Rock mass quality evaluation method, system and device based on hyperspectral imaging technology
By combining hyperspectral imaging technology and deep learning models, real-time, rapid and comprehensive evaluation of tunnel rock mass quality has been achieved, solving the problems of inaccurate evaluation results and time-consuming and laborious processes in traditional methods, and improving the efficiency and accuracy of evaluation.
Patent Information
- Application Number
- CN202410469819.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-04-18
- Publication Date
- 2024-11-01
- Estimated Expiration
- 2044-04-18
AI Technical Summary
Existing technologies are insufficient for quickly and accurately evaluating the quality of tunnel rock mass, especially since they cannot effectively combine rock mass structural and strength characteristics, resulting in inaccurate evaluation results and being time-consuming and labor-intensive.
Hyperspectral imaging technology is used to acquire images and spectral information of the tunnel face in a non-contact manner. Deep learning models are used to extract rock mass structure and strength characteristics, and fusion analysis is performed to achieve real-time and rapid rock mass quality evaluation.
It enables real-time, rapid, and comprehensive evaluation of tunnel rock mass quality, avoiding sampling bias and the time-consuming and labor-intensive problems of traditional methods, and improving the accuracy and efficiency of the evaluation.
Smart Images

Figure CN118503905B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of tunnel rock mass quality evaluation, and in particular relates to a rock mass quality evaluation method, system and device based on hyperspectral imaging technology. Background Art
[0002] Rock mass quality evaluation can generally reflect the quality of various types of engineering rock masses in underground engineering construction, predict possible rock mass mechanics problems, and to a certain extent provide a basis and guiding opinions for engineering design, construction methods, lining support, etc., to ensure construction safety to the greatest extent, which is of great significance in avoiding casualties and property losses.
[0003] Rock mass quality evaluation mainly involves classifying rock mass by linking engineering geological conditions with rock mass mechanical property parameters through some rock mass indicators, such as the spatial distribution of rock mass discontinuities and strength characteristics such as mechanical properties of rock mass. The development of air-space-ground remote sensing technology has made it possible to detect the geometric shape and discontinuities of surface rocks, but it is not possible to accurately detect the structural characteristics of rock mass at the tunnel face.
[0004] In addition, the existing technology mainly uses on-site in-situ testing and laboratory testing to obtain the strength characteristic indicators of tunnel rock mass. The mechanical properties of in-situ rock mass are more complex than those of rock blocks. In addition to being affected by the mechanical properties of the rock blocks themselves, they are also affected and controlled by rock mass structure and environmental factors. Laboratory testing requires multiple processes such as on-site sampling, sample transportation, sample preparation, and laboratory testing. It is expensive and time-consuming, and the samples obtained on-site cannot accurately represent the rock mass strength of the entire face. At present, there are also some studies that use digital image processing technology and deep learning methods to interpret the geological information parameters of the rock mass structure surface contained in the image and establish a rapid rock mass quality evaluation system, but this evaluation method rarely considers the influence of rock mass strength and has certain limitations. Summary of the invention
[0005] In order to overcome the deficiencies of the above-mentioned prior art, the present invention provides a rock quality evaluation method, system and device based on hyperspectral imaging technology. By performing non-contact scanning and testing on the target rock mass, the structural characteristic parameters and strength characteristic parameters of the hyperspectral three-dimensional data are fused and analyzed, so as to realize real-time, rapid and comprehensive evaluation and grading of the rock quality of the face.
[0006] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions:
[0007] A first aspect of the present invention provides a rock mass quality evaluation method based on hyperspectral imaging technology.
[0008] The rock mass quality evaluation method based on hyperspectral imaging technology includes the following steps:
[0009] Collect the original image information and spectral information of the target rock tunnel face, and perform pre-processing and gridding;
[0010] The gridded image information and spectral information are respectively input into the image and rock mass structure feature deep learning model and the spectrum and rock mass strength feature deep learning model to extract the rock mass structure feature and rock mass strength feature;
[0011] The rock mass structure characteristics and rock mass strength characteristics are input into the fused image spectrum-rock mass quality evaluation model to perform feature dimension reduction and multi-feature vector fusion. The fused features are compressed using an activation function and the rock mass quality grade is classified using a Logistic regression classifier to obtain the rock mass quality evaluation grade within a single grid.
[0012] The rock mass quality evaluation and classification conditions of each grid are integrated and the rock mass quality evaluation results are output.
[0013] Optionally, the deep learning model of image and rock structure characteristics and the deep learning model of spectrum and rock strength characteristics are both convolutional neural network models, which initially extract features through the convolution layer, extract main features through the pooling layer, and perform multiple rounds of convolution and pooling. Finally, the fully connected layer summarizes the features of each part to achieve the extraction of features related to cracks and rock strength.
[0014] Optionally, extract rock mass structural features, including:
[0015] The preprocessed image information is segmented based on the SLIC superpixel segmentation algorithm to obtain superpixel blocks of the image;
[0016] Calculate the merging weights between adjacent superpixel blocks and construct a regional adjacency graph;
[0017] Execute the hierarchical region merging algorithm to merge the regional adjacency graphs, and transform the contour of the merged result into a joint and fissure distribution map through coordinate transformation;
[0018] The crack distribution map is input into the trained image and rock mass structural characteristics deep learning model to obtain the structural characteristics of the rock mass.
[0019] Optionally, extract rock mass strength characteristics, including:
[0020] Based on the different mineral compositions of different rock masses and the changes in the spectral radiation brightness and band characteristics of the same rock mass under different stresses, a database on the relationship between spectrum and rock strength is constructed;
[0021] After performing first-order differential and continuum removal transformation on the divided spectral information, characteristic wavelengths are selected;
[0022] Based on the characteristic wavelength and the constructed database of the relationship between spectrum and rock strength, the spectrum information matching method is adopted to obtain the relevant information of rock mass strength;
[0023] The spectrum and rock strength feature deep learning model is used to transform the relevant information of rock strength into the strength characteristics of the rock mass.
[0024] Optionally, image preprocessing mainly includes graying, enhancing image brightness and contrast, image sharpening, and median filtering; spectral preprocessing includes spectrum smoothing, derivative, normalization, wavelet transform, and convolution smoothing based on the least squares principle.
[0025] Optionally, the gridding process specifically includes:
[0026] Set the mesh density according to the actual project needs;
[0027] Identify the grid boundaries where the number of rock mass cracks and joints and spectral information change significantly;
[0028] The mesh is divided based on the mesh density and the identified mesh boundaries.
[0029] Optionally, non-contact scanning and testing is performed on the target rock mass to collect original image information and spectral information of the tunnel face of the target rock mass.
[0030] A second aspect of the present invention provides a rock quality evaluation system based on hyperspectral imaging technology.
[0031] The rock mass quality evaluation system based on hyperspectral imaging technology includes:
[0032] The information acquisition module is configured to: collect original image information and spectral information of the target rock tunnel face, and perform pre-processing and gridding processing;
[0033] The feature extraction module is configured to: input the gridded image information and spectral information into the image and rock mass structure feature deep learning model and the spectrum and rock strength feature deep learning model respectively, and extract the rock mass structure feature and rock mass strength feature;
[0034] The fusion analysis module is configured to: input rock mass structure features and rock mass strength features into the fusion constructed image spectrum-rock mass quality evaluation model, perform feature dimension reduction and multi-feature vector fusion, compress the fused features using an activation function, and use a Logistic regression classifier to classify the rock mass quality level, and obtain the rock mass quality evaluation classification within a single grid;
[0035] The integration module is configured to integrate the rock mass quality evaluation and classification conditions of each grid and output the rock mass quality evaluation results.
[0036] A third aspect of the present invention provides a rock mass quality evaluation device based on the rock mass quality evaluation system based on hyperspectral imaging technology described in the second aspect.
[0037] The rock mass quality evaluation device based on hyperspectral imaging technology includes a carrying platform, a data acquisition device, an auxiliary device and a main control module. The data acquisition device includes an imaging spectrometer, a camera, a first lifting platform and a slide rail. The auxiliary device includes a full-spectrum halogen lamp and a second lifting platform:
[0038] The slide rail is arranged on the top of the carrying platform, the first lifting platform is slidably connected to the slide rail, the imaging spectrometer and the camera are arranged on the top of the first lifting platform; the second lifting platform is arranged on the top of the carrying platform, and the full-spectrum halogen lamp is arranged on the top of the second lifting platform; the bottom of the carrying platform is also provided with track wheels and a track wheel driving mechanism;
[0039] The main control module is electrically connected to the first lifting platform, the second lifting platform, the imaging spectrometer, the camera, the full-spectrum halogen lamp and the track wheel driving mechanism respectively.
[0040] Optionally, it also includes a data processing module, a data storage module, and a result display module, wherein:
[0041] The data processing module is electrically connected to the camera, the imaging spectrometer, and the result display module, respectively, receives the original image information and spectral information of the target rock mass tunnel face sent back by the camera and the imaging spectrometer, processes the information, and obtains the rock mass quality evaluation result;
[0042] The result display module is electrically connected to the data processing module to display the rock mass quality evaluation results;
[0043] The data storage module is electrically connected to the data processing module, the camera and the imaging spectrometer respectively, and stores the rock mass quality evaluation result, the original image information and the spectral information of the target rock mass tunnel face.
[0044] One or more of the above technical solutions have the following beneficial effects:
[0045] (1) The present invention provides a rock mass quality evaluation method, system and device based on hyperspectral imaging technology, which introduces imaging hyperspectral technology, obtains the structural characteristics and strength characteristics of the rock mass, and fuses and analyzes the structural characteristic parameters and strength characteristic parameters of the hyperspectral three-dimensional data, so as to realize real-time, rapid and comprehensive evaluation and classification of the rock mass quality of the face, thereby avoiding misjudgment caused by evaluating the rock mass quality using a single factor.
[0046] (2) Compared with the traditional rock quality evaluation method which requires sampling, block making, testing and other processes, the present invention uses imaging hyperspectral technology to photograph, scan and analyze the rock mass at the tunnel face in a non-contact manner, which simplifies the testing process and can quickly and efficiently obtain the rock quality status.
[0047] (3) Use imaging hyperspectral technology to scan and analyze the entire rock mass to obtain all the information about the rock mass at the face, effectively avoiding the problem that the samples cannot accurately represent the rock mass information at the face due to sampling bias during on-site sampling.
[0048] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS
[0049] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.
[0050] Figure 1 Schematic diagram of the rock quality evaluation method based on hyperspectral imaging technology.
[0051] Figure 2 Schematic diagram of the data flow of the image spectrum-rock quality evaluation model.
[0052] Figure 3 Schematic diagram of the rock quality evaluation device based on hyperspectral imaging technology.
[0053] In the figure:
[0054] 1-main control module; 2-full spectrum halogen lamp; 3-imaging spectrometer; 4-camera; 5-lens; 601-first lifting platform; 602-second lifting platform; 7-carrying platform; 8-slide rail; 9-power supply; 10-track wheel. DETAILED DESCRIPTION
[0055] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.
[0056] It should be noted that the terms used herein are for describing specific embodiments only and are not intended to be limiting of exemplary embodiments according to the present invention.
[0057] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.
[0058] Embodiment 1
[0059] This embodiment discloses a rock mass quality evaluation method based on hyperspectral imaging technology.
[0060] like Figure 1 , Figure 2 As shown, the rock mass quality evaluation method based on hyperspectral imaging technology includes the following steps:
[0061] Collect the original image information and spectral information of the target rock tunnel face, and perform pre-processing and gridding;
[0062] The gridded image information and spectral information are respectively input into the image and rock mass structure feature deep learning model and the spectrum and rock mass strength feature deep learning model to extract the rock mass structure feature and rock mass strength feature;
[0063] The rock mass structure characteristics and rock mass strength characteristics are input into the fused image spectrum-rock mass quality evaluation model to perform feature dimension reduction and multi-feature vector fusion. The fused features are compressed using an activation function and the rock mass quality grade is classified using a Logistic regression classifier to obtain the rock mass quality evaluation grade within a single grid.
[0064] The rock mass quality evaluation and classification conditions of each grid are integrated and the rock mass quality evaluation results are output.
[0065] More specifically, they include:
[0066] (1) Pretreatment
[0067] The preprocessing of information obtained from the data acquisition module includes image preprocessing and spectral preprocessing. Image preprocessing mainly includes four steps: grayscale, enhancing image brightness and contrast, image sharpening, and median filtering. A high-resolution face image is obtained, which provides a basis for accurately extracting the characteristics of image cracks. There are many spectral preprocessing methods, such as smoothing, derivative, normalization, wavelet transform, etc. The more commonly used method is the convolution smoothing method based on the least squares principle proposed by Savitzky-Golay, which removes redundant information in the original spectrum that will interfere with the spectral information and improves the accuracy and performance of spectral processing.
[0068] (2) Grid division
[0069] The complete tunnel face area after pretreatment is divided into grids, and the rock quality of the tunnel face in the grid is evaluated by region. The density of the grid division directly affects the speed of rock quality evaluation. The grid division density can be set according to the needs of the actual project. At the same time, the grid boundaries where the number of rock cracks and joints and spectral information changes significantly are identified, and the grid is re-divided for further analysis.
[0070] (3) Image and spectral information analysis
[0071] The superpixel blocks of the collected and processed images are obtained based on the SLIC superpixel segmentation algorithm, the merging weights between adjacent superpixel blocks are calculated, the regional adjacency graph is constructed and the hierarchical regional merging algorithm is executed, the contour of the merged result is transformed into the final joint and crack distribution map by coordinate conversion, and the crack distribution map is further input into the trained deep learning model to obtain the structural characteristics of the rock mass;
[0072] The first-order differential and continuum removal transformations are performed on the divided spectral information to select characteristic wavelengths. Based on the spectral-rock strength database constructed with a large amount of spectral test data in the early stage and the deep learning model of spectral and rock strength characteristics, the spectral information matching method is used to obtain the relevant information of rock strength and integrate it into rock strength characteristics.
[0073] The rock mass structural characteristics and strength characteristics are further input into the fused image spectrum-rock mass quality evaluation model to obtain the rock mass quality evaluation information of each grid and integrate and visualize it.
[0074] like Figure 2 As shown in the figure, since the spectral information obtained by hyperspectral imaging is a digital image with multi-spectral resolution, the deep learning model of image and rock structure characteristics and the deep learning model of spectrum and rock strength characteristics use the most advanced convolutional neural network model (CNN) in the current computer vision field. The convolution layer initially extracts features, the pooling layer extracts the main features, and multiple rounds of convolution and pooling are performed. Finally, the fully connected layer summarizes the features of each part to achieve the extraction of features related to cracks and rock strength.
[0075] The fused image spectrum-rock quality evaluation model is a comprehensive processing of the above information after preprocessing and feature extraction. The model adopts a multi-CNN feature dimension reduction fusion algorithm, which mainly includes the following processes: the classification layer is removed from the previous two convolutional neural network models, and the feature vectors of rock structure and rock strength are extracted respectively. Then, each feature is reduced to a feature vector with a length of N for multi-feature vector fusion. The sigmoid activation function is used on the fused features to obtain the final result P, and then the Logistic regression classifier is used to realize the final rock quality grade classification.
[0076] (4) Result output
[0077] The information integrated in (3) is output to obtain a visualized rock mass quality evaluation result.
[0078] For various types of rocks with different strengths, hyperspectral imaging tests were carried out under the same conditions to obtain a large number of images and spectral information libraries of rocks with different strengths. Further preprocessing was used to eliminate irrelevant information in the image spectrum, restore useful real information, simplify data and enhance information monitorability, obtain high-resolution tunnel face image dimensional information, and remove redundant information in the original spectrum that would interfere with the spectral information, thereby improving the accuracy and performance of spectral processing.
[0079] The mineral composition of different rock masses is obviously different. The image hyperspectral technology can be used to identify the mineral composition of the rock mass, porosity and other factors that affect the rock strength, as well as the degree of fragmentation and crack distribution of the rock mass at the face. The corresponding image spectral characteristics will be different.
[0080] Based on the above analysis, a deep learning model of image information and rock structure characteristics and a deep learning model of spectral information and rock strength characteristics are established to realize the extraction of structural features such as cracks and rock strength-related features respectively.
[0081] On this basis, the structural characteristics and intensity characteristics of the hyperspectral image are fused and analyzed to construct an image spectrum-rock quality evaluation model. A large amount of fusion training is carried out using the image spectrum data acquired in the early stage. At the same time, the rock image spectrum information under different rock qualities is continuously acquired to optimize the effect of deep learning.
[0082] Due to different tunnel environments, the tunnel faces and target rock mass ranges to be identified and judged are different, and there may be different rock mass quality levels in different areas of the same tunnel face. To address this problem, the present invention adopts the form of grid division to divide the target rock mass into a certain number of grids and evaluate the rock mass quality separately.
[0083] When using this system, the image spectrum information of the target rock mass is obtained and preprocessed, and further gridded. After processing, the collected rock mass images are subjected to the trained deep learning model to obtain the structural characteristics of the rock mass. Different rock masses have different mineral compositions, and the types of mineral components and the amount of each mineral content affect the strength of the rock. At the same time, the spectral radiation brightness and band characteristics of the same rock mass also change under different stresses. According to the database of the relationship between spectrum and rock strength to be constructed, the deep learning model of spectrum and rock strength characteristics is used to convert the spectral characteristics into the strength characteristics of the rock mass. The image spectrum-rock quality evaluation model constructed by fusion is input to evaluate and grade the rock quality in a single grid, and finally the conditions of each grid are integrated and output as the rock quality evaluation results.
[0084] Embodiment 2
[0085] This embodiment discloses a rock quality evaluation system based on hyperspectral imaging technology.
[0086] The rock mass quality evaluation system based on hyperspectral imaging technology includes:
[0087] The information acquisition module is configured to: collect original image information and spectral information of the target rock tunnel face, and perform pre-processing and gridding processing;
[0088] The feature extraction module is configured to: input the gridded image information and spectral information into the image and rock mass structure feature deep learning model and the spectrum and rock strength feature deep learning model respectively, and extract the rock mass structure feature and rock mass strength feature;
[0089] The fusion analysis module is configured to: input rock mass structure features and rock mass strength features into the fusion constructed image spectrum-rock mass quality evaluation model, perform feature dimension reduction and multi-feature vector fusion, compress the fused features using an activation function, and use a Logistic regression classifier to classify the rock mass quality level, and obtain the rock mass quality evaluation classification within a single grid;
[0090] The integration module is configured to integrate the rock mass quality evaluation and classification conditions of each grid and output the rock mass quality evaluation results.
[0091] Embodiment 3
[0092] The purpose of this embodiment is to provide a rock mass quality evaluation device based on hyperspectral imaging technology. As shown in the figure, the rock mass quality evaluation device based on hyperspectral imaging technology includes a carrying platform 7, a data acquisition device, an auxiliary device and a main control module 1. The data acquisition device includes an imaging spectrometer 3, a camera 4, a first lifting platform 601 and a slide rail 8. The auxiliary device includes a full-spectrum halogen lamp 2 and a second lifting platform 602:
[0093] The slide rail 8 is arranged on the top of the carrying platform 7, the first lifting platform 601 is slidably connected to the slide rail 8, the imaging spectrometer 3 and the camera 4 are arranged on the top of the first lifting platform 601; the second lifting platform 602 is arranged on the top of the carrying platform 7, and the full-spectrum halogen lamp 2 is arranged on the top of the second lifting platform 602; the bottom of the carrying platform 7 is also provided with a track wheel 10 and a track wheel 10 driving mechanism;
[0094] The main control module 1 is electrically connected to the first lifting platform 601, the second lifting platform 602, the imaging spectrometer 3, the camera 4, the full-spectrum halogen lamp 2 and the driving mechanism of the track wheel 10 respectively.
[0095] It also includes a data processing module, a data storage module, and a result display module, wherein:
[0096] The data processing module is electrically connected to the camera 4, the imaging spectrometer 3, and the result display module, respectively, receives the original image information and spectral information of the target rock mass tunnel face transmitted back by the camera 4 and the imaging spectrometer 3, processes the information, and obtains the rock mass quality evaluation result;
[0097] The result display module is electrically connected to the data processing module to display the rock mass quality evaluation results;
[0098] The data storage module is electrically connected to the data processing module, the camera 4 and the imaging spectrometer 3 respectively, and stores the rock mass quality evaluation results, the original image information and the spectral information of the target rock mass tunnel face.
[0099] More specifically, this embodiment includes a data acquisition module, an auxiliary module, a data processing module, a data storage module, a result display module, a main control module 1 , a mounting platform 7 and a power supply 9 .
[0100] The data acquisition module is an imaging spectrometer 3 and its supporting devices, which are used to collect image information and spectral information of the rock mass at the tunnel face, and transmit the collected information to the data processing and display module;
[0101] The auxiliary module is used to assist in collecting tunnel face images and spectral information, and provides functions such as dedicated light source illumination and position adjustment, including a full-spectrum halogen lamp 2, a lifting platform and a slide rail 8, to ensure all-round detection of the rock mass on the tunnel face;
[0102] The data processing module is used to transmit the information transmitted by the data acquisition module through four parts: data preprocessing, grid division, image and spectrum information analysis, and result output, and transmit the analysis results of the image and spectrum information to the result display module;
[0103] The data storage module is used to store the tunnel face image and spectrum information obtained by the data acquisition module and the rock mass quality evaluation information analyzed by the data processing module;
[0104] The result display module is used to integrate and display the processing results;
[0105] The main control module 1 is connected to the data acquisition module, auxiliary module, data processing module, data storage module and result display module, and is used to receive information from each module, control the movement of the data acquisition module, auxiliary module and carrying platform 7, and cooperate with the collection of image and spectral information.
[0106] The specific installation methods of the rock quality evaluation system based on hyperspectral imaging technology can be drone-mounted or vehicle-mounted.
[0107] This embodiment takes the vehicle-mounted type as an example, and the system mainly includes a data acquisition module, an auxiliary module, a data processing module, a data storage module, a result display module, a main control module 1, a carrying platform 7, a power supply 9 and a track wheel 100. The data acquisition module includes an imaging spectrometer 3, a micro camera 4 and a lens 5, and a first lifting platform 601; the auxiliary module includes a full-spectrum halogen lamp 2, a second lifting platform 602 and a slide rail 8 that can adjust the position of the data acquisition and storage module.
[0108] First, the trolley is controlled to move to a suitable position at a distance from the tunnel face through the crawler wheels 100. The crawler wheels 100 have a low center of gravity, a large adhesion coefficient and traction, and a good ability to resist tipping and sliding, and are suitable for movement in complex tunnel terrain.
[0109] Then adjust the vehicle body, align one side of the data acquisition module with the tunnel face, turn on the power supply 9, the full-spectrum halogen lamp 2, the imaging spectrometer 3 and the micro camera 4. Turn on the switch of the second lifting platform 602, adjust the position of the full-spectrum halogen lamp 2 up and down, so that the illumination conditions of the tunnel face meet the requirements of hyperspectral imaging, and then adjust the positions of the imaging spectrometer 3 and the micro camera 4 through the first lifting platform 601 and the slide rail 8 to ensure that the tunnel face can be completely scanned and photographed.
[0110] During the scanning and shooting process, the data acquisition module transmits the collected image and spectral data to the data processing module for real-time processing, and then stores the input image and spectral information and the processed results in the data storage module and displays them on the screen of the main control module 1 through the result display module.
[0111] After scanning and shooting, turn off the full-spectrum halogen lamp 2, the imaging spectrometer 3, and the micro camera 4, retract the lifting platform, turn off the power 9, and leave the working area of the equipment.
[0112] The implementation process of each module is as follows:
[0113] When the trolley is moved to the vicinity of the platform under the control of the main control module 1, the auxiliary devices, mainly including the full-spectrum halogen lamp 2, the lifting platform and the slide rail 8, are adjusted to ensure sufficient illumination of the tunnel face and the integrity of the shooting of the spectrometer and the camera 4. Then, the tunnel face is scanned and photographed through the data acquisition module to obtain image information and spectral information.
[0114] The data storage module mainly stores the original image and spectral information obtained by the data acquisition module in the above process and the processed output information obtained by the data processing module in a one-to-one correspondence for subsequent research and verification.
[0115] The main control module 1 is connected to the above modules, controls the movement of the data acquisition module, the auxiliary module and the carrying platform 7, cooperates with the collection of image and spectral information, and integrates these operations into one interface.
[0116] The steps involved in the apparatus of the above embodiments 2 and 3 correspond to the method embodiment 1. For the specific implementation, please refer to the relevant description of embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood to include any medium that can store, encode or carry an instruction set for execution by a processor and enable the processor to execute any method of the present invention.
[0117] Those skilled in the art should understand that the modules or steps of the present invention described above can be implemented by a general-purpose computer device, or alternatively, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.
[0118] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.
Claims
1. Rock mass quality evaluation method based on hyperspectral imaging technology, It is characterized in that The following steps are involved: Collect the original image information and spectral information of the target rock tunnel face, and perform pre-processing and gridding; The gridded image information and spectral information are respectively input into the image and rock mass structure feature deep learning model and the spectrum and rock mass strength feature deep learning model to extract the rock mass structure feature and rock mass strength feature; The rock mass structure characteristics and rock mass strength characteristics are input into the fused image spectrum-rock mass quality evaluation model to perform feature dimension reduction and multi-feature vector fusion. The fused features are compressed using an activation function and the rock mass quality grade is classified using a Logistic regression classifier to obtain the rock mass quality evaluation grade within a single grid. The rock mass quality evaluation and grading conditions of each grid are integrated and the rock mass quality evaluation results are output.
2. The rock mass quality evaluation method based on hyperspectral imaging technology according to claim 1, It is characterized in that The deep learning model of image and rock structure characteristics and the deep learning model of spectrum and rock strength characteristics are both convolutional neural network models. The convolution layer initially extracts features, the pooling layer extracts main features, and multiple rounds of convolution and pooling are performed. Finally, the fully connected layer summarizes the features of each part to achieve the extraction of crack and rock strength related features.
3. The rock mass quality evaluation method based on hyperspectral imaging technology according to claim 1, It is characterized in that Extract rock mass structural features, including: The preprocessed image information is segmented based on the SLIC superpixel segmentation algorithm to obtain superpixel blocks of the image; Calculate the merging weights between adjacent superpixel blocks and construct a regional adjacency graph; Execute the hierarchical region merging algorithm to merge the regional adjacency graphs, and transform the contour of the merged result into a joint and fissure distribution map through coordinate transformation; The crack distribution map is input into the trained image and rock mass structural characteristics deep learning model to obtain the structural characteristics of the rock mass.
4. The rock mass quality evaluation method based on hyperspectral imaging technology according to claim 1, It is characterized in that Extract rock mass strength characteristics, including: Based on the different mineral compositions of different rock masses and the changes in the spectral radiation brightness and band characteristics of the same rock mass under different stresses, a database on the relationship between spectrum and rock strength is constructed; After performing first-order differential and continuum removal transformation on the divided spectral information, characteristic wavelengths are selected; Based on the characteristic wavelength and the constructed database of the relationship between spectrum and rock strength, the spectrum information matching method is adopted to obtain the relevant information of rock mass strength. The spectrum and rock strength feature deep learning model is used to transform the relevant information of rock strength into the strength characteristics of the rock mass.
5. The rock mass quality evaluation method based on hyperspectral imaging technology according to claim 1, It is characterized in that Image preprocessing includes grayscale, enhancing image brightness and contrast, image sharpening, and median filtering; Spectral preprocessing includes spectrum smoothing, derivative, normalization, wavelet transform, and convolution smoothing based on the least squares principle.
6. The rock mass quality evaluation method based on hyperspectral imaging technology according to claim 1, It is characterized in that Grid processing specifically includes: Set the mesh density according to the actual project needs; Identify the grid boundaries where the number of rock mass cracks and joints and spectral information change significantly; The mesh is divided based on the mesh density and the identified mesh boundaries.
7. The rock mass quality evaluation method based on hyperspectral imaging technology according to claim 1, It is characterized in that The target rock mass is scanned and tested non-contactly to collect the original image information and spectral information of the target rock mass tunnel face.
8. Rock mass quality evaluation system based on hyperspectral imaging technology, Features: include: The information acquisition module is configured to: collect original image information and spectral information of the target rock tunnel face, and perform pre-processing and gridding processing; The feature extraction module is configured to: input the gridded image information and spectral information into the image and rock mass structure feature deep learning model and the spectrum and rock strength feature deep learning model respectively, and extract the rock mass structure feature and rock mass strength feature; The fusion analysis module is configured to: input rock mass structure features and rock mass strength features into the fusion constructed image spectrum-rock mass quality evaluation model, perform feature dimension reduction and multi-feature vector fusion, compress the fused features using an activation function, and use a Logistic regression classifier to classify the rock mass quality level, and obtain the rock mass quality evaluation classification within a single grid; The integration module is configured to integrate the rock mass quality evaluation and classification conditions of each grid and output the rock mass quality evaluation results.
9. A rock mass quality evaluation device based on the rock mass quality evaluation system based on hyperspectral imaging technology as claimed in claim 8, It is characterized in that It includes a carrying platform, a data acquisition device, an auxiliary device and a main control module. The data acquisition device includes an imaging spectrometer, a camera, a first lifting platform and a slide rail. The auxiliary device includes a full-spectrum halogen lamp and a second lifting platform: The slide rail is arranged on the top of the carrying platform, the first lifting platform is slidably connected to the slide rail, the imaging spectrometer and the camera are arranged on the top of the first lifting platform; the second lifting platform is arranged on the top of the carrying platform, and the full-spectrum halogen lamp is arranged on the top of the second lifting platform; the bottom of the carrying platform is also provided with track wheels and a track wheel driving mechanism; The main control module is electrically connected to the first lifting platform, the second lifting platform, the imaging spectrometer, the camera, the full-spectrum halogen lamp and the track wheel driving mechanism respectively.
10. The rock mass quality evaluation device based on hyperspectral imaging technology according to claim 9, It is characterized in that It also includes a data processing module, a data storage module, and a result display module, wherein: The data processing module is electrically connected to the camera, the imaging spectrometer, and the result display module, respectively, receives the original image information and spectral information of the target rock mass tunnel face sent back by the camera and the imaging spectrometer, processes the information, and obtains the rock mass quality evaluation result; The result display module is electrically connected to the data processing module to display the rock mass quality evaluation results; The data storage module is electrically connected to the data processing module, the camera and the imaging spectrometer respectively, and stores the rock mass quality evaluation result, the original image information and the spectral information of the target rock mass tunnel face.
