Large-scale dangerous rock mass identification method based on digital image deformation processing

Abstract

The present invention relates to the field of image data processing technology, and discloses a large-scale dangerous rock mass identification method based on digital image deformation processing, which aims to solve the problem of low efficiency and accuracy of the existing methods. The scheme mainly includes: obtaining two images of the target rock mass collected within two cycles after preprocessing, and generating two-dimensional grayscale matrices of the two images in the spatial domain respectively; performing two-dimensional discrete Fourier transform on the two-dimensional grayscale matrices of the two images respectively, calculating the cross-correlation function of the two images, calculating the first phase difference corresponding to the frequency domain according to the cross-correlation function, and determining the second phase difference corresponding to the spatial domain according to the first phase difference; obtaining the displacement field after phase demodulation according to the second phase difference, and calculating the displacement gradient and strain field according to the displacement field; determining the danger level of each area of ​​the target rock mass based on the pre-trained dangerous rock mass identification model. The present invention improves the efficiency and accuracy of dangerous rock mass identification, and is particularly suitable for large-scale dangerous rock mass identification.

Description

Technical Field

[0001] The invention relates to the technical field of image data processing, and in particular to a large dangerous rock body identification method based on digital image deformation processing. Background Art

[0002] Dangerous rock masses refer to rocks separated by cracks on steep slopes. These rocks may fall or collapse from the slopes under the influence of external factors such as vibration, heavy rain or strong wind, and have potential collapse risks. The formation of dangerous rock masses is related to many factors, including rock and soil types, geological structures, weathering, etc. In geological disaster prevention and control, the identification and monitoring of dangerous rock masses is an important part of ensuring engineering safety. It is necessary to regularly monitor and evaluate dangerous rock masses and take necessary reinforcement measures to reduce their collapse risks.

[0003] In the monitoring of dangerous rock masses, Digital Image Correlation (DIC) is widely used in real-time monitoring and early warning. DIC can accurately measure the displacement, deformation and other parameters of an object by analyzing the digital image of the object surface. It has the characteristics of non-contact, high precision and strong real-time performance. Its main principle is to obtain the displacement vector of the pixel by comparing the position change of the same pixel point in the two images before and after the deformation of the rock surface, so as to obtain the full-field displacement of the surface of the rock sample to be tested, and then determine whether the rock sample to be tested is a dangerous rock mass.

[0004] The basic principle of traditional DIC to determine pixel displacement is to grid the area of ​​interest in the image before deformation and treat each sub-region as rigid motion. Then, for each sub-region, a certain search method is used to perform correlation calculations according to a pre-defined correlation function, and the area with the maximum mutual correlation coefficient with the sub-region is found in the deformed image, that is, the position of the sub-region after deformation is determined, and then the displacement of the sub-region is obtained. By calculating all sub-regions, the deformation information of the entire field can be obtained. However, in complex geological environments, especially for the identification of large dangerous rock bodies on steep slopes, a large amount of image data needs to be processed. The traditional DIC implementation method is inefficient when processing a large amount of image data, and the recognition ability for weak deformation is limited, resulting in poor accuracy in the identification of dangerous rock bodies. Summary of the invention

[0005] The present invention aims to solve the problem of low efficiency and accuracy of the existing digital image correlation method in identifying dangerous rock bodies, and proposes a large-scale dangerous rock body identification method based on digital image deformation processing.

[0006] The technical solution adopted by the present invention to solve the above technical problems is:

[0007] A large dangerous rock mass identification method based on digital image deformation processing, the method comprising:

[0008] Collect images of the target rock mass according to a preset period and pre-process the images;

[0009] Obtain two images collected in two cycles after preprocessing, and generate two-dimensional grayscale matrices of the two images in the spatial domain respectively;

[0010] Performing a two-dimensional discrete Fourier transform on the two-dimensional grayscale matrices of the two images respectively to obtain frequency domain matrices of the two images, calculating a cross-correlation function of the two images according to the frequency domain matrices, calculating a first phase difference corresponding to the frequency domain according to the cross-correlation function, and determining a second phase difference corresponding to the spatial domain according to the first phase difference;

[0011] According to the second phase difference and based on the pre-fitted displacement function, a displacement field after phase demodulation is obtained, and a displacement gradient and a strain field are calculated according to the displacement field;

[0012] The danger level of each area of ​​the target rock mass is determined according to the displacement field, displacement gradient and strain field and based on a pre-trained dangerous rock mass identification model.

[0013] Furthermore, the calculation formula of the two-dimensional discrete Fourier transform is as follows:

[0014] ;

[0015] in, express The frequency domain matrix, Represents the two-dimensional grayscale matrix of the image in the spatial domain, Represents the spatial coordinates of the pixel point, represents the frequency domain coordinates, , , Indicates the number of pixels in the horizontal direction of the image. Indicates the number of pixels in the vertical direction of the image. represents a natural constant, Represents an imaginary unit.

[0016] Furthermore, the calculation formula of the cross-correlation function is as follows:

[0017] ;

[0018] in, represents the cross-correlation function, Represents the frequency domain matrix corresponding to the image collected in the first cycle, Represents the conjugate complex matrix of the frequency domain matrix corresponding to the image collected in the second period;

[0019] The calculation formula of the first phase difference is as follows:

[0020] ;

[0021] in, represents the first phase difference, express The imaginary part of express The real part of .

[0022] Furthermore, the displacement field includes the horizontal displacement and vertical displacement of each pixel point, and the calculation formula is as follows:

[0023] ;

[0024] ;

[0025] in, represents the horizontal displacement, represents the vertical displacement, represents the second phase difference, , , and Represents the coefficient.

[0026] Furthermore, the displacement gradient includes a horizontal displacement gradient and a vertical displacement gradient, and the calculation formula is as follows:

[0027] ;

[0028] ;

[0029] in, represents the horizontal displacement gradient, represents the vertical displacement gradient, Indicates the horizontal displacement in The partial derivative in the direction, Indicates the vertical displacement in The partial derivative in the direction, Indicates the vertical displacement in The partial derivative in the direction, Indicates the horizontal displacement in Partial derivatives in the direction.

[0030] Furthermore, the strain field includes horizontal strain, vertical strain and shear strain, and the calculation formula is as follows:

[0031] ;

[0032] ;

[0033] ;

[0034] in, represents the horizontal strain, represents the vertical strain, represents shear strain.

[0035] Furthermore, the method further comprises:

[0036] Before acquiring an image of the target rock mass, a speckle material is coated on the surface of the target rock mass to form a uniform speckle pattern on the surface of the target rock mass.

[0037] Further, the image is preprocessed, including: removing noise, enhancing contrast, registering the image, and cropping the image;

[0038] The noise removal method includes high-pass filtering, mean filtering and Gaussian filtering, the contrast enhancement method includes adaptive histogram equalization and Gamma correction, and the image registration method includes feature point matching, phase correlation method and cross-correlation method.

[0039] Furthermore, the training method of the dangerous rock mass identification model includes:

[0040] Obtain images of rock samples in two cycles and preprocess them;

[0041] Determine the displacement field, displacement gradient and strain field of the rock sample based on the preprocessed image, and mark the hazard level of each area of ​​the rock sample based on the deformation area, deformation rate and deformation gradient;

[0042] The displacement field, displacement gradient and strain field of the rock sample are taken as input features, and the danger level of the corresponding area is taken as the output label to construct a training data set, and the training data set is used to train a dangerous rock mass identification model.

[0043] Furthermore, the method further comprises:

[0044] According to the danger level of each area of ​​the target rock mass, corresponding graphic annotations are made in the rock mass model;

[0045] Determine the area corresponding to each danger level, determine the warning level of the target rock mass according to the size and change rate of the area, and issue a warning according to the warning level.

[0046] The beneficial effects of the present invention are as follows: the large-scale dangerous rock mass identification method based on digital image deformation processing provided by the present invention converts the two-dimensional grayscale matrix of the digital image in the spatial domain into a frequency domain matrix, calculates the cross-correlation function and phase difference of the two images before and after the target rock mass is deformed in the frequency domain, and then calculates the displacement field, displacement gradient and strain field. Since the complex convolution operation in the spatial domain is changed into a simple product operation in the frequency domain, the deformation position determination process of the traditional DIC is also avoided, and useful information can be quickly and effectively extracted from a large amount of image data, thereby simplifying the image processing process and improving the efficiency of dangerous rock mass identification. Moreover, the calculation of cross-correlation functions and phase differences based on frequency domain analysis can reduce the interference of factors such as noise, capture the tiny displacement changes of the rock mass, and more accurately reveal the differences and changes between images, improve the accuracy of the analysis results, and provide a basis for the subsequent accurate calculation of the displacement field. In addition, the present invention constructs a dangerous rock mass identification model and inputs the displacement field, displacement gradient and strain field of the target rock mass into the dangerous rock mass identification model to obtain the danger level of the target rock mass. It not only comprehensively describes the mechanical state of the rock mass from multiple dimensions, but also simplifies the dangerous rock mass identification process, and can accurately identify the slight deformation on the surface of the dangerous rock mass, further improving the efficiency and accuracy of dangerous rock mass identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 A schematic flow chart of a large dangerous rock mass identification method based on digital image deformation processing provided in an embodiment;

[0048] Figure 2 A schematic diagram of a speckle pattern provided for an embodiment;

[0049] Figure 3 A schematic diagram of the installation of a high-definition camera provided in an embodiment;

[0050] Figure 4 A schematic diagram of an image collected in the first cycle after preprocessing provided in an embodiment;

[0051] Figure 5 A schematic diagram of an image collected in the second cycle after preprocessing provided in an embodiment;

[0052] Figure 6 A diagrammatic illustration of a diagram provided for an embodiment. DETAILED DESCRIPTION

[0053] In order to enable those skilled in the art to better understand the solution of the present invention, the technical solution in this embodiment will be clearly and completely described below in conjunction with the drawings in this embodiment.

[0054] In some of the processes described in the specification of the present invention and the above-mentioned figures, multiple operations appearing in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this document or may be executed in parallel. The sequence numbers of the operations are only used to distinguish different operations, and the sequence numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel.

[0055] In order to improve the efficiency and accuracy of dangerous rock mass identification, the technical solution of the present invention is proposed. In the present invention, images of the target rock mass are collected according to a preset period and the images are preprocessed; two images collected within two periods after preprocessing are obtained, and two-dimensional grayscale matrices of the two images in the spatial domain are generated respectively; two-dimensional discrete Fourier transform is performed on the two-dimensional grayscale matrices of the two images respectively to obtain frequency domain matrices of the two images, the cross-correlation function of the two images is calculated according to the frequency domain matrix, the first phase difference corresponding to the frequency domain is calculated according to the cross-correlation function, and the second phase difference corresponding to the spatial domain is determined according to the first phase difference; according to the second phase difference and based on the pre-fitted displacement function, the displacement field after phase demodulation is obtained, and the displacement gradient and strain field are calculated according to the displacement field; according to the displacement field, the displacement gradient and the strain field and based on the pre-trained dangerous rock mass identification model, the danger level of each area of ​​the target rock mass is determined.

[0056] Specifically, the present invention first preprocesses the images of the target rock mass collected within two periods and generates a two-dimensional grayscale matrix of the two images in the spatial domain; then performs a two-dimensional discrete Fourier transform on them to generate a corresponding frequency domain matrix, calculates the cross-correlation function and phase difference between each pixel point of the two images based on the frequency domain matrix, and calculates the displacement of each pixel point according to the phase difference to obtain the displacement field, and then calculates the displacement gradient and strain field according to the displacement field; finally, the displacement field, displacement gradient and strain field are input into a pre-trained dangerous rock mass identification model to obtain the danger level of each area of ​​the target rock mass.

[0057] The technical solution in this embodiment will be clearly and completely described below in conjunction with the drawings in this embodiment. Obviously, the described embodiment is only a part of the embodiments of the present invention, rather than all the embodiments.

[0058] Figure 1 A schematic diagram of a large dangerous rock mass identification method based on digital image deformation processing is shown in FIG. Figure 1 , the method comprises the following steps:

[0059] Step 1: Collect images of the target rock mass according to a preset period and pre-process the images.

[0060] In this embodiment, before acquiring an image of the target rock mass, a speckle material is coated on the surface of the target rock mass to form a uniform speckle pattern on the surface of the target rock mass.

[0061] In practical applications, a high-definition camera can be used to preliminarily identify the target rock mass to be identified. After taking pictures, it can be observed to see whether the surface feature points of the target rock mass are clear and whether the cracks, bedding and other features on the surface of the target rock mass can be observed. If the feature points are obvious, the next step can be directly performed. If not, it is necessary to coat the surface of the target rock mass with appropriate speckle material (i.e., target) to form a uniform speckle pattern (non-contact spraying operations can be directly performed using drones, etc., which can ensure the safety of workers in construction and is convenient and fast). The speckle pattern should have sufficient contrast to facilitate subsequent image processing. The following points should be noted when spraying:

[0062] (1) Layered spraying: First spray a layer of base color (such as white), and then spray another color (such as black) on it to create contrast. You can use staggered spraying to form a spotted pattern, such as Figure 2 As shown. (2) Spraying distance: Keep the spraying equipment about 15-30 cm away from the rock surface to ensure uniform spraying and no dripping. (3) Spraying material: The spraying material should have good adhesion and weather resistance, such as matte version spray paint, to ensure that it is not easy to fall off in the natural environment. The material particle size should be between 1-5 mm, depending on the roughness of the rock surface. (4) Pattern design: When spraying, you can spray randomly to form an irregular speckle pattern, or spray at specific intervals to form a regular grid or dot pattern. Ensure that the speckle is evenly distributed and avoid large areas of single color. Among them, there are two types of targets, namely natural light sources and artificial light sources. The former can only be used to collect images of the target rock surface during the day (the speckle material described in this embodiment is mainly suitable for natural light sources), and the latter can emit a certain frequency band of light sources, which can be used to collect images of the target rock surface throughout the day.

[0063] In practical applications, multiple fixed piles can be set up in the area where the large target rock mass is located. A high-definition camera can be installed on each pile. By setting different shooting positions and angles, images covering the surface of the target rock mass can be obtained to ensure full coverage of the surface of the target rock mass. Generally, three high-definition cameras can be set up, namely, on both sides and in front of the target rock mass. Figure 3As shown. According to the on-site lighting conditions and monitoring range, adjust the camera parameters such as focal length, aperture, exposure time, etc. to obtain a clear and stable image sequence. The image acquisition frequency is set according to the deformation rate of the dangerous rock mass and is collected regularly. Generally, it varies from every hour to every day. In order to ensure the shooting effect of the dangerous rock mass in different time periods before and after, once the high-definition camera is installed, the angle, position, etc. cannot be moved. At the same time, in order to ensure that the high-definition camera works for a long time, it is best to install a small solar cell on each camera to ensure that the camera can work 24 hours a day.

[0064] After the installation and debugging of the high-definition camera is completed, it can be automatically set to collect the surface image of the target rock mass. The collection frequency can be set to every few hours or every day. After rain, snow, strong wind and other weather conditions, the image collection frequency can be increased. The collected images need to be transferred back to the computer for analysis. There are two ways to upload images here. One is to directly connect the high-definition camera and the computer with a data connection cable such as a network cable. After the image is collected, it is automatically transferred back. The other is to install a wireless transmitter on the high-definition camera. After the image is collected, it is automatically uploaded to the network platform, and then the image can be downloaded from the platform. The former is suitable for working conditions where the distance between the high-definition camera and the computer is close, but the process is complicated; the latter is not subject to distance restrictions and the procedure is simple, but a wireless transmitter needs to be installed, and the right to use the network platform needs to be applied for, and the cost is relatively high.

[0065] After image acquisition, some preprocessing operations are usually required to improve the accuracy of subsequent analysis. In this embodiment, the image is preprocessed, including: noise removal, contrast enhancement, image registration and image cropping.

[0066] Specifically, the captured images may contain noise caused by factors such as ambient light and shooting angle. At this time, image denoising technology, such as high-pass filtering, mean filtering or Gaussian filtering, is needed to remove background noise and improve image quality. Using methods such as adaptive histogram equalization (CLAHE) or gamma correction for contrast enhancement can enhance the local contrast of the image, making the details of the rock mass in the image more obvious, which is convenient for subsequent analysis. In addition, multiple acquisitions or multi-angle shooting require registration of multiple images at different time points. Common registration methods include feature point matching, phase correlation or cross-correlation methods, so that images at different time points are aligned in the same coordinate system, and images taken at different angles are fused and spliced ​​to obtain images covering the surface of the target rock mass, ensuring that the subsequent deformation analysis has high accuracy. In addition, the captured images can also be cropped to retain key areas and remove unnecessary background information to reduce the complexity of subsequent processing.

[0067] This embodiment can accurately obtain the state information of the rock mass at different time points by collecting target rock mass images at preset periods and performing a series of processing. Image collection and processing are performed according to preset periods, so that the monitoring process has good repeatability. Data from different periods can be easily compared and analyzed, which helps to observe the changing trend of rock mass status over time, better predict the development and changes of rock mass, and provide strong support for long-term geological disaster monitoring and prevention.

[0068] Step 2: Obtain two images collected within two cycles after preprocessing, and generate two-dimensional grayscale matrices of the two images in the spatial domain respectively.

[0069] In this embodiment, the image obtained after preprocessing the image collected in the first cycle can be found in Figure 4 , the image obtained after preprocessing the image collected in the second cycle can be found in Figure 5 The corresponding two-dimensional grayscale matrix is ​​generated according to the preprocessed image. In this embodiment, it is assumed that the two-dimensional grayscale matrix of the image collected in the first cycle is , the image collected in the second cycle is a two-dimensional grayscale matrix , Represents the spatial coordinates of a pixel.

[0070] Step 3: perform a two-dimensional discrete Fourier transform on the two-dimensional grayscale matrices of the two images respectively to obtain frequency domain matrices of the two images, calculate the cross-correlation function of the two images according to the frequency domain matrix, calculate the first phase difference corresponding to the frequency domain according to the cross-correlation function, and determine the second phase difference corresponding to the spatial domain according to the first phase difference.

[0071] In this embodiment, the calculation formula of the two-dimensional discrete Fourier transform is as follows:

[0072] ;

[0073] in, express The frequency domain matrix, Represents the two-dimensional grayscale matrix of the image in the spatial domain, Represents the spatial coordinates of the pixel point, represents the frequency domain coordinates, , , Indicates the number of pixels in the horizontal direction of the image. Indicates the number of pixels in the vertical direction of the image. represents a natural constant, Represents an imaginary unit.

[0074] In practical applications, the two-dimensional grayscale matrices of the two images are subjected to two-dimensional discrete Fourier transform according to the above formula to obtain the frequency domain matrix of the image collected in the first cycle: And the frequency domain matrix of the image collected in the second cycle .

[0075] In this embodiment, the calculation formula of the cross-correlation function is as follows:

[0076] ;

[0077] in, represents the cross-correlation function, Represents the frequency domain matrix corresponding to the image collected in the first cycle, Represents the conjugate complex matrix of the frequency domain matrix corresponding to the image collected in the second period.

[0078] It can be understood that there are two common ways to calculate the cross-correlation function, namely using conjugate complex numbers and directly using the frequency domain matrix. The conjugate complex numbers are used here because in the theory of Fourier transform, the expression of the signal correlation operation in the frequency domain is closely related to the use of conjugate complex numbers. According to the properties of Fourier transform, the calculation of the cross-correlation function in the frequency domain is intrinsically related to the conjugate operation of the frequency domain matrix. This consistency enables the use of the conjugate complex numbers of the deformed image spectrum when calculating the cross-correlation function in the entire Fourier transform-based image processing process to better connect with other steps (such as phase difference calculation, displacement field solution, etc.), ensuring the logic and coherence of the algorithm.

[0079] If you use it directly Without using conjugate, the combination of phase information will be different during the calculation process, resulting in the cross-correlation function not being able to accurately represent the similarity between the two images at different frequencies. Using conjugate complex numbers can make the cross-correlation function reach its maximum value at a specific frequency under certain ideal conditions (such as no noise, small deformation, etc.), indicating that the two images have the highest similarity at this frequency, which is crucial for the subsequent accurate extraction of phase changes caused by deformation.

[0080] In this embodiment, the calculation formula of the first phase difference is as follows:

[0081] ;

[0082] in, represents the first phase difference, express The imaginary part of express The real part of .

[0083] Through the above formula, the first phase difference corresponding to each frequency domain variable can be calculated in the frequency domain: By calculating the phase difference in the frequency domain, the characteristics of Fourier transform can be used to better analyze the relationship between the periodic structure, texture and other information in the image and the displacement. The corresponding distribution in the spatial domain determines the second phase difference corresponding to the spatial domain .

[0084] This embodiment uses two-dimensional discrete Fourier transform to process image data, which can quickly and effectively extract useful information from image data, thereby improving the efficiency of data processing. Moreover, based on frequency domain analysis to calculate cross-correlation functions and phase differences, etc., compared with direct processing in the spatial domain, it can reduce the interference of factors such as noise, capture small displacement changes of rock mass, and thus more accurately reveal the differences and changes between images, improve the accuracy of analysis results, and provide a basis for subsequent accurate calculation of displacement fields.

[0085] Step 4: Obtain a phase-demodulated displacement field according to the second phase difference and based on a pre-fitted displacement function, and calculate a displacement gradient and a strain field according to the displacement field.

[0086] In a physical sense, in digital image correlation technology, image deformation will lead to relative displacement between pixels, and this displacement information is related to the phase difference obtained by phase demodulation. The phase difference reflects the change in optical path caused by deformation, and the change in optical path has a certain geometric relationship with the displacement of the object surface.

[0087] Assume that the object is The displacement in the direction, that is, the horizontal displacement, is ,exist The displacement in the direction, i.e. the vertical displacement, is , and there is a certain linear relationship between displacement and phase difference. Specifically, the phase difference of the image is caused by the displacement of the object surface, so it can be expressed as the following functional relationship:

[0088] ;

[0089] ;

[0090] in, represents the horizontal displacement, represents the vertical displacement, represents the second phase difference, , , and Represents the coefficient.

[0091] In practical applications, the least squares fitting method can be used to determine the coefficients , , and The goal of the least squares method is to minimize the error function so that the sum of squares of the errors between the displacement field calculated based on the phase difference and the displacement actually measured (indirectly reflected by the phase difference) is minimized. The error function can be defined as:

[0092] ;

[0093] in represents the sum of squared errors, and is the displacement estimate calculated from the second phase difference (the initial value may not be accurate), and We hope to find the displacement closest to the true value through the least squares method. Indicates The spatial coordinates of the pixels.

[0094] Will and Substitute into the error function E, and make , , and Find the partial derivative and set it equal to 0 to get the least squares equation system. By solving the equation system, we can get the coefficients , , and , and then the displacement field is obtained and , realizing the conversion from phase difference to displacement field and completing the phase demodulation process.

[0095] It can be understood that the displacement gradient describes the spatial distribution of displacement changes between points on the surface of an object, and represents the rate of change of displacement in space, that is, the degree of change of displacement between two adjacent points. In this embodiment, the displacement gradient includes a horizontal displacement gradient and a vertical displacement gradient, and the calculation formula is as follows:

[0096] ;

[0097] ;

[0098] in, represents the horizontal displacement gradient, represents the vertical displacement gradient, Indicates the horizontal displacement in The partial derivative in the direction, Indicates the vertical displacement in The partial derivative in the direction, Indicates the vertical displacement in The partial derivative in the direction, Indicates the horizontal displacement in Partial derivatives in the direction.

[0099] The horizontal displacement gradient and the vertical displacement gradient have different effects on the stability of the rock mass. When the horizontal displacement gradient is large, it may cause the rock mass to slip, crack, and other damage in the horizontal direction; when the vertical displacement gradient is too large, it may cause the rock mass to collapse, bulge, and other problems. This embodiment comprehensively considers the horizontal displacement gradient and the vertical displacement gradient, and can evaluate the damage risk of the rock mass from multiple angles, more comprehensively judge whether the rock mass is in a stable state, and further improve the accuracy of dangerous rock mass identification.

[0100] It can be understood that the strain field is an important derivation of the displacement gradient, reflecting the relative deformation of a certain area on the surface or in the body of an object. In this embodiment, the strain field includes horizontal strain, vertical strain and shear strain, and the calculation formula is as follows:

[0101] ;

[0102] ;

[0103] ;

[0104] in, Represents horizontal strain, i.e. horizontal displacement exist Changes in direction, Represents the vertical strain, i.e. the vertical displacement exist Changes in direction, represents the shear strain, i.e. the horizontal displacement and vertical displacement Respectively Direction and Cross-change of direction.

[0105] Different types of strain have different effects on the strength and stability of rock mass. Excessive horizontal and vertical strains may cause tensile or compressive failure of the rock mass, while excessive shear strain may cause shear failure. This embodiment takes into account the horizontal strain, vertical strain and shear strain, and can fully present the deformation state of the rock mass from multiple dimensions, and can more accurately evaluate the strength reserve and stability of the rock mass under different stress conditions, thereby further improving the accuracy of dangerous rock mass identification.

[0106] In practical applications, strain field characteristic parameters (such as local strain concentration area and strain gradient mutation point) can be extracted according to displacement field, displacement gradient and strain field to identify deformation characteristics of target rock mass, such as crack extension and local slip, etc. At the same time, DIC technology can also be used to obtain detailed information on rock surface deformation, including deformation distribution, size, direction, etc., to provide data support for subsequent risk assessment.

[0107] This embodiment processes the phase difference based on the pre-fitted displacement function to obtain the displacement field, makes full use of the known physical relationships and mathematical models, further improves the accuracy and reliability of the rock mass displacement analysis, can more realistically reflect the actual displacement of the rock mass, and lays a solid foundation for subsequent calculations and analysis.

[0108] Step 5: Determine the danger level of each area of ​​the target rock mass according to the displacement field, displacement gradient and strain field and based on a pre-trained dangerous rock mass identification model.

[0109] Specifically, the displacement field can intuitively reflect the position change of each point of the rock mass, the displacement gradient can reflect the rate of change of displacement at different positions, and the strain field can reveal the degree and distribution of deformation inside the rock mass. This embodiment combines the displacement field, displacement gradient and strain field, which can fully present the mechanical state and deformation characteristics of the target rock mass from different dimensions. The three types of data are interrelated and complementary. The multi-parameter comprehensive analysis using the dangerous rock mass identification model can avoid misjudgment or omission caused by single factor judgment, and greatly improve the accuracy and reliability of dangerous rock mass identification. In addition, in actual engineering, geological conditions are often very complex, and the rock mass properties and stress conditions in different regions may vary greatly. Using displacement field, displacement gradient and strain field for analysis can better adapt to this complexity, comprehensively consider the impact of various factors on the stability of the rock mass, and accurately identify dangerous rock masses of different types and sizes.

[0110] In this embodiment, the training method of the dangerous rock mass identification model includes: obtaining images of rock samples within two cycles and preprocessing them; determining the displacement field, displacement gradient and strain field of the rock samples based on the images of the preprocessed rock samples, and marking the danger level of each area of ​​the rock sample according to the deformation area, deformation rate and deformation gradient; using the displacement field, displacement gradient and strain field of the rock sample as input features, and the danger level of the corresponding area as an output label, to construct a training data set, and use the training data set to train the dangerous rock mass identification model.

[0111] In practical applications, we first need to preprocess the images of rock samples using the same preprocessing method as the target rock mass, then determine the displacement field, displacement gradient and strain field of the rock samples using the same method as the target rock mass image, and mark the hazard level of each area of ​​the rock samples. Finally, we use the displacement field, displacement gradient and strain field of the rock samples as input features, and the hazard level of the corresponding area as the output label to train the dangerous rock recognition model.

[0112] Understandably, if the deformation in some places is much larger than the surrounding areas, this may mean that the rock mass here is loosening or cracking. When the strain in some areas suddenly jumps from a small value to a large value, it means that there may be cracks or potential instability in these areas. Not only the size of the deformation, but also the speed of change, if a certain area is deforming faster and faster, it may mean that damage is about to happen.

[0113] Based on this, this embodiment marks the danger level of each area of ​​the rock sample according to the deformation area, deformation rate and deformation gradient, and divides each area on the rock surface into three danger levels: Stable: No abnormal deformation, the rock is basically stable. Transition: Small-scale deformation occurs, which may have potential risks. Dangerous: Significant deformation, high risk of damage.

[0114] After obtaining the displacement field, displacement gradient and strain field of the target rock mass, they are input into the trained dangerous rock mass identification model to obtain the danger level of each area of ​​the target rock mass.

[0115] In this embodiment, the method also includes: making corresponding graphic annotations in the rock model according to the danger level of each area of ​​the target rock mass; determining the area of ​​the region corresponding to each danger level, determining the warning level of the target rock mass according to the size and change rate of the area of ​​the region, and issuing a warning based on the warning level.

[0116] See also Figure 6In this embodiment, different colors are used to mark areas of different danger levels in the rock model, so that technicians can intuitively view the deformation of the target rock mass. In addition, this embodiment sets the warning level according to the danger level of each area of ​​the rock mass: Level 1 warning (high risk): The area of ​​the danger zone exceeds a certain proportion, and the strain change rate increases rapidly, and collapse may occur soon. Level 2 warning (medium risk): The scope of the transition zone is expanded, and the strain change rate is slow, but it still needs attention. Level 3 warning (low risk): There is only a small range of deformation or the change rate is extremely low, and there is no obvious danger for the time being. Corresponding warnings are issued according to the warning level, so that technicians can understand and deal with it in time. For example, in the case of a level 1 warning, immediately evacuate the surrounding people and take emergency protective measures (such as unloading dangerous rocks and reinforcing supports). In the case of a level 2 warning, monitor the dangerous area, strengthen patrols, and carry out local reinforcement when necessary. In the case of a level 3 warning, regular reviews and records are conducted to remain vigilant.

[0117] In summary, the large-scale dangerous rock mass identification method based on digital image deformation processing provided in this embodiment converts the two-dimensional grayscale matrix of the digital image in the spatial domain into a frequency domain matrix, calculates the cross-correlation function and phase difference of the two images before and after the target rock mass is deformed in the frequency domain, and then calculates the displacement field, displacement gradient and strain field. Since the complex convolution operation in the spatial domain is converted into a simple product operation in the frequency domain, the deformation position determination process of the traditional DIC is also avoided, and useful information can be quickly and effectively extracted from a large amount of image data, thereby simplifying the image processing process and improving the efficiency of dangerous rock mass identification. Moreover, the calculation of cross-correlation functions and phase differences based on frequency domain analysis can reduce the interference of factors such as noise, capture the small displacement changes of the rock mass, and more accurately reveal the differences and changes between images, improve the accuracy of the analysis results, and provide a basis for the subsequent accurate calculation of the displacement field. In addition, the present invention constructs a dangerous rock mass identification model and inputs the displacement field, displacement gradient and strain field of the target rock mass into the dangerous rock mass identification model to obtain the danger level of the target rock mass. It not only comprehensively describes the mechanical state of the rock mass from multiple dimensions, but also simplifies the dangerous rock mass identification process, and can accurately identify the slight deformation on the surface of the dangerous rock mass, further improving the efficiency and accuracy of dangerous rock mass identification.

Claims

1. A large-scale dangerous rock mass identification method based on digital image deformation processing is characterized by: The method comprises: Collect images of the target rock mass according to a preset period and pre-process the images; Obtain two images collected in two cycles after preprocessing, and generate two-dimensional grayscale matrices of the two images in the spatial domain respectively; Performing a two-dimensional discrete Fourier transform on the two-dimensional grayscale matrices of the two images respectively to obtain frequency domain matrices of the two images, calculating a cross-correlation function of the two images according to the frequency domain matrices, calculating a first phase difference corresponding to the frequency domain according to the cross-correlation function, and determining a second phase difference corresponding to the spatial domain according to the first phase difference; According to the second phase difference and based on the pre-fitted displacement function, a displacement field after phase demodulation is obtained, and a displacement gradient and a strain field are calculated according to the displacement field; The danger level of each area of ​​the target rock mass is determined according to the displacement field, displacement gradient and strain field and based on a pre-trained dangerous rock mass identification model.

2. The large-scale dangerous rock body identification method based on digital image deformation processing according to claim 1 is characterized in that: The calculation formula of the two-dimensional discrete Fourier transform is as follows: ； in, express The frequency domain matrix, Represents the two-dimensional grayscale matrix of the image in the spatial domain, Represents the spatial coordinates of the pixel point, represents the frequency domain coordinates, , , Indicates the number of pixels in the horizontal direction of the image. Indicates the number of pixels in the vertical direction of the image. represents a natural constant, Represents an imaginary unit.

3. The large-scale dangerous rock body identification method based on digital image deformation processing according to claim 2 is characterized in that: The calculation formula of the cross-correlation function is as follows: ； in, represents the cross-correlation function, Represents the frequency domain matrix corresponding to the image collected in the first cycle, Represents the conjugate complex matrix of the frequency domain matrix corresponding to the image collected in the second period; The calculation formula of the first phase difference is as follows: ； in, represents the first phase difference, express The imaginary part of express The real part of .

4. The large-scale dangerous rock body identification method based on digital image deformation processing according to claim 3 is characterized in that: The displacement field includes the horizontal displacement and vertical displacement of each pixel point, and the calculation formula is as follows: ； ； in, represents the horizontal displacement, represents the vertical displacement, represents the second phase difference, , , and Represents the coefficient.

5. The large-scale dangerous rock body identification method based on digital image deformation processing according to claim 4 is characterized in that: The displacement gradient includes horizontal displacement gradient and vertical displacement gradient, and the calculation formula is as follows: ； ； in, represents the horizontal displacement gradient, represents the vertical displacement gradient, Indicates the horizontal displacement in The partial derivative in the direction, Indicates the vertical displacement in The partial derivative in the direction, Indicates the vertical displacement in The partial derivative in the direction, Indicates the horizontal displacement in Partial derivatives in the direction.

6. The large-scale dangerous rock body identification method based on digital image deformation processing according to claim 5 is characterized in that: The strain field includes horizontal strain, vertical strain and shear strain, and the calculation formula is as follows: ； ； ； in, represents the horizontal strain, represents the vertical strain, represents shear strain.

7. The large-scale dangerous rock body identification method based on digital image deformation processing according to claim 1 is characterized in that: The method further comprises: Before acquiring an image of the target rock mass, a speckle material is coated on the surface of the target rock mass to form a uniform speckle pattern on the surface of the target rock mass.

8. The large-scale dangerous rock body identification method based on digital image deformation processing according to claim 1 is characterized in that: Preprocess the image, including: noise removal, contrast enhancement, image registration and image cropping; The noise removal method includes high-pass filtering, mean filtering and Gaussian filtering, the contrast enhancement method includes adaptive histogram equalization and Gamma correction, and the image registration method includes feature point matching, phase correlation method and cross-correlation method.

9. The large-scale dangerous rock body identification method based on digital image deformation processing according to claim 1 is characterized in that: The training method of the dangerous rock mass identification model includes: Obtain images of rock samples in two cycles and preprocess them; Determine the displacement field, displacement gradient and strain field of the rock sample based on the preprocessed image, and mark the hazard level of each area of ​​the rock sample based on the deformation area, deformation rate and deformation gradient; The displacement field, displacement gradient and strain field of the rock sample are taken as input features, and the danger level of the corresponding area is taken as the output label to construct a training data set, and the training data set is used to train a dangerous rock mass identification model.

10. The large-scale dangerous rock body identification method based on digital image deformation processing according to claim 1 is characterized in that: The method further comprises: According to the danger level of each area of ​​the target rock mass, corresponding graphic annotations are made in the rock mass model; Determine the area corresponding to each danger level, determine the warning level of the target rock mass according to the size and change rate of the area, and issue a warning according to the warning level.

Images