Rock-soil core sample information recording system based on spectrum and image recognition

Abstract

The invention discloses a rock-soil core sample information logging system based on spectrum and image recognition, and the system comprises a spectrum imaging module which carries out the collection of spectrum data, and carries out the classification and recognition of a rock-soil layer; the image recognition module is used for acquiring an image and recognizing rock-soil layer characteristics in the image; the data processing module is used for receiving and processing the spectral data and the image data; the naming and determining module is used for determining the naming and description of the rock-soil core sample; the data logging module is used for integrating output results and generating a logging report and a database; and the database is used for storing and managing data and providing data and query functions for each module. According to the invention, the image features and spectral features of the rock-soil core sample can be rapidly acquired through the spectrum and image acquisition device, the image feature information of each layer of rock-soil can be automatically identified by using the image identification module, the name of each layer of rock-soil can be determined according to the image feature information of each layer of rock-soil, and accurate naming is carried out by combining the spectral feature information. And the efficiency and correctness are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of rock and soil core sample identification, and more specifically, particularly relates to a rock and soil core sample information cataloging system based on spectrum and image recognition. Background Art

[0002] The detailed information cataloging of geotechnical core samples constitutes the cornerstone of geotechnical engineering investigation work and is crucial for the investigation of construction projects. The accuracy and efficiency of the identification of core samples obtained by drilling are directly related to the overall progress of the project. In the traditional geotechnical investigation process, the cataloging of core sample properties involves many tedious steps, significant repetitive labor, and high reliance on manual operations, which not only increases the labor intensity, but also complicates key links such as stratum naming and property determination. Usually, these judgments need to be combined with geotechnical tests, on-site in-situ test results, and follow established processes for comprehensive analysis before they can be completed. This process is time-consuming and error-prone.

[0003] This traditional method is significantly limited by the tediousness of manual data entry, the lack of intelligence, the diversity of data format standards, and the difficulty of integrating multi-source data. In addition, the establishment of the evaluation system often relies too much on the professional skills and personal experience of the catalogers, which makes it difficult to ensure the accuracy of trait information identification and improve the efficiency of information cataloging. More importantly, it fails to make full use of cutting-edge scientific and technological forces such as artificial intelligence, visual recognition, and big data analysis. The huge potential of these technologies in promoting the transformation and upgrading of the geotechnical exploration industry has been seriously neglected. Therefore, the traditional geotechnical exploration model has obvious shortcomings in ensuring information accuracy and improving work efficiency, and it is urgent to introduce new technologies to overcome its limitations.

[0004] However, there are some problems with the existing technology: the traditional rock and soil core sample information cataloging method is highly dependent on the professional ability and personal experience of the cataloger, and there are problems of strong subjectivity and inconsistent judgment standards, which makes it difficult to ensure the accuracy of the naming and property identification of rock and soil core samples. In addition, the traditional cataloging method has a long process, many repetitive tasks, and high manual labor intensity, resulting in low efficiency in information cataloging and unable to meet the needs of rapid advancement of modern engineering. Finally, when traditional methods process multi-source data, such as spectral data, image data, etc., there are problems with inconsistent data format standards and difficulty in effective integration, which limits the comprehensive utilization of data. Therefore, we propose a rock and soil core sample information cataloging system based on spectral and image recognition. Summary of the invention

[0005] In view of the problems existing in the prior art, the purpose of the present invention is to provide a rock and soil core sample information cataloging system based on spectral and image recognition, which can effectively improve the accuracy and reliability of rock and soil core sample property information identification and improve the efficiency of information cataloging.

[0006] To achieve the above object, the present invention provides the following technical solution: a rock and soil core sample information cataloging system based on spectrum and image recognition, comprising:

[0007] Spectral imaging module: Use a spectrometer to collect spectral data of rock and soil core samples, compare the spectral curves recorded in the spectral database, and classify and identify rock and soil layers through matching algorithms;

[0008] Image recognition module: captures images of rock and soil core samples, compares them with the preset image sample library, and identifies rock and soil layer features in the image, including texture information, granularity information, color information, integrity information, and crack development degree information;

[0009] Data processing module: receives spectral data and image data, and performs tasks such as data preprocessing, feature extraction, and data fusion to provide data support for the naming and determination module;

[0010] Naming and determination module: Based on the output of the data processing module, the spectral characteristics and image characteristics are integrated to determine the naming and description of the rock and soil core sample;

[0011] Data cataloging module: Integrate the output results of all modules such as spectral imaging, image recognition, and naming determination to generate geotechnical core sample information cataloging reports and databases;

[0012] Database: Stores and manages spectral database, image sample library, catalog report and other data, and provides data and query functions for each module.

[0013] Optionally, the spectral imaging module includes mineral composition information, water content information, color center information and energy level information;

[0014] The mineral component information determination unit: identifies the mineral wavelength information of each layer of rock and soil through the spectrum, matches it with the spectrum database, sets the mineral wavelength identifier corresponding to the mineral component with the greatest similarity, and uses it as the mineral component identifier of the corresponding layer of rock and soil;

[0015] The moisture content information determination unit: by detecting the fluorescence emitted by each layer of rock and soil under high-energy X-ray irradiation, analyzing the element composition of the rock and soil material, and calculating the moisture content according to the relative content of the elements;

[0016] The color center information determination unit: by comparing the trench database, sets the color center mark corresponding to the rock and soil color center feature with the greatest similarity as the color center mark of the corresponding layer of rock and soil;

[0017] The classification and identification of rock and soil layers are specifically as follows:

[0018] 1) Use a spectrometer to collect spectral data of rock and soil core samples and measure the wavelength distribution of light reflected or emitted from the surface of the rock and soil core samples;

[0019] 2) matching the collected spectral data with known spectral curves in the spectral database;

[0020] 3) Preliminary identification of the characteristics of the rock and soil layers based on the matching results of the spectral data;

[0021] 4) Classify the rock and soil layers according to the identification results.

[0022] Optionally, the matching of the collected spectral data with known spectral curves in a spectral database is specifically:

[0023] Set the spectrum vector x = (x1, x2, ..., x n ) and y=(y1,y2,...,y n ), the formula of spectral correlation coefficient r is:

[0024]

[0025] Where: n is the number of wavelength points of the spectrum;

[0026] and are the means of x and y, namely:

[0027]

[0028] The value range of the correlation coefficient r is [-1,1];

[0029] When r = 1, it means that the contrast spectra are completely positively correlated;

[0030] When r = -1, it means that the contrast spectra are completely negatively correlated;

[0031] When r=0, it means that there is no linear correlation between the contrast spectra.

[0032] Optionally, the specific workflow of the data processing module is:

[0033] Preprocessing: preprocess the collected data separately, including denoising, enhancement, etc.

[0034] Spectral feature extraction: extract spectral features from spectral data, including peak wavelength, absorption or reflection intensity, spectral slope, etc.;

[0035] Image feature extraction: extracting features such as shape, texture, and color from image data;

[0036] Data fusion: Fusion of extracted spectral data with image data;

[0037] Data output: Output the processed data and use it as input for subsequent steps;

[0038] The processed data includes feature vectors and fused data;

[0039] The subsequent steps include classification, identification, naming, etc.

[0040] Optionally, the collected data are preprocessed separately, specifically:

[0041] The process of spectral data preprocessing is as follows:

[0042] 1) Baseline correction: remove baseline offset or drift caused by instrument, sample container or other factors;

[0043] 2) Wavelet transform denoising: Split the original signal into wavelet coefficients of different frequencies, set a threshold based on the data, remove the noise part below the threshold, perform inverse wavelet transform on the retained part, and resynthesize it into a denoised signal;

[0044] 3) Smoothing: Reduce the curve fluctuation of spectral data through filtering technology;

[0045] 4) Normalization: Normalize the spectral data and synchronize the intensity ratio of the spectrum;

[0046] 5) Spectral enhancement processing: Transformation based on the multispectral database band, including principal component transformation, independent component transformation, color space transformation and color stretching to enhance the spectrum;

[0047] The process of image data preprocessing is as follows:

[0048] 1) Filtering: Use the Gaussian template to scan each pixel in the image.

[0049] 2) Calculate the product of each pixel in the Gaussian template coverage area and the corresponding Gaussian kernel weight, add the obtained products, and output the weighted sum;

[0050] 3) Assign the calculated weighted average value to the central pixel point and output it as the filtered pixel value;

[0051] 4) Feature enhancement: including operations such as adjusting the grayscale distribution of the image, sharpening, spatial domain enhancement, and frequency domain enhancement.

[0052] Optionally, the extracting spectral features is specifically:

[0053] Peak identification: Analyze the spectral data, identify the light intensity peak at each wavelength point, record the corresponding wavelength of each peak, and record it as the peak wavelength;

[0054] Calculate intensity: At the peak wavelength or specific wavelength, measure the light intensity value at that wavelength, calculate the absorption or reflection intensity of the spectral data, and analyze the change trend of the absorption or reflection intensity at different wavelengths;

[0055] Slope extraction: Calculate the spectral slope within a specific interval of spectral data and analyze the changes in the spectral slope within different intervals;

[0056] Characteristic parameter definition: According to actual needs, define a series of characteristic parameters to describe the spectral characteristics, including peak wavelength, absorption or reflection intensity, spectral slope, etc. According to the defined characteristic parameters, calculate and extract the corresponding values;

[0057] Result analysis: Analyze the extracted spectral features to understand the optical characteristics, chemical properties and other information of the object or light source, and compare the extracted features with known standards.

[0058] Optionally, the formula of the slope extraction algorithm is:

[0059]

[0060] Where: x i is the wavelength of the spectral data;

[0061] y i is the spectral intensity of the corresponding wavelength;

[0062] w i is the weight of the i-th observation;

[0063] and The slope and intercept need to be estimated;

[0064] The slope and intercept estimates are calculated as follows:

[0065]

[0066] in: and is the weighted mean;

[0067]

[0068] The w i Determined based on the signal-to-noise ratio and data quality of the spectral data.

[0069] Optionally, the image feature extraction is specifically:

[0070] 1) Receive and process the image;

[0071] 2) Color feature extraction: Count the distribution of each color component in the image to form a color histogram, convert the color space through color index, and extract relevant features;

[0072] The color distribution is represented by a color moment;

[0073] 3) Texture feature extraction:

[0074] A. Analyze the gray level dependency between a pixel and its neighboring pixels to extract texture features;

[0075] B. Compare the grayscale values ​​of the central pixel with those of its neighboring pixels and encode the result into a binary number to extract features;

[0076] C. Use wavelet transform to decompose the image and extract texture features at different scales and directions;

[0077] 4) Shape feature extraction: extract edge information, shape, contour, and collection attributes from the image through relevant algorithms;

[0078] 5) Feature dimensionality reduction: Use principal component analysis, linear discriminant analysis and other methods to reduce feature dimensions and select relevant feature subsets.

[0079] Optionally, the data fusion is specifically:

[0080] Spatial registration: align the spectral data and image data in the same coordinate system and adjust them to the same spatial resolution and coverage;

[0081] Select fusion methods: including pixel-level fusion, feature-level fusion and decision-level fusion;

[0082] Perform data fusion: fuse the pre-processed spectral data and image data through data processing software or algorithm according to the selected fusion method;

[0083] Quality inspection and evaluation: Perform quality inspection on the fused data, including checking the color, contrast, clarity, etc. of the image.

[0084] Optionally, the naming and description of the rock and soil core sample is determined by spectral identification to determine the mineral composition and its proportion in the rock core;

[0085] Said nomenclature is based on the principles of rock and mineral identification;

[0086] The database also has data backup, recovery and security access control functions.

[0087] Technical effects and advantages of the present invention: The rock and soil core sample identification system of the present invention can quickly obtain the image characteristics and spectral characteristics of the rock and soil core sample through the spectrum and image acquisition equipment, and can automatically identify the image characteristic information of each layer of rock and soil by using the image recognition module, and then determine the name of each layer of rock and soil by the image characteristic information of each layer of rock and soil, and combine the spectral characteristic information to more accurately and quickly perform precise naming, thereby improving efficiency and correctness. In addition, the image characteristic information, spectral characteristic information and naming of each layer of rock and soil are stored in a preset electronic table to form an electronic catalog table.

[0088] In this way, through spectral and image recognition technology, the system can objectively and accurately identify the mineral composition and properties of rock and soil core samples, reduce the interference of human factors, and improve the accuracy and reliability of cataloging; the automated and rapid data processing and analysis capabilities enable the system to significantly shorten the cataloging cycle and improve work efficiency. This helps to speed up the progress of the project and reduce labor costs; and the system can unify data format standards, realize the effective integration of multi-source data, and facilitate the comprehensive use of data.

[0089] Further features and advantages of the present invention will become apparent from the following detailed description of exemplary embodiments of the present invention with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0090] Figure 1 It is a system workflow diagram provided by the present invention;

[0091] Figure 2 It is a working schematic diagram of each module provided by the present invention;

[0092] Figure 3 It is a flow chart of rock and soil layer classification and identification provided by the present invention. DETAILED DESCRIPTION

[0093] In order to make the purpose, technical scheme and advantages of the present invention clearer, the present invention is further described in detail below in conjunction with specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0094] like Figures 1 to 3 As shown, an embodiment of the present invention provides a rock and soil core sample information cataloging system based on spectrum and image recognition, comprising:

[0095] Spectral imaging module: Use a spectrometer to collect spectral data of rock and soil core samples, compare the spectral curves recorded in the spectral database, and classify and identify rock and soil layers through matching algorithms;

[0096] Image recognition module: captures images of rock and soil core samples, compares them with the preset image sample library, and identifies rock and soil layer features in the image, including texture information, granularity information, color information, integrity information, and crack development degree information;

[0097] Data processing module: receives spectral data and image data, and performs tasks such as data preprocessing, feature extraction, and data fusion to provide data support for the naming and determination module;

[0098] Naming and determination module: Based on the output of the data processing module, the spectral characteristics and image characteristics are integrated to determine the naming and description of the rock and soil core sample;

[0099] Data cataloging module: Integrate the output results of all modules such as spectral imaging, image recognition, and naming determination to generate geotechnical core sample information cataloging reports and databases;

[0100] Database: Stores and manages spectral database, image sample library, catalog report and other data, and provides data and query functions for each module.

[0101] Furthermore, by integrating spectral imaging, image recognition, data processing, naming, data cataloging and database management modules into rock and soil core sample analysis, significant improvements have been achieved. The system uses a spectrometer to accurately collect spectral data and combines image recognition technology to capture the intuitive features of rock and soil core samples, thus achieving accurate classification and identification of rock and soil layer components;

[0102] The automated and intelligent data processing process not only improves work efficiency, but also reduces human errors, ensuring the accuracy and comprehensiveness of the analysis results. The naming and determination module integrates spectral and image features to provide detailed naming and description for rock and soil core samples;

[0103] Finally, all information was integrated into the catalog report and database, realizing the centralized management and efficient use of data. This system not only provides reliable data support for geotechnical engineering investigation, design and construction, but also promotes the in-depth development of geology and related scientific research fields, and is an important manifestation of technological progress in the field of geotechnical engineering.

[0104] Specifically, the spectral imaging module includes mineral composition information, water content information, color center information and energy level information;

[0105] The mineral component information determination unit: identifies the mineral wavelength information of each layer of rock and soil through the spectrum, matches it with the spectrum database, sets the mineral wavelength identifier corresponding to the mineral component with the greatest similarity, and uses it as the mineral component identifier of the corresponding layer of rock and soil;

[0106] The moisture content information determination unit: by detecting the fluorescence emitted by each layer of rock and soil under high-energy X-ray irradiation, analyzing the element composition of the rock and soil material, and calculating the moisture content according to the relative content of the elements;

[0107] The color center information determination unit: by comparing the trench database, sets the color center mark corresponding to the rock and soil color center feature with the greatest similarity as the color center mark of the corresponding layer of rock and soil;

[0108] The classification and identification of rock and soil layers are specifically as follows:

[0109] 1) Use a spectrometer to collect spectral data of rock and soil core samples and measure the wavelength distribution of light reflected or emitted from the surface of the rock and soil core samples;

[0110] 2) matching the collected spectral data with known spectral curves in the spectral database;

[0111] 3) Preliminary identification of the characteristics of the rock and soil layers based on the matching results of the spectral data;

[0112] 4) Classify the rock and soil layers according to the identification results.

[0113] Furthermore, the spectral imaging module includes precise determination of mineral composition, water content, color center and energy level information in the analysis of rock and soil core samples. The mineral composition information determination unit uses spectral technology to identify the mineral wavelength of rock and soil, and accurately identifies the mineral composition of each layer of rock and soil after matching with the database, laying the foundation for geological analysis; the water content information determination unit cleverly uses high-energy X-ray fluorescence analysis to calculate the water content of rock and soil, providing key data for evaluating rock and soil stability and engineering design; the color center information determination unit identifies the color center characteristics of rock and soil by comparing the trench database, further enriching the understanding of the appearance characteristics of rock and soil; it not only improves the accuracy and comprehensiveness of rock and soil core sample analysis, but also promotes the scientific nature of engineering decision-making and ensures the safety and reliability of geotechnical engineering.

[0114] Specifically, the collected spectral data is matched with known spectral curves in the spectral database, specifically:

[0115] Set the spectrum vector x = (x1, x2, ..., x n ) and y=(y1,y2,...,y n ), the formula of spectral correlation coefficient r is:

[0116]

[0117] Where: n is the number of wavelength points of the spectrum;

[0118] and are the means of x and y, namely:

[0119]

[0120] The value range of the correlation coefficient r is [-1,1];

[0121] When r = 1, it means that the contrast spectra are completely positively correlated;

[0122] When r = -1, it means that the contrast spectra are completely negatively correlated;

[0123] When r=0, it means that there is no linear correlation between the contrast spectra.

[0124] Furthermore, when r = 1, it means that the two spectral data are highly consistent in terms of waveform, peak position and intensity, showing a trend of synchronous change, indicating that the rocks and soils in different layers or different regions may have similar mineral compositions or chemical characteristics.

[0125] When r = -1, it means that the two spectral data are mirror-symmetric in the waveform, that is, the peak of one spectrum corresponds to the valley of the other spectrum, indicating a strong inverse relationship between the two, indicating the presence of different mineral compositions or chemical environments.

[0126] When r = 0, it means that the two spectral data have no significant synchronous or reverse change trend in terms of waveform, peak position and intensity, indicating that there are significant differences in mineral composition, chemical characteristics and other aspects of the rocks and soils in different layers or regions.

[0127] Specifically, the specific workflow of the data processing module is as follows:

[0128] Preprocessing: preprocess the collected data separately, including denoising, enhancement, etc.

[0129] Spectral feature extraction: extract spectral features from spectral data, including peak wavelength, absorption or reflection intensity, spectral slope, etc.;

[0130] Image feature extraction: extracting features such as shape, texture, and color from image data;

[0131] Data fusion: Fusion of extracted spectral data with image data;

[0132] Data output: Output the processed data and use it as input for subsequent steps;

[0133] The processed data includes feature vectors and fused data;

[0134] The subsequent steps include classification, identification, naming, etc.

[0135] Specifically, the collected data are preprocessed separately, specifically as follows:

[0136] The process of spectral data preprocessing is as follows:

[0137] 1) Baseline correction: remove baseline offset or drift caused by instrument, sample container or other factors;

[0138] 2) Wavelet transform denoising: Split the original signal into wavelet coefficients of different frequencies, set a threshold based on the data, remove the noise part below the threshold, perform inverse wavelet transform on the retained part, and resynthesize it into a denoised signal;

[0139] 3) Smoothing: Reduce the curve fluctuation of spectral data through filtering technology;

[0140] 4) Normalization: Normalize the spectral data and synchronize the intensity ratio of the spectrum;

[0141] 5) Spectral enhancement processing: Transformation based on the multispectral database band, including principal component transformation, independent component transformation, color space transformation and color stretching to enhance the spectrum;

[0142] The process of image data preprocessing is as follows:

[0143] 1) Filtering: Use the Gaussian template to scan each pixel in the image.

[0144] 2) Calculate the product of each pixel in the Gaussian template coverage area and the corresponding Gaussian kernel weight, add the obtained products, and output the weighted sum;

[0145] 3) Assign the calculated weighted average value to the central pixel point and output it as the filtered pixel value;

[0146] 4) Feature enhancement: including operations such as adjusting the grayscale distribution of the image, sharpening, spatial domain enhancement, and frequency domain enhancement.

[0147] Furthermore, the spectral data preprocessing effectively removed interference factors such as instrument noise and baseline offset through baseline correction, wavelet transform denoising, smoothing, normalization and spectral enhancement, and improved the accuracy and reliability of spectral data. It not only reduced the volatility and error of the data, but also made the spectral data between different samples more comparable through normalization and enhancement processing, providing a high-quality data foundation for subsequent analysis.

[0148] Secondly, image data preprocessing removes image noise through filtering and uses Gaussian filtering and other techniques to smooth the image and improve image quality. At the same time, feature enhancement processing further improves the visual effect of the image, including adjusting the grayscale distribution, sharpening the edges, and enhancing the spatial and frequency domain features, so that the geotechnical features in the image are more prominent, which is convenient for subsequent image recognition and feature extraction.

[0149] It should be noted that when preprocessing spectral and image data, the accuracy of baseline correction directly affects the authenticity and reliability of spectral data. Therefore, when selecting a baseline correction method, it is necessary to determine it based on the actual situation and data characteristics to ensure the best correction effect. At the same time, if the threshold in the wavelet transform denoising process is too high, important information in the signal may be deleted by mistake, while if the threshold is too low, the noise may not be effectively removed. Therefore, it is necessary to focus on testing and then determine it.

[0150] Specifically, the spectral feature extraction is specifically as follows:

[0151] Peak identification: Analyze the spectral data, identify the light intensity peak at each wavelength point, record the corresponding wavelength of each peak, and record it as the peak wavelength;

[0152] Calculate intensity: At the peak wavelength or specific wavelength, measure the light intensity value at that wavelength, calculate the absorption or reflection intensity of the spectral data, and analyze the change trend of the absorption or reflection intensity at different wavelengths;

[0153] Slope extraction: Calculate the spectral slope within a specific interval of spectral data and analyze the changes in the spectral slope within different intervals;

[0154] Characteristic parameter definition: According to actual needs, define a series of characteristic parameters to describe the spectral characteristics, including peak wavelength, absorption or reflection intensity, spectral slope, etc. According to the defined characteristic parameters, calculate and extract the corresponding values;

[0155] Result analysis: Analyze the extracted spectral features to understand the optical characteristics, chemical properties and other information of the object or light source, and compare the extracted features with known standards.

[0156] Specifically, the formula of the slope extraction algorithm is:

[0157]

[0158] Where: x i is the wavelength of the spectral data;

[0159] y i is the spectral intensity of the corresponding wavelength;

[0160] w i is the weight of the i-th observation;

[0161] and are the slope and intercept that need to be estimated;

[0162] The slope and intercept estimates are calculated as follows:

[0163]

[0164] in: and is the weighted mean;

[0165]

[0166] The w i Determined based on the signal-to-noise ratio and data quality of the spectral data.

[0167] Furthermore, calculating the slope within a specific interval of spectral data can reveal the changing trend of the spectral curve in the interval, and further reflect the changes in the physical or chemical properties of the sample within the band. The present invention assigns different weights to different observations, and weighted linear regression can more accurately reflect the real relationship in the data, thereby improving the accuracy and reliability of slope calculation.

[0168] Secondly, the characteristics of weighted linear regression are its flexibility and adaptability. In the calculation of spectral slope, the appropriate weight distribution method can be selected according to the actual situation, such as weight setting based on factors such as data quality, measurement error, and sample importance. Weighted linear regression can better adapt to the characteristics and needs of different spectral data sets and improve the applicability and effectiveness of the analysis results.

[0169] In addition, weighted linear regression can also reduce the impact of outliers or noise on slope calculation to a certain extent. By reducing the weight of unreliable observations, weighted linear regression can make the model more robust and avoid slope calculation deviations caused by individual outliers.

[0170] Specifically, the image feature extraction is specifically as follows:

[0171] 1) Receive and process the image;

[0172] 2) Color feature extraction: Count the distribution of each color component in the image to form a color histogram, convert the color space through color index, and extract relevant features;

[0173] The color distribution is represented by a color moment;

[0174] 3) Texture feature extraction:

[0175] A. Analyze the gray level dependency between a pixel and its neighboring pixels to extract texture features;

[0176] B. Compare the grayscale values ​​of the central pixel with those of its neighboring pixels and encode the result into a binary number to extract features;

[0177] C. Use wavelet transform to decompose the image and extract texture features at different scales and directions;

[0178] 4) Shape feature extraction: extract edge information, shape, contour, and collection attributes from the image through relevant algorithms;

[0179] 5) Feature dimensionality reduction: Use principal component analysis, linear discriminant analysis and other methods to reduce feature dimensions and select relevant feature subsets.

[0180] Exemplarily, spectral recognition also has feature dimensionality reduction, which is used to achieve data dimensionality reduction and denoising, reduce computational complexity, optimize data storage, and improve signal-to-noise ratio (SNR). Subspace projection maps high-dimensional spectral data to low-dimensional subspace, providing higher quality data for subsequent spectral recognition and unmixing. The relevant formula (I) is:

[0181] Y′=PY=P(MX+N)=PMX+N′

[0182] Where: M is the characteristic component of the spectrum of pure substance, called end member;

[0183] Y is the observed original hyperspectral data matrix, which contains the spectral vectors of multiple pixels;

[0184] Y' is the spectral data matrix after subspace projection, which is located in a low-rank subspace;

[0185] P is the low-rank subspace projection matrix;

[0186] X is the abundance matrix, which contains the contribution (i.e. abundance) of each end member in each pixel;

[0187] N is the noise matrix in the original spectral data;

[0188] N' is the additional error or noise that may be introduced during the subspace projection process;

[0189] It should be noted that, since the number of end members M is often much smaller than the number of bands L, in order to improve the accuracy of spectral recognition, the spectral vector is in a low-dimensional subspace, and the subspace projection allows the observed spectral vector to be represented in the low-dimensional subspace. In addition, given that the subspace projection has obvious gain effects in terms of computational complexity, data storage, and signal-to-noise ratio (SNR), the subspace projection P is introduced into the hyperspectral unmixing, so that the spectral variability can be processed by solving the denoising problem;

[0190] Furthermore, in the learned subspace, in order to improve the correlation of similar pixels and transform the original hyperspectral data into a robust subspace, we embed low-rank attributes into it and construct an overall unmixing model through the projected data;

[0191] The overall unmixing model related algorithm formula (II) can be expressed as:

[0192] Y=Y′+N′

[0193] stY′=PY,Y′=PMX+N″

[0194] in:

[0195] N" is the error after low-rank subspace projection processing;

[0196] Furthermore, unlike the matrix decomposition-based unmixing method that estimates endmembers and abundance simultaneously, HySime is used to first determine the number of endmembers, and then the VCA algorithm is used to extract each endmember from the hyperspectral image, and then the abundance map is estimated. The LRSU algorithm is proposed through comprehensive analysis, and the algorithm formula (III) related to the constrained optimization problem is expressed as follows:

[0197]

[0198] stX≥0

[0199] Among them: α>0, β>0, γ>0 are all penalty parameters;

[0200] In the model, the above scheme has dual advantages. On the one hand, because end members are extracted directly from hyperspectral data, it can retain the physical meaning of the spectrum to the maximum extent and avoid the influence of some meaningless errors, thereby improving the stability of the subsequent abundance estimation process. On the other hand, the process of preprocessing to obtain end members requires fewer variables to be optimized, which further simplifies the complexity of the model and facilitates the subsequent acquisition of better solutions.

[0201] It should be noted that, since problem (III) is indeterminate, the variables X and P need to be regularized;

[0202] Regarding the solution of problem (III), considering that it is a convex problem, an optimization algorithm based on ADMM is used to solve it efficiently and quickly. By introducing multiple auxiliary variables G, I and J, the relevant formula (IV) is:

[0203]

[0204] stP=G,X=I,X=J

[0205] in,(·) +is an operator that extracts the non-negative part from each component of the matrix, that is, Defined as J ≥ 0. Formula (IV) can be equivalent to solving the following minimized augmented Lagrangian function, formula (V) is:

[0206]

[0207] Among them, Λ1, Λ2, Λ3 are all Lagrange multipliers, μ and α are penalty parameters;

[0208] Furthermore, in spectral image processing, by minimizing the augmented Lagrangian function, it can be ensured that the physical characteristics and constraints of the data are fully considered during the optimization process, thereby avoiding meaningless errors caused by inappropriate data processing or model assumptions. In addition, the regularization terms in the augmented Lagrangian function (such as sparsity constraints, low-rank constraints, etc.) can further guide the optimization process in a direction that is more in line with physical meaning and practical application requirements, thereby improving the accuracy and stability of the solution.

[0209] Specifically, the data fusion is specifically as follows:

[0210] Spatial registration: align the spectral data and image data in the same coordinate system and adjust them to the same spatial resolution and coverage;

[0211] Select fusion methods: including pixel-level fusion, feature-level fusion and decision-level fusion;

[0212] Perform data fusion: fuse the pre-processed spectral data and image data through data processing software or algorithm according to the selected fusion method;

[0213] Quality inspection and evaluation: Perform quality inspection on the fused data, including checking the color, contrast, clarity, etc. of the image.

[0214] Furthermore, through spatial registration, the spectral data and image data can be accurately aligned in the same coordinate system, providing accurate spatial correspondence for subsequent data fusion, which helps to improve the accuracy and reliability of the fusion results;

[0215] It should be noted that the choice of fusion method should be determined according to the specific data characteristics, application requirements and expected results. The pixel-level fusion method is simple and direct, but may lose some feature information; the feature-level fusion method can retain more feature information, but the computational complexity is higher; the decision-level fusion method pays more attention to the decision-making value of the fusion result;

[0216] Furthermore, appropriate fusion methods can give full play to the advantages of spectral data and image data, realize complementary advantages, and improve the comprehensiveness and accuracy of fusion results. At the same time, it helps to meet different application needs and provide richer data support for geotechnical engineering surveys;

[0217] Specifically, the naming and description of the rock and soil core samples are determined by spectral identification to determine the mineral composition and its proportion in the core;

[0218] Said nomenclature is based on the principles of rock and mineral identification;

[0219] The database also has data backup, recovery and security access control functions

[0220] Exemplary: List the description and naming principles of mudstone (shale):

[0221] a) Description: including the content and distribution of terrigenous debris components, clay minerals, endogenous debris, authigenic minerals, carbonized plant debris, heavy oil, asphalt, etc., and the structure and texture of the rock.

[0222] b) Naming of mineral components: When the content of a certain mineral is greater than or equal to 50%, the name of the mineral is directly added before the main name "mud (shale) rock", such as kaolinite mudstone; when the content is greater than or equal to 25% and less than 50%, it is named "xx quality", such as gray mudstone; when the content is greater than or equal to 10% and less than 25%, it is generally not involved in the naming. When the content of two or more minerals is greater than or equal to 25% and less than 50%, the order of less first and more last is used for naming.

[0223] c) Naming of sand and silt content: When the total amount of debris greater than or equal to 0.0156 mm is estimated to be greater than or equal to 10% and less than 25%, it is named "containing xx", such as silt-containing mudstone; when the content is greater than or equal to 25% and less than 50%, it is named "xx quality", such as silt-containing mudstone.

[0224] d) Shale with a bituminous substance or kerogen content greater than or equal to 5% and mainly composed of clay minerals is named oil shale.

[0225] e) Naming of endogenous debris and carbonaceous materials: When the content of a certain component is greater than or equal to 10% and less than 25%, it is named "containing xx", such as oolitic mudstone; when the content is greater than or equal to 25% and less than 50%, it is named "xx", such as carbonaceous mudstone.

[0226] f) Naming of pyroclastic materials of the same period: When the content of pyroclastic materials of the same period with a particle size less than 2 mm is greater than or equal to 10% and less than 50%, it is named "tuffaceous xx"

[0227] In summary, the rock and soil core sample identification system of the present invention can quickly obtain the image features and spectral features of the rock and soil core sample through the spectrum and image acquisition device, and can automatically identify the image feature information of each layer of rock and soil by using the image recognition module, and then determine the name of each layer of rock and soil by the image feature information of each layer of rock and soil, and combine the spectral feature information to more accurately and quickly perform precise naming, thereby improving efficiency and correctness. In addition, the image feature information, spectral feature information and naming of each layer of rock and soil are stored in a preset electronic table to form an electronic catalog table.

[0228] In this way, through spectral and image recognition technology, the system can objectively and accurately identify the mineral composition and properties of rock and soil core samples, reduce the interference of human factors, and improve the accuracy and reliability of cataloging; the automated and rapid data processing and analysis capabilities enable the system to significantly shorten the cataloging cycle and improve work efficiency. This helps to speed up the progress of the project and reduce labor costs; and the system can unify data format standards, realize the effective integration of multi-source data, and facilitate the comprehensive use of data.

[0229] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A rock and soil core sample information cataloging system based on spectrum and image recognition, characterized in that: include: Spectral imaging module: Use a spectrometer to collect spectral data of rock and soil core samples, compare the spectral curves recorded in the spectral database, and classify and identify rock and soil layers through matching algorithms; Image recognition module: captures images of rock and soil core samples, compares them with the preset image sample library, and identifies rock and soil layer features in the image, including texture information, granularity information, color information, integrity information, and crack development degree information; Data processing module: receives spectral data and image data, and performs tasks such as data preprocessing, feature extraction, and data fusion to provide data support for the naming and determination module; Naming and determination module: Based on the output of the data processing module, the spectral characteristics and image characteristics are integrated to determine the naming and description of the rock and soil core sample; Data cataloging module: Integrate the output results of all modules such as spectral imaging, image recognition, and naming determination to generate geotechnical core sample information cataloging reports and databases; Database: Stores and manages spectral database, image sample library, catalog report and other data, and provides data and query functions for each module.

2. The rock and soil core sample information cataloging system based on spectrum and image recognition according to claim 1 is characterized by: The spectral imaging module includes mineral composition information, water content information, color center information and energy level information; The mineral component information determination unit: identifies the mineral wavelength information of each layer of rock and soil through the spectrum, matches it with the spectrum database, sets the mineral wavelength identifier corresponding to the mineral component with the greatest similarity, and uses it as the mineral component identifier of the corresponding layer of rock and soil; The moisture content information determination unit: by detecting the fluorescence emitted by each layer of rock and soil under high-energy X-ray irradiation, analyzing the element composition of the rock and soil material, and calculating the moisture content according to the relative content of the elements; The color center information determination unit: by comparing the trench database, sets the color center mark corresponding to the rock and soil color center feature with the greatest similarity as the color center mark of the corresponding layer of rock and soil; The classification and identification of rock and soil layers are specifically as follows: 1) Use a spectrometer to collect spectral data of rock and soil core samples and measure the wavelength distribution of light reflected or emitted from the surface of the rock and soil core samples; 2) matching the collected spectral data with known spectral curves in the spectral database; 3) Preliminary identification of the characteristics of the rock and soil layers based on the matching results of the spectral data; 4) Classify the rock and soil layers according to the identification results.

3. The rock and soil core sample information cataloging system based on spectrum and image recognition according to claim 1 is characterized by: The matching of the collected spectral data with the known spectral curves in the spectral database is specifically as follows: Set the spectrum vector x = (x1, x2, ..., x n ) and y=(y1,y2,...,y n ), the formula of spectral correlation coefficient r is: Where: n is the number of wavelength points of the spectrum; and are the means of x and y, namely: The value range of the correlation coefficient r is [-1,1]; When r = 1, it means that the contrast spectra are completely positively correlated; When r = -1, it means that the contrast spectra are completely negatively correlated; When r=0, it means that there is no linear correlation between the contrast spectra.

4. The rock and soil core sample information cataloging system based on spectrum and image recognition according to claim 1 is characterized by: The specific workflow of the data processing module is as follows: Preprocessing: preprocess the collected data separately, including denoising, enhancement, etc. Spectral feature extraction: extract spectral features from spectral data, including peak wavelength, absorption or reflection intensity, spectral slope, etc.; Image feature extraction: extracting features such as shape, texture, and color from image data; Data fusion: Fusion of extracted spectral data with image data; Data output: Output the processed data and use it as input for subsequent steps; The processed data includes feature vectors and fused data; The subsequent steps include classification, identification, naming, etc.

5. The rock and soil core sample information cataloging system based on spectrum and image recognition according to claim 4 is characterized in that: The collected data are preprocessed respectively, specifically: The process of spectral data preprocessing is as follows: 1) Baseline correction: remove baseline offset or drift caused by instrument, sample container or other factors; 2) Wavelet transform denoising: Split the original signal into wavelet coefficients of different frequencies, set a threshold based on the data, remove the noise part below the threshold, perform inverse wavelet transform on the retained part, and resynthesize it into a denoised signal; 3) Smoothing: Reduce the curve fluctuation of spectral data through filtering technology; 4) Normalization: Normalize the spectral data and synchronize the intensity ratio of the spectrum; 5) Spectral enhancement processing: Transformation based on the multispectral database band, including principal component transformation, independent component transformation, color space transformation and color stretching to enhance the spectrum; The process of image data preprocessing is as follows: 1) Filtering: Use the Gaussian template to scan each pixel in the image. 2) Calculate the product of each pixel in the Gaussian template coverage area and the corresponding Gaussian kernel weight, add the obtained products, and output the weighted sum; 3) Assign the calculated weighted average value to the central pixel point and output it as the filtered pixel value; 4) Feature enhancement: including operations such as adjusting the grayscale distribution of the image, sharpening, spatial domain enhancement, and frequency domain enhancement.

6. The rock and soil core sample information cataloging system based on spectrum and image recognition according to claim 4 is characterized in that: The extracted spectral features are specifically: Peak identification: Analyze the spectral data, identify the light intensity peak at each wavelength point, record the corresponding wavelength of each peak, and record it as the peak wavelength; Calculate intensity: At the peak wavelength or specific wavelength, measure the light intensity value at that wavelength, calculate the absorption or reflection intensity of the spectral data, and analyze the change trend of the absorption or reflection intensity at different wavelengths; Slope extraction: Calculate the spectral slope within a specific interval of spectral data and analyze the changes in the spectral slope within different intervals; Characteristic parameter definition: According to actual needs, define a series of characteristic parameters to describe the spectral characteristics, including peak wavelength, absorption or reflection intensity, spectral slope, etc. According to the defined characteristic parameters, calculate and extract the corresponding values; Result analysis: Analyze the extracted spectral features to understand the optical characteristics, chemical properties and other information of the object or light source, and compare the extracted features with known standards.

7. A rock and soil core sample information cataloging system based on spectrum and image recognition according to claim 6, characterized in that: The formula for the slope extraction algorithm is: Where: x i is the wavelength of the spectral data; y i is the spectral intensity of the corresponding wavelength; w i is the weight of the i-th observation; and The slope and intercept need to be estimated; The slope and intercept estimates are calculated as follows: in: and is the weighted mean; The w i Determined based on the signal-to-noise ratio and data quality of the spectral data.

8. The rock and soil core sample information cataloging system based on spectrum and image recognition according to claim 1 is characterized by: The image feature extraction is specifically as follows: 1) Receive and process the image; 2) Color feature extraction: Count the distribution of each color component in the image to form a color histogram, convert the color space through color index, and extract relevant features; The color distribution is represented by a color moment; 3) Texture feature extraction: A. Analyze the gray level dependency between a pixel and its neighboring pixels to extract texture features; B. Compare the grayscale values ​​of the central pixel with those of its neighboring pixels and encode the result into a binary number to extract features; C. Use wavelet transform to decompose the image and extract texture features at different scales and directions; 4) Shape feature extraction: extract edge information, shape, contour, and collection attributes from the image through relevant algorithms; 5) Feature dimensionality reduction: Use principal component analysis, linear discriminant analysis and other methods to reduce feature dimensions and select relevant feature subsets.

9. The rock and soil core sample information cataloging system based on spectrum and image recognition according to claim 4, characterized in that: The data fusion is specifically as follows: Spatial registration: align the spectral data and image data in the same coordinate system and adjust them to the same spatial resolution and coverage; Select fusion methods: including pixel-level fusion, feature-level fusion and decision-level fusion; Perform data fusion: fuse the pre-processed spectral data and image data through data processing software or algorithm according to the selected fusion method; Quality inspection and evaluation: Perform quality inspection on the fused data, including checking the color, contrast, clarity, etc. of the image.

10. The rock and soil core sample information cataloging system based on spectrum and image recognition according to claim 1, characterized in that: The naming and description of the rock and soil core samples are determined by spectral identification to determine the mineral composition and its proportion in the core; Said nomenclature is based on the principles of rock and mineral identification; The database also has data backup, recovery and security access control functions.

Images