CN115713645A - Lithology identification method and system based on spectral imaging technology - Google Patents

Abstract

The invention belongs to the technical field of lithology identification of engineering geology, and provides a lithology identification method and a lithology identification system based on a spectral imaging technology. Acquiring image spectrum data of a tunnel face of a tunnel, and rasterizing an interested area; carrying out regional trace detection by using image data of a tunnel face, and judging whether a trace exists in a grid; performing second rasterization processing on the grids with the traces; extracting all traceless grids and spectral features and image features of the grids obtained by rasterization again; the spectral features and the image features of the same grid are fused, and then the lithology of each grid is identified through a pre-trained lithology prediction model; and checking and correcting the lithology recognition result of the grid obtained by the second rasterization, and finally obtaining the lithology of all the grids.

Description

Lithology identification method and system based on spectral imaging technology

technical field

The invention belongs to the technical field of engineering geological lithology identification, and in particular relates to a lithology identification method and system based on spectral imaging technology.

Background technique

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

Lithology identification has always been a very important and basic issue in the fields of geology, resource exploration, poor geological identification of tunnels and underground engineering, and disaster prevention and mitigation. The lithological identification of unfavorable geological areas is the premise and basis for tunnel engineering geological prediction, and has important guiding significance for tunnel engineering design optimization, safety assessment and risk assessment. The traditional lithology identification methods, such as naked eye observation and thin section identification, rely too much on manual experience, which is not only time-consuming and highly professional, but also easily affected by subjective factors, resulting in unsatisfactory accuracy. Geologists have carried out research on lithology identification based on images, but the result is that due to the high similarity of images caused by similar rock composition or texture, the fine features of lithology are easily lost during the feature extraction process, and weathering or human activities destroy the visible features. The identification accuracy is inaccurate due to problems such as rock features, poor imaging quality due to shooting conditions or technical differences. There are many ways to obtain rock mineral types and content information, among which spectroscopic technology is currently the most commonly used method, among which XRF and XRD spectral tests require contact measurement, sample grinding processing, etc., which will consume a lot of time, and only rely on a single feature of the rock , it is easy to produce misclassification and omission phenomenon, leading to problems such as large lithology identification error and low lithology interpretation accuracy.

The inventors found that rocks can be simply classified with the help of image information or spectral information of rock ore, but there are phenomena of "same object with different spectrum" and "same spectrum with different object". Such methods often have low classification accuracy. Nowadays, using spectral information Most of the lithology identification relies on contact or sample grinding, which has certain limitations when used in tunnel sites.

Contents of the invention

In order to solve the technical problems in the above-mentioned background technology, the present invention provides a lithology identification method and system based on spectral imaging technology, which can identify the lithology and spatial distribution of the tunnel face in real time, and make a prediction of the front of the tunnel. The judgment of the geological condition of the rock mass provides an important reference for mastering the geological condition of the rock mass in front of the tunnel. The present invention utilizes image spectrum technology and image spectrum information obtained by long-distance photography to obtain not only image data, but also spectral data of each pixel on the image, that is, a three-dimensional cubic data volume.

In order to achieve the above object, the present invention adopts the following technical solutions:

A first aspect of the present invention provides a method for lithology identification based on spectral imaging technology, which includes:

Obtain the image spectrum data of the tunnel face and perform rasterization on the region of interest;

Use the image data of the tunnel face to detect area traces, and judge whether there are traces in the grid;

Perform a second rasterization process on the grid with traces;

Extract the spectral features and image features of all grids without traces and the grids obtained by rasterization again;

The spectral features and image features of the same grid are fused, and then the lithology of each grid is identified through the pre-trained lithology prediction model;

The lithology identification results of the grids obtained by the second rasterization are checked and corrected, and finally the lithology of all grids is obtained.

As an implementation manner, the lithology of all the finally obtained grids is displayed on the image of the tunnel face by filling in a map.

The advantage of the above technical solution is that the lithology and spatial distribution on the face of the tunnel can be recognized intuitively in real time, and the geological conditions of the rock mass in front of the tunnel can be judged, which provides important information for mastering the geological conditions of the rock mass in front of the tunnel. reference basis.

As an implementation manner, the lithology of the grid obtained by the second rasterization is constrained by using the nine-square grid test method and the regional trace, and the lithology identification result of the corresponding grid is corrected.

Among them, the Jiugongge test method is to use the Jiugongge diagram to test the grids of the abnormal recognition results of the second divided small grids; firstly, the grids with different recognition results in the nine grids are calibrated, and the Jiugongge is established with the abnormal grid as the center. , classify it as the lithology with the largest proportion among the eight adjacent grids.

The advantage of the above technical solution is to correct the error of the lithology identification result of the small grid crossed by the regional trace, and improve the accuracy of the lithology identification result of the second divided small grid.

As an implementation, before fusing the spectral features and image features of the same grid, it also includes:

Normalize the spectral and image features of the same grid.

The advantage of the above technical solution is to ensure that the dimension of the feature vector of the image dimension is consistent with the dimension of the feature vector of the spectral dimension, and finally improve the accuracy of the lithology identification result of the grid.

As an implementation manner, after acquiring the image spectrum data of the tunnel face, it also includes:

Preprocessing of the spectral data of the tunnel face.

Among them, there are many preprocessing methods, such as convolution smoothing, area normalization, baseline Correction, first derivative, standard normal variable transformation, multivariate scatter correction, etc. The most commonly used noise removal method (SG) convolution smoothing method performs spectral smoothing on the collected hyperspectral curve, which not only eliminates noise but also preserves the spectral profile.

As an implementation manner, the lithology prediction model is a classifier.

What needs to be explained here is that the classifier can use models such as extreme learning machine, partial least squares regression, and support vector machine.

A second aspect of the present invention provides a lithology identification system based on spectral imaging technology, which includes:

The initial rasterization module, which is used to obtain the image spectrum data of the tunnel face, and perform rasterization processing on the region of interest;

An area trace detection module, which is used to perform area trace detection using the image data of the tunnel face to determine whether there are traces in the grid;

A rasterization module again, which is used to perform a second rasterization process on the grid with traces;

A feature extraction module, which is used to extract the spectral features and image features of all grids without traces and the grids obtained by rasterization again;

The lithology identification module is used to fuse the spectral features and image features of the same grid, and then identify the lithology of each grid through the pre-trained lithology prediction model;

The lithology correction module is used to check and correct the lithology identification results of the grids obtained by the second rasterization, and finally obtain the lithology of all the grids.

As an implementation, the lithology identification system based on spectral imaging technology also includes:

The identification result display module is used to display the lithology of all grids finally obtained on the image of the tunnel face in the form of mapping.

As an implementation manner, in the lithology correction module, the lithology of the grid obtained from the second rasterization is constrained by using the Jiugongge test method and the regional trace, and the lithology identification result of the corresponding grid is corrected.

As an implementation manner, in the lithology identification module, before the spectral features and image features of the same grid are fused, it also includes:

Normalize the spectral and image features of the same grid.

As an implementation manner, in the initial rasterization module, after acquiring the image spectrum data of the face of the tunnel, it also includes:

Preprocessing of the spectral data of the tunnel face.

As an implementation manner, in the lithology identification module, the lithology prediction model is a classifier.

A third aspect of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, the steps in the method for identifying lithology based on spectral imaging technology as described above are implemented.

A fourth aspect of the present invention provides a computer device, including a memory, a processor, and a computer program stored on the memory and operable on the processor. When the processor executes the program, the above-mentioned based Steps in the lithology identification method of spectral imaging technique.

Compared with prior art, the beneficial effect of the present invention is:

(1) Due to the large data volume of the three-dimensional data obtained by the image spectrum technology, the large amount of full-band spectral data of the sample, and the mixed information, the present invention uses the acquired image spectral data of the tunnel face to perform rasterization processing on the region of interest , calculate the average value of the full band of the pixels of the spectral image in each grid, and fuse the image features in the grid with the average spectrum to identify the lithology, which can not only ensure the full acquisition of face information, but also reduce the Reduce the amount of model computation, improve model robustness and work efficiency.

(2) The present invention fuses the image data and spectral data of the face, and extracts all grids without traces and the spectral features and image features of the grids obtained by gridding again through rasterization processing, and undergoes feature fusion, Identify the lithology of each grid; and then check and correct the lithology identification results of the grids obtained by the second rasterization, and finally obtain the lithology of all grids, which realizes the tunnel face alignment at the tunnel site To identify the lithology, it is only necessary to collect the information of the tunnel face through hyperspectral imaging technology. After data processing and analysis, the lithology and spatial distribution of the tunnel face can be identified in real time, and the geological conditions of the rock mass in front of the tunnel can be made. The judgment provided an important reference for mastering the geological conditions of the rock mass in front of the tunnel.

Advantages of additional aspects of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention.

Description of drawings

The accompanying drawings constituting a part of the present invention are used to provide a further understanding of the present invention, and the schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations to the present invention.

Fig. 1 is the flow chart of the lithology identification method based on spectral imaging technology of the embodiment of the present invention;

Fig. 2 is a schematic diagram of the face mesh division of the embodiment of the present invention;

Fig. 3 is a schematic structural diagram of a lithology identification system based on spectral imaging technology according to an embodiment of the present invention.

Among them, 1. Standard whiteboard; 2. Spectral imager; 3. Imaging lens; 4. Battery; 5. Retractable platform; 6. Telescopic rod; 7. Data analysis platform; 8. Light source;

Detailed ways

The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

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

It should be noted that the terminology used here is only for describing specific embodiments, and is not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly dictates otherwise, the singular is intended to include the plural, and it should also be understood that when the terms "comprising" and/or "comprising" are used in this specification, they mean There are features, steps, operations, means, components and/or combinations thereof.

Embodiment one

With reference to Fig. 1, the present embodiment provides a kind of lithology identification method based on spectral imaging technology, and it specifically comprises the following steps:

Step 1: Obtain the image spectral data of the tunnel face and perform rasterization on the region of interest.

In the specific implementation process, the image spectrum data of the tunnel face can be realized by using the data acquisition and storage system, and its structure is shown in Figure 3. Among them, the data acquisition and storage system includes a hyperspectral imaging camera, a pan/tilt 9 , a standard whiteboard 1 , a battery 4 and a light source 8 . Wherein, the hyperspectral imaging camera includes a spectral imager 2 and an imaging lens 3 . The pan-tilt 9 is fixed on the robot for installing and fixing the hyperspectral imaging camera, and the pan-tilt 9 can be rotated arbitrarily, which is convenient for actual on-site operation. The bottom of the cloud platform 9 is also provided with a telescopic platform 5. The light source 8 is fixed on the robot and is located around the hyperspectral imaging camera. Its type is a halogen lamp, and the height and angle of the light source 8 are adjustable. The standard white board 1 is mainly used for optical calibration measurement of spectral analysis.

In this embodiment, the imaging hyperspectral instrument has the advantage of "integration of maps and spectra". The scanning of the face rock can not only obtain the rock image synchronously, but also obtain the mineral spectral information on the entire rock surface. Through the analysis and fusion of hyperspectral three-dimensional data , making up for the lack of single data, and can quickly and efficiently obtain the spatial distribution information of face lithology.

In order to reduce the influence of uneven illumination and camera dark current, it is necessary to perform black and white correction on the collected original hyperspectral data. ref, to obtain rectified hyperspectral images.

Adjust the distance between the robot and the face of the tunnel, the angle of the gimbal, and the height of the imaging spectrometer according to different tunnel environments. It mainly depends on what spatial resolution scale the details of the target need to be clearly seen and distinguished without being disturbed by others. .

Adjust and calibrate the standard whiteboard 1 of the imaging spectrometer through the telescopic rod 5 connected to the standard whiteboard 1, move the position of the standard whiteboard 1 to the front of the camera, and align the camera probe vertically with the standard whiteboard for calibration.

Among them, the data collection and storage system is connected to the data analysis platform 7, and the image spectrum acquired by the data collection and storage system is transmitted to the data analysis platform 7 for corresponding rasterization and other processing.

Because the image data collected by the image spectrum system has many pixels, each pixel can extract a complete high-resolution spectral curve. The large amount of data and the increase of wave bands will lead to information redundancy and increase the complexity of data processing. If all the spectral data of the entire face are averaged, the overall recognition accuracy will be reduced. In this embodiment, the face is divided into grids, that is, the spectral values of N pixels in the small grid under the same band are taken as Mean value, and finally each grid can extract a complete spectral curve.

The grid division is based on the actual tunnel engineering environment (depending on the shooting range of the tunnel face and the size of the tunnel face). When other settings are correct, the denser the grid, the higher the solution accuracy; but the denser the grid , the larger the number, the longer the solution time; therefore, a trade-off needs to be made between solution efficiency and solution accuracy. For example: when first debugging the grid division, the area of a single grid can be set to about 0.1 square meters. If the area of a tunnel face is 40 square meters, divide the tunnel face into 400 grids equally.

As a specific implementation, after acquiring the image spectrum data of the tunnel face, it also includes:

Preprocessing of the spectral data of the tunnel face.

Among them, there are many preprocessing methods, such as convolution smoothing, area normalization, baseline Correction, first derivative, standard normal variable transformation, multivariate scatter correction, etc. The most commonly used noise removal method (SG) convolution smoothing method performs spectral smoothing on the collected hyperspectral curve, which not only eliminates noise but also preserves the spectral profile.

The effects of preprocessing the spectral data of the tunnel face include functions such as black and white correction and spectral smoothing. In order to overcome the influence of the inhomogeneity of light source intensity in each band and the influence of dark current in the process of hyperspectral image acquisition, the acquired hyperspectral curve is spectrally smoothed, which not only eliminates noise but also preserves the spectral profile.

Step 2: Use the image data of the tunnel face to detect area traces, and judge whether there are traces in the grid.

In the specific implementation process, the image data of the tunnel face is used to detect the regional traces, and the detection results include two cases: no traces in the grid and traces in the grid, as shown in Figure 2.

When the detection result is that there is no trace in the grid, the grid does not need to be rasterized again.

Step 3: Perform a second rasterization process on the grid with traces.

Step 4: Extract the spectral features and image features of all grids without traces and the grids obtained by rasterization again.

Specifically, the spectral features include spectral value averaging, feature band extraction, and the number of spectral bands. Spectral averaging, mean the spectral values of all pixels in the same band in the small grid, and finally extract a spectral curve from one grid. Extraction of characteristic bands, screening and extraction of characteristic wavelengths.

Specifically, spectral averaging means the average value of the spectral values of all pixels in the same band in the small grid, that is, each grid hyperspectral image is a three-dimensional tensor of N×N×P, where N×N is the spatial dimension, and P is Spectral dimension, expand this three-dimensional tensor along the third dimension to obtain (N*N)×P, which means that each band corresponds to N*N pixel points, and then average these pixel points to obtain a 1×P vector, which can also be understood as m is the inner image of the grid

—The number of prime points, Iij is the spectral value of the i-th pixel in the j-th band, and I _j is the j-th wave in a grid

- the average spectral value under the segment, these I _j (j=1, 2...) form a complete spectral curve, and finally a grid extracts a spectral curve.

For the extraction of characteristic bands, due to the large number of spectral data variables, there may be redundant information. If each spectral value is substituted into the model analysis, it will not only affect the accuracy of recognition and prediction, but also increase the amount of calculations for system processing and analysis, and reduce the model. Therefore, it is necessary to screen and extract the characteristic wavelengths. In the practical application analysis of hyperspectral images, spectroscopy requires that there be obvious peaks or troughs in the spectral curve of the selected characteristic band, that is, the target to be measured will absorb or reflect the light of the characteristic band, and perform principal component analysis on the original data to remove the band The redundant information between the multi-band image information is compressed into a few conversion bands that are more effective than the original bands. According to the factor load between the principal component and the original feature, the feature with the greatest correlation with the principal component in the original feature can be obtained, so as to realize the feature wavelength extraction. Finally, the characteristic bands b1, b2, b3... are selected to form the characteristic space, and the average spectral curve of the sample under the characteristic space is obtained.

Among them, the image features are very rich, and there is a close relationship between each feature, and the information fusion between various features is an important factor for recognition. The texture features can be calculated from the image at four angles of 0°, 45°, 90°, and 135° by using gray-level co-occurrence matrix and other methods, and the energy, entropy, moment of inertia, and correlation of the image spectrum image can be extracted as texture feature.

Step 5: The spectral features and image features of the same grid are fused, and then the lithology of each grid is identified through the pre-trained lithology prediction model.

Specifically, before fusing the spectral features and image features of the same grid, it also includes:

Normalize the spectral and image features of the same grid.

The advantage of the above technical solution is to ensure that the dimension of the feature vector of the image dimension is consistent with the dimension of the feature vector of the spectral dimension, and finally improve the accuracy of the lithology identification result of the grid.

The normalized spectral features and image features are fused, and the fused features are sent to the pre-trained lithology prediction model for lithology classification.

Wherein, the lithology prediction model is a classifier.

What needs to be explained here is that the classifier can use models such as extreme learning machine, partial least squares regression, and support vector machine.

Step 6: Check the lithology identification results of the grids obtained by the second rasterization and make corrections, and finally obtain the lithology of all grids.

Through the model established in the early stage, the model will describe the characteristics of the target data through certain rules, and then qualitatively classify or quantitatively predict the data according to the characteristics, and identify the lithology of each small grid. At the same time, the nine-square grid test method is adopted, and the regional trace detection results processed by image information are jointly determined, which can reduce the impact of local area recognition on the overall recognition result.

In some embodiments, the lithology of the grid obtained from the second rasterization is constrained by using the nine-square grid test method and the regional trace, and the lithology identification result of the corresponding grid is corrected.

Among them, the Jiugongge test method is to use the Jiugongge diagram to test the grids of the abnormal recognition results of the second divided small grids; firstly, the grids with different recognition results in the nine grids are calibrated, and the Jiugongge is established with the abnormal grid as the center. , classify it as the lithology with the largest proportion among the eight adjacent grids.

Generally, there is an error in the lithology identification result of the small grid crossed by the regional trace, because there may be two kinds of lithology in this grid. At this time, continue to divide the grid again for the grid crossed by the trace. This time The divided area is about 0.01m ² (that is, divided into 9 grids). Using the same method as above, an extracted spectral curve is fused with the image features in the small grid to identify the lithology. For the second division of small grids, the nine-square grid test method is used to correct the recognition results.

The nine-grid test method is to utilize the nine-grid diagram to test the grid of the abnormal recognition result of the small grid divided for the second time; first, the grids with different recognition results in the nine grids are marked, and the nine-grid is established with the abnormal grid as the center. , classify it as the lithology with the largest proportion among the eight adjacent grids.

The advantage of the above technical solution is to correct the error of the lithology identification result of the small grid crossed by the regional trace, and improve the accuracy of the lithology identification result of the second divided small grid.

In one or more embodiments, the finally obtained lithology of all the grids can be displayed on the image of the tunnel face by filling in a map.

The advantage of the above-mentioned technical solution is that the lithology and spatial distribution on the face of the tunnel can be recognized intuitively in real time, and the geological conditions of the rock mass in front of the tunnel can be judged, which provides important information for mastering the geological conditions of the rock mass in front of the tunnel. reference basis.

Embodiment two

This embodiment provides a lithology identification system based on spectral imaging technology, which includes an initial rasterization module, an area trace detection module, a second rasterization module, a feature extraction module, a lithology identification module and a lithology correction module .

(1) The initial rasterization module, which is used to acquire the image spectrum data of the tunnel face, and perform rasterization processing on the region of interest.

Specifically, in the initial rasterization module, after acquiring the image spectrum data of the tunnel face, it also includes:

Preprocessing of the spectral data of the tunnel face.

(2) A regional trace detection module, which is used to detect regional traces by using the image data of the tunnel face, and judge whether there are traces in the grid.

(3) A re-rasterization module, which is used to perform a second rasterization process on the grid with traces.

(4) A feature extraction module, which is used to extract spectral features and image features of all traceless grids and grids obtained by rasterization again.

(5) The lithology identification module, which is used to fuse the spectral features and image features of the same grid, and then identify the lithology of each grid through the pre-trained lithology prediction model.

Specifically, in the lithology identification module, before the spectral features and image features of the same grid are fused, it also includes:

Normalize the spectral and image features of the same grid.

Specifically, in the lithology identification module, the lithology prediction model is a classifier.

For example: the classifier can use models such as extreme learning machine, partial least squares regression, support vector machine and so on.

(6) A lithology correction module, which is used to check and correct the lithology identification results of the grids obtained by the second rasterization, and finally obtain the lithology of all the grids.

Specifically, in the lithology correction module, the lithology of the grid obtained from the second rasterization is constrained by using the Jiugongge test method and the regional trace, and the lithology identification result of the corresponding grid is corrected.

In one or more embodiments, the lithology identification system based on spectral imaging technology also includes:

The identification result display module is used to display the lithology of all grids finally obtained on the image of the tunnel face in the form of mapping.

It should be noted here that each module in this embodiment corresponds to each step in Embodiment 1, and the specific implementation process is the same, so it will not be repeated here.

Embodiment three

This embodiment provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, the steps in the method for identifying lithology based on spectral imaging technology as described above are implemented.

Embodiment four

This embodiment provides a computer device, including a memory, a processor, and a computer program stored on the memory and operable on the processor. When the processor executes the program, the above-mentioned spectrum-based imaging technology is implemented. steps in the lithology identification method.

The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the invention. It should be understood that each procedure and/or block in the flowchart and/or block diagram and a combination of procedures and/or blocks in the flowchart and/or block diagram can be realized by computer program instructions. These computer program instructions may be provided to a general purpose computer, special purpose computer, embedded processor, or processor of other programmable data processing equipment to produce a machine such that the instructions executed by the processor of the computer or other programmable data processing equipment produce a An apparatus for realizing the functions specified in one or more procedures of the flowchart and/or one or more blocks of the block diagram.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and changes. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims (10)

1. A lithology identification method based on spectral imaging technology, characterized in that, comprising: Obtain the image spectrum data of the tunnel face and perform rasterization on the region of interest; Use the image data of the tunnel face to detect area traces, and judge whether there are traces in the grid; Perform a second rasterization process on the grid with traces; Extract the spectral features and image features of all grids without traces and the grids obtained by rasterization again; The spectral features and image features of the same grid are fused, and then the lithology of each grid is identified through the pre-trained lithology prediction model; The lithology identification results of the grids obtained by the second rasterization are checked and corrected, and finally the lithology of all grids is obtained. 2. The lithology identification method based on spectral imaging technology as claimed in claim 1, characterized in that, the lithology of all the grids finally obtained is displayed on the image of the tunnel face by filling in a map. 3. the lithology identification method based on spectral imaging technology as claimed in claim 1, is characterized in that, adopts Jiugongge test method and area trace to constrain the lithology of the grid that the second rasterization obtains, corrects corresponding The lithology identification results of the grid. 4. The lithology identification method based on spectral imaging technology as claimed in claim 1, is characterized in that, before the spectral features and image features of the same grid are fused, it also includes: Normalize the spectral and image features of the same grid. 5. the lithology identification method based on spectral imaging technology as claimed in claim 1, is characterized in that, after obtaining the image spectral data of tunnel face, also comprises: Preprocessing of the spectral data of the tunnel face. 6. The lithology identification method based on spectral imaging technology as claimed in claim 1, wherein the lithology prediction model is a classifier. 7. A lithology identification system based on spectral imaging technology, characterized in that it comprises: The initial rasterization module, which is used to obtain the image spectrum data of the tunnel face, and perform rasterization processing on the region of interest; An area trace detection module, which is used to perform area trace detection using the image data of the tunnel face to determine whether there are traces in the grid; A rasterization module again, which is used to perform a second rasterization process on the grid with traces; A feature extraction module, which is used to extract the spectral features and image features of all grids without traces and the grids obtained by rasterization again; The lithology identification module is used to fuse the spectral features and image features of the same grid, and then identify the lithology of each grid through the pre-trained lithology prediction model; The lithology correction module is used to check and correct the lithology identification results of the grids obtained by the second rasterization, and finally obtain the lithology of all the grids. 8. the lithology identification system based on spectral imaging technology as claimed in claim 7, is characterized in that, the lithology identification system based on spectral imaging technology also comprises: The identification result display module is used to display the lithology of all grids finally obtained on the image of the tunnel face in the form of mapping; or In the lithology correction module, the lithology of the grid obtained by the second rasterization is constrained by using the Jiugongge test method and the regional trace, and the lithology identification result of the corresponding grid is corrected; or In the lithology identification module, before the spectral features and image features of the same grid are fused, it also includes: Normalize the spectral features and image features of the same grid; or In the initial rasterization module, after acquiring the image spectrum data of the face of the tunnel, it also includes: Preprocessing the spectral data of the tunnel face; or In the lithology identification module, the lithology prediction model is a classifier. 9. A computer-readable storage medium, on which a computer program is stored, characterized in that, when the program is executed by a processor, the lithology identification based on spectral imaging technology according to any one of claims 1-6 is realized steps in the method. 10. A computer device, comprising a memory, a processor, and a computer program stored on the memory and operable on the processor, characterized in that, when the processor executes the program, it realizes any of claims 1-6. A step in the lithology identification method based on spectral imaging technology.

Images