrity, and tunnel face texture. The secondary indicators of color include the first-order moment, second-order moment, and third-order moment of the a component and b component in the Lab color space. The secondary indicators of surrounding rock integrity include the number and length of joints and fissures per unit area. The secondary indicators of tunnel face texture include texture contrast, correlation, entropy, stationarity, and second-order moment.

2. The method for automatically identifying the weathering degree of surrounding rocks based on image recognition according to claim 1 is characterized in that: The quantitative evaluation model of surrounding rock weathering degree is obtained by training through the following steps: Constructing a tunnel face image database, wherein tunnel face image samples in the database are labeled with image tags, wherein the image tags include lithology and weathering degree, and the tunnel face image samples are images obtained after preprocessing and illumination correction of the collected images; performing image segmentation on the tunnel face image sample based on weathered surrounding rock features; Obtaining the corresponding index value of each sample based on the surrounding rock weathering degree determination index system, and obtaining the model input and output data set of each sample after normalization processing; The samples are randomly divided into a training set and a test set, and the quantitative evaluation model of the weathering degree of the surrounding rock is obtained through training.

3. The method for automatically identifying the weathering degree of surrounding rocks based on image recognition according to claim 1 or 2, characterized in that: The preprocessing includes image denoising and image size unification.

4. The method for automatically identifying the weathering degree of surrounding rocks based on image recognition according to claim 3 is characterized in that: The image denoising process includes Gaussian low-pass filtering and median filtering.

5. The method for automatically identifying the weathering degree of surrounding rocks based on image recognition according to claim 1 or 2, characterized in that: The illumination correction is specifically as follows: Convert the preprocessed image from RGB color space to Lab color space; Separate the Lab color space into three independent channel components, forming L component map, a component map and b component map, where L is the brightness component, a and b are color components and are independent of light brightness, and perform brightness correction on the L component map; Extract texture features from the original RGB image to form a feature grayscale image; The L component image, a component image, b component image and feature grayscale image are fused.

6. The method for automatically identifying the weathering degree of surrounding rocks based on image recognition according to claim 2 is characterized in that: The tunnel face images in the database cover various weathering degrees including unweathered, slightly weathered, moderately weathered, strongly weathered, and fully weathered.

7. The method for automatically identifying the weathering degree of surrounding rocks based on image recognition according to claim 2 is characterized in that: When training the quantitative evaluation model for the degree of surrounding rock weathering, a variety of kernel functions are used to construct an SVM classifier. The classification accuracy is compared between the training set and the test set to select the optimal kernel function. A grid partitioning method is then used to search for the global optimal hyperparameters to construct an SVM classifier that meets the optimal degree of weathering of the surrounding rock of the tunnel face.

8. The method for automatically identifying the weathering degree of surrounding rocks based on image recognition according to claim 1 is characterized in that: The quantitative evaluation model of surrounding rock weathering degree is expressed as: in: β1=A1+A2+A3+A4+A5+A6 β2=A7+A8 β3=A9+A 10 +A 11 +A 12 +A 13 Where: c - color index, obtained by combining the six secondary indicators; w——surrounding rock integrity index, obtained by combining the two secondary indicators; t——Tunnel face texture index, which is obtained by combining the five secondary indicators; β1——weight corresponding to color index; β2——weight corresponding to surrounding rock integrity index; β3——weight corresponding to the face texture index; n i ——The index of the corresponding indicator, i = 1, 2, ..., 13; x i ——Normalized indicator value.

9. An electronic identification device, characterized in that: include: one or more processors; Memory; and One or more programs stored in a memory, wherein the one or more programs include instructions for executing the fully automatic recognition method according to any one of claims 1 to 8.

Images