Digital speckle-based solidified soil damage mode identification method and system

Abstract

The invention discloses a solidified soil damage mode recognition method and system based on digital speckles, and the method comprises the steps: obtaining a left group of speckle image series and a right group of speckle image series of measured solidified soil through a calibrated camera, and obtaining two groups of preprocessed image series; decomposing the reference image of the first group of preprocessed image series into first reference sub-region images through a sub-region matching algorithm; performing stereo matching on the second group of preprocessed image series based on a feature algorithm of scale invariant feature transformation to obtain a second reference subarea image; and performing time sequence matching on the first reference subarea image and the second reference subarea image to obtain an initial deformation displacement value of the solidified soil so as to confirm an identification damage mode of the measured solidified soil according to the displacement deformation value of the solidified soil. The method solves the problem that a traditional determination method of a traditional solidified soil damage mode mainly depends on analysis of a damage form and a stress-strain curve of cement solidified soil, but the analysis is only on a qualitative level, and the damage mode of the solidified soil is difficult to accurately identify.

Description

Technical Field

[0001] The present invention relates to the field of optical measurement technology, and in particular to a method and system for identifying solidified soil failure patterns based on digital speckle. Background Art

[0002] As my country's dredging industry expands, the resource utilization of the ever-increasing amount of dredged soil is a critical issue that needs to be addressed. One such resource utilization option is to convert dredged soil into relatively stable solidified soil through solidification treatment, which can be used as an alternative material for roadbed, foundation pit, and slope reinforcement. However, while solidified soil improves engineering performance and reduces costs, it also faces multiple risks, including structural degradation, reduced durability, and quality issues. Therefore, establishing a maintenance and deformation monitoring system for solidified soil is crucial to ensuring its long-term performance.

[0003] Technical problem solved: The traditional method for identifying the failure mode of cement-solidified soil mainly relies on the analysis of the failure morphology and stress-strain curve of cement-solidified soil. However, this is only a qualitative analysis and it is difficult to accurately identify the failure mode of cement-solidified soil. Therefore, there is an urgent need for a method that combines digital speckle technology to achieve quantitative characterization of the surface deformation field of cement-solidified soil. A method is proposed to accurately identify the failure mode of cement-solidified soil by analyzing the cracking coefficient index of cement-solidified soil cracks. Summary of the Invention

[0004] The present invention provides a method and system for identifying failure patterns of solidified soil based on digital speckle to overcome the above technical problems.

[0005] In order to achieve the above object, the technical solution of the present invention is:

[0006] A method for identifying failure patterns of stabilized soil based on digital speckle patterns specifically comprises the following steps:

[0007] S1: Obtain two sets of left and right speckle image series of the solidified soil under test through the calibrated camera;

[0008] The two sets of speckle image series are subjected to grayscale conversion, interpolation and denoising respectively to obtain two sets of pre-processed image series;

[0009] The left pre-processed image series is defined as the first set of pre-processed image series;

[0010] The right pre-processed image series is defined as the second set of pre-processed image series;

[0011] S2: decomposing the reference image of the first set of pre-processed image series into first reference sub-region images using a preset sub-region matching algorithm;

[0012] S3: Stereo matching is performed on the second set of pre-processed image series based on a feature algorithm based on scale-invariant feature transformation to obtain a second reference sub-area image, specifically comprising the following steps:

[0013] S31: constructing multiple groups of Gaussian pyramids for scale space division according to the second group of preprocessed image series;

[0014] S32: Obtaining Gaussian blurred images of different scales in each layer structure of the multiple groups of Gaussian pyramids;

[0015] S33: subtracting the Gaussian blurred images of different scales in adjacent layers within each group of the multiple Gaussian pyramids to obtain a difference image, thereby obtaining a Gaussian difference pyramid, namely, a DOG pyramid;

[0016] Obtaining differential image feature points according to the DOG pyramid; and the differential image feature points are local pixel extreme points of the differential image obtained by the DOG pyramid space;

[0017] and obtaining pixel fitting parameters for confirming the deformation of the solidified soil based on the local pixel extreme points, and determining a second reference sub-area image in the reference image of the second set of speckle image series according to the pixel fitting parameters;

[0018] S4: performing time-series matching on the first reference sub-area image and the second reference sub-area image to obtain the initial deformation displacement value of the solidified soil, so as to confirm the identified failure mode of the solidified soil under test according to the displacement deformation value of the solidified soil.

[0019] Furthermore, the method for constructing multiple groups of Gaussian pyramids for scale space division in S31 specifically includes the following steps:

[0020] S311: obtaining the number of groups of Gaussian pyramids according to the second pre-processed image series, and encoding the number of groups of Gaussian pyramids to obtain a group sequence table of multiple Gaussian pyramids;

[0021] And the formula for obtaining the number of Gaussian pyramid groups is:

[0022] O=[log2(min(W,H))]-t (1)

[0023] Where: O represents the number of groups of Gaussian pyramids; W, H represent the width and height of the second group of preprocessed images respectively; t represents the image size of the minimum resolution image;

[0024] S312: Obtain multi-layer image data of any Gaussian pyramid group Q in a list of multiple Gaussian pyramid groups;

[0025] The method for acquiring multi-layer image data includes:

[0026] S3121: Define the total number of layers of the Gaussian pyramid group Q to be L, and the initial image set of each layer is an empty set;

[0027] After doubling the size of the images in the second pre-processed image series in the Qth group, the images are used as the image data of the first layer of the Qth group of multiple Gaussian pyramids;

[0028] The images in the first layer of the Qth group of multiple Gaussian pyramids are subjected to Gaussian convolution as the image data of the second layer of the Qth group;

[0029] And the function expression of Gaussian convolution is

[0030]

[0031] Where: G(x,y) represents the Gaussian function; σ represents the scale space factor in the SIFT operator; x,y represent the position points of the image pixels in the second set of preprocessed image series; x0,y0 represent the position points of the image pixels after expansion in the second set of preprocessed image series;

[0032] S3122: Update the scale space factor to k*σ to obtain a new Gaussian convolution function;

[0033] The image of the second layer of the Q group is subjected to Gaussian convolution by the new Gaussian convolution function and used as the image data of the third layer of the Q group;

[0034] S3123: Update the scale space factor to K again 2 *σ, to obtain the updated Gaussian convolution function;

[0035] The image data of the third layer of the Q group is subjected to the Gaussian convolution function that is updated again, and the resulting image data is used as the image data of the fourth layer of the Q group; ......

[0037] Update and obtain the scale space factor K of the Gaussian convolution function in the Lth layer of the Qth group (L-1) σ, to obtain the Gaussian convolution function of the Lth layer of the Qth group;

[0038] The image of the L-1th layer of the Qth group is subjected to the Gaussian convolution function of the Lth layer of the Qth group, and the resulting image is used as the image data of the Lth layer of the Qth group;

[0039] S313: Repeat step S312 to obtain O groups of Gaussian pyramids, each group having L layers of image data, and a total of O*L images to achieve the construction of multiple groups of Gaussian pyramids.

[0040] Furthermore, the method for obtaining multiple groups of Gaussian blurred images of different scales in each layer structure of the Gaussian pyramid in S32 includes:

[0041] S321: Obtain the total number of layers S of each group in multiple Gaussian pyramids;

[0042] Calculate and obtain the image Gaussian blur coefficient of each layer in the same group;

[0043] And the formula for obtaining the image Gaussian blur coefficient is:

[0044]

[0045] Where: σ(O,S) represents the Gaussian blur coefficient of the image; σ0 represents the initial scale of the image, that is, the initial value of the Gaussian blur; S i Indicates the layer number code; S indicates the total number of layers in each group;

[0046] S322: Based on the image Gaussian blur coefficient, obtain the image Gaussian blur coefficient of each group in the multiple groups of Gaussian pyramids to obtain Gaussian blurred images of different scales in each layer of the multiple groups of Gaussian pyramids.

[0047] Furthermore, the step S33 specifically includes the following steps:

[0048] S331: Subtracting Gaussian blurred images of different scales in adjacent layers within each of the multiple Gaussian pyramids to obtain a difference image, and then obtaining all the difference images in the multiple Gaussian pyramids group by group and layer by layer;

[0049] And build a Gaussian difference pyramid, namely DOG pyramid, based on all difference images;

[0050] S332: Obtaining differential image feature points according to the DOG pyramid; and the differential image feature points are local pixel extreme points of the differential image obtained from the DOG pyramid space;

[0051] And the formula for obtaining the local pixel extreme point is:

[0052]

[0053] Where: represents the local pixel extreme point of the differential image; D represents the DOG function about the feature point of the differential image; X represents the feature point of the differential image; σ represents the abbreviation of σ(O,S);

[0054] S333: Obtain pixel fitting parameters for confirming the deformation of solidified soil based on local pixel extreme points; and the pixel fitting parameter acquisition formula is:

[0055]

[0056] Where: represents the pixel fitting parameter used to confirm the deformation of solidified soil; T represents transposition;

[0057] The estimated positions of the differential image feature points in the reference image of the second set of speckle image series are determined according to the pixel fitting parameters, and the positions of the corresponding reference sub-areas in the reference image are confirmed according to the estimated positions based on the SURF feature matching algorithm.

[0058] Furthermore, the S4 specifically includes the following steps:

[0059] S41: Based on the phase correlation method, the first reference sub-area image and the second reference sub-area image are time-series matched to obtain two sets of initial deformation displacement values ​​of the solidified soil;

[0060] S42: using the inverse synthetic Gauss-Newton iteration method, solving and obtaining the zero-mean normalized minimum distance square criterion based on the two sets of initial deformation displacement values, and iteratively optimizing the zero-mean normalized minimum distance square criterion to obtain two sets of estimated displacements;

[0061] S43: performing displacement field calculation on the two groups of estimated displacements to convert the image displacement into actual displacement;

[0062] S44: determining the actual strain value of the solidified soil under test according to the actual displacement;

[0063] The cracking coefficient of the cracks after the solidified soil is destroyed is obtained based on the actual strain value, and the cracking coefficient is used to characterize the cracking situation of the solidified soil and identify the failure mode.

[0064] Furthermore, the formula for obtaining the cracking coefficient of the crack after the solidified soil is damaged in S44 is:

[0065]

[0066] Where: λ represents the cracking coefficient of the crack after the solidified soil is damaged; ε xmax Indicates the transverse strain value at the maximum crack; The average transverse strain value of the neighborhood around the maximum crack is the actual strain value of the tested solidified soil.

[0067] A system based on a digital speckle-based solidified soil failure pattern recognition method comprises an image acquisition module, an image preprocessing module, a reference sub-area image acquisition module, an image data processing module, and a solidified soil failure pattern recognition module;

[0068] An image acquisition module is used to acquire two sets of speckle image series of the solidified soil under test;

[0069] An image preprocessing module is used to perform grayscale conversion, interpolation, and denoising on the two sets of speckle image series in sequence to obtain two sets of preprocessed image series;

[0070] a reference sub-region image acquisition module, configured to decompose the reference images of the first set of pre-processed image series into first reference sub-region images using a preset sub-region matching algorithm; and simultaneously perform stereo matching on the second set of pre-processed image series based on a feature algorithm of scale-invariant feature transformation to obtain second reference sub-region images;

[0071] An image data processing module is used to perform time-series matching on the first reference sub-area image and the second reference sub-area image based on a phase correlation method to obtain two sets of initial deformation and displacement values ​​of the solidified soil;

[0072] The inverse synthetic Gauss-Newton iteration method is used to solve and obtain the zero-mean normalized minimum distance square standard based on the two sets of initial deformation displacement values, and the zero-mean normalized minimum distance square standard is iteratively optimized to obtain two sets of estimated displacements;

[0073] The displacement fields of the two sets of estimated displacements are calculated to convert the image displacements into actual displacements;

[0074] The solidified soil failure mode recognition module is used to determine the actual strain value of the solidified soil under test based on the actual displacement;

[0075] The cracking coefficient of the cracks after the solidified soil is destroyed is obtained based on the actual strain value, and the cracking coefficient is used to characterize the cracking situation of the solidified soil and identify the failure mode.

[0076] Beneficial effects: The present invention provides a method and system for identifying failure patterns of solidified soil based on digital speckle patterns. Stereo matching is performed on a second set of preprocessed image series using a feature algorithm based on scale-invariant feature transformation to obtain a second reference sub-area image. Time-series matching is performed on the first reference sub-area image and the second reference sub-area image to obtain the initial deformation displacement value of the solidified soil. The identified failure pattern of the solidified soil under test is confirmed based on the displacement deformation value of the solidified soil. The present invention obtains the deformation result of the solidified soil after failure through high-precision detection of digital speckle patterns, and quantitatively analyzes the failure pattern of the solidified soil to achieve more accurate identification, providing a solid technical guarantee for the safety management and life prediction of engineering structures. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following is a brief introduction to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0078] Figure 1 This is a flow chart of the method for identifying failure patterns of solidified soil based on digital speckle according to the present invention;

[0079] Figure 2 Schematic diagram of finding extreme points in the Gaussian difference pyramid in the present invention;

[0080] Figure 3 Schematic diagram of the target image area before and after deformation in the present invention;

[0081] Figure 4 Schematic diagram of the overall process of IC-GN in the present invention;

[0082] Figure 5 Schematic diagram of cracking of solidified soil at time 3 / 5T0 in the present invention;

[0083] Figure 6 The failure morphology diagram of soil solidified by ordinary Portland cement and silica fume modified cement with different cement content in the present invention;

[0084] Figure 7 The stress-strain curves of ordinary Portland cement-cured soil with different cement content of OPC-3d in the present invention;

[0085] Figure 8 The stress-strain curves of ordinary Portland cement-cured soil with different cement content of OPC-28d in the present invention are shown;

[0086] Figure 9 Schematic diagram of selecting the analysis area in the present invention;

[0087] Figure 10 This is a simulation diagram of displacement results of failure mode identification of solidified soil in the present invention;

[0088] Figure 11 This is a simulation diagram of strain results for failure mode identification of solidified soil in the present invention;

[0089] Figure 12 It is a box plot of the lateral strain in the neighborhood of the maximum crack of ordinary Portland cement-stabilized soil in the present invention. DETAILED DESCRIPTION

[0090] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0091] This embodiment provides a method for identifying failure patterns of solidified soil based on digital speckle. Figure 1 As shown, the specific steps include:

[0092] S1: Obtain two sets of left and right speckle image series of the solidified soil under test through the calibrated camera;

[0093] Specifically, this embodiment can obtain the internal and external parameters of the camera by calibrating the camera. The internal parameters include focal length, principal point position, pixel size, etc.; the external parameters include the position and posture of the camera. These parameters can accurately convert the collected image information into the spatial dimensions in the physical coordinate system. At the same time, calibration can perform distortion correction. All camera lenses have a certain amount of lens distortion. Through calibration, the distortion parameters can be obtained, and then the lens distortion in the image can be corrected to obtain a true representation of the shape and size of the measured object. Different interpolation methods are used for different data situations, and denoising through Gaussian filtering can obtain more accurate displacement information, which is better for accuracy and convergence characteristics.

[0094] Grayscale conversion, interpolation, and denoising are performed on the two sets of speckle image series, respectively, to obtain two sets of pre-processed image series. The left pre-processed image series is defined as the first set of pre-processed image series; and the right pre-processed image series is defined as the second set of pre-processed image series. The grayscale conversion, interpolation, and denoising methods are all known in the art and will not be elaborated on here.

[0095] For example, interpolation and Gaussian filtering are performed on the collected images. Because the image of the object to be measured is composed of grayscale values ​​of different sizes, the correlation analysis is performed based on the minimum unit value of the pixel, and the minimum unit is the size of the entire pixel. However, the actual situation is not like this. Due to the deformation requirements of the object to be measured and the measurement accuracy requirements, the above situation where a single pixel is the minimum unit is difficult to meet the above requirements; therefore, this embodiment further divides the pixels and performs a difference on the grayscale values, so that more accurate displacement information can be obtained, and the alignment accuracy and convergence characteristics are better. This embodiment provides four interpolation methods, corresponding to different situations, linear interpolation: the linear interpolation method is simple and has a fast calculation speed, and is suitable for situations where data changes relatively slowly; Spline interpolation: the cubic spline interpolation method uses a cubic polynomial to approximate the data, and is suitable for situations where smooth curve approximation is required; Makima interpolation: the Modified Akima interpolation method has good adaptability in places where the data curvature changes greatly, and is suitable for situations where better adaptation to changes in data curvature is required; Pchip interpolation: the piecewise cubic Hermite interpolation method ensures a smooth interpolation curve, and is suitable for situations where the curve changes greatly between data points; Gaussian filtering is a method for removing image noise while maintaining the edge information of the image; in DIC analysis, since this embodiment aims to extract accurate shape, displacement, and deformation information from the image, it is necessary to minimize the influence of noise as much as possible. Gaussian filtering can effectively remove random noise in the image, thereby improving the accuracy of the measurement;

[0096] S2: Decomposing the reference image of the first set of preprocessed image series into first reference sub-region images by using a preset sub-region matching algorithm, so as to be used for subsequent stereo matching and time series matching, thereby finding the sub-region of the image after deformation of the solidified soil;

[0097] Specifically, the method of obtaining the first reference sub-region image by using a preset sub-region matching algorithm is a conventionally known technical means. For example, by selecting the sub-region size (unit pixel), the sub-region step size (pixel unit), and the sub-region shape (square or circular), where the sub-region step size represents the interval between two adjacent sub-regions, and after the selection is made, the entire detection area is divided into sub-regions according to the sub-region size and step size through grid division, thereby obtaining the decomposed first reference sub-region image. The specific implementation process will not be elaborated on here.

[0098] S3: Perform stereo matching on the second set of pre-processed image series based on a feature algorithm of scale-invariant feature transformation to obtain a second reference sub-area image; specifically, the steps include:

[0099] S31: constructing multiple groups of Gaussian pyramids for scale space partitioning based on the second group of preprocessed image series; wherein the multiple groups of Gaussian pyramids include several network layers obtained based on the second group of preprocessed image series;

[0100] In a specific embodiment, the method for constructing multiple groups of Gaussian pyramids for scale space division in S31 specifically includes the following steps:

[0101] S311: obtaining the number of groups of Gaussian pyramids according to the second pre-processed image series, and encoding the number of groups of Gaussian pyramids to obtain a group sequence table of multiple Gaussian pyramids;

[0102] And the formula for obtaining the number of Gaussian pyramid groups is:

[0103] O=[log2(min(W,H))]-t (1)

[0104] Where: O represents the number of groups of Gaussian pyramids; W, H represent the width and height of the image respectively; t represents the image size of the minimum resolution image, which refers to the standard deviation and can also be the Gaussian blur coefficient. For example, 3x3 is set as the minimum size, and images of the same size are considered as one group.

[0105] S312: Obtain multi-layer image data of any one of the multiple Gaussian pyramid groups Q in the multiple Gaussian pyramid group sequence list;

[0106] The method for acquiring multi-layer image data includes:

[0107] S3121: Define the total number of layers of the Gaussian pyramid group Q as L, and the initial image set of each layer is an empty set; double the size of the images in the second pre-processed image series in the Qth group and use them as the image data of the first layer of the Qth group of the multiple Gaussian pyramids;

[0108] The image in the first layer of the Qth group is subjected to Gaussian convolution (e.g., Gaussian smoothing or Gaussian filtering) as the image data of the second layer of the Qth group;

[0109] And the function expression of Gaussian convolution is

[0110]

[0111] Where: G(x,y) represents the Gaussian function; σ represents the scale space factor in the SIFT operator, preferably with a fixed value of 1.6; x,y represent the position points of the image pixels in the second set of preprocessed image series; x0,y0 represent the position points of the image pixels after expansion in the second set of preprocessed image series;

[0112] S3122: Update the scale space factor to k*σ to obtain a new Gaussian convolution function;

[0113] The image of the second layer of the Q group is subjected to Gaussian convolution by the new Gaussian convolution function and used as the image data of the third layer of the Q group;

[0114] For example, the original image is first doubled and then used as the first layer of the first group of multiple Gaussian pyramids. The first layer of the first group of images is subjected to Gaussian convolution and used as the second layer of the first group of pyramids. In this embodiment, a new smoothing factor σ = k * σ is obtained by multiplying σ by a scaling factor k, and the new smoothing factor σ = k * σ is used to smooth the second layer of the first group of images. The resulting image is used as the third layer of the first group.

[0115] S3123: Update the scale space factor to K again 2 *σ, to obtain the updated Gaussian convolution function;

[0116] The image data of the third layer of the Q group is subjected to the Gaussian convolution function that is updated again, and the resulting image data is used as the image data of the fourth layer of the Q group; ......

[0118] Update and obtain the scale space factor K of the Gaussian convolution function in the Lth layer of the Qth group (L-1) σ, to obtain the Gaussian convolution function of the Lth layer of the Qth group;

[0119] The image of the L-1th layer of the Qth group is subjected to the Gaussian convolution function of the Lth layer of the Qth group, and the resulting image is used as the image data of the Lth layer of the Qth group;

[0120] Specifically, step S2122 is repeated in this way to finally obtain L layers of images. In the same group, the size of each layer of images is the same, but the smoothing coefficients are different, and the smoothing coefficients corresponding to the group are: 0, σ, Kσ, K 2 σ, K 3 σ……K (L-1) σ;

[0121] S313: Repeat step S312 to obtain O groups of Gaussian pyramids, each group including L layers of image data, totaling O*L images to achieve the construction of multiple groups of Gaussian pyramids;

[0122] Specifically, the third-to-last layer image of the first group is downsampled with a scale factor of 2, and the obtained image is used as the first layer of the second group. Then, the first layer image of the second group is Gaussian smoothed with a smoothing factor of σ to obtain the second layer of the second group, just like in step 2. In this way, the L layer image of the second group is obtained. The sizes of the images in the same group in this embodiment are the same, and the corresponding smoothing coefficients are: 0, σ, Kσ, K 2 σ, K 3 σ……K (L-1) σ, but the size of the second group is half of the first group of images; this is repeated to obtain a total of O groups, each with L layers, for a total of O*L images, which together constitute multiple groups of Gaussian pyramids;

[0123] S32: Obtain multiple sets of Gaussian blurred images of different scales in each layer of the Gaussian pyramid;

[0124] In a specific embodiment, the method for obtaining multiple groups of Gaussian blurred images of different scales in each layer of the Gaussian pyramid in S32 includes:

[0125] S321: Obtain the total number of layers S of each group in multiple Gaussian pyramids;

[0126] And the number of layers of each group in multiple Gaussian pyramids is expressed as S=n+t;

[0127] Where n represents the number of images to be extracted; S represents the number of layers of multiple Gaussian pyramids; t is the image size of the minimum resolution image;

[0128] Calculate and obtain the image Gaussian blur coefficient of each layer in the same group;

[0129] And the formula for obtaining the image Gaussian blur coefficient is:

[0130]

[0131] Where: σ(O,S) represents the Gaussian blur coefficient of the image; σ0 represents the initial scale of the image, that is, the initial value of the Gaussian blur; S iIndicates the layer number code; S indicates the total number of layers in each group;

[0132] For example: In this embodiment, for the convenience of calculation, the number of groups or layers is recorded starting from 0;

[0133] The Gaussian blur coefficient in the corresponding image pyramid can be calculated by formula (3), as follows:

[0134] Group 0, Layer 0:

[0135] Group 0, Layer 1:

[0136] Group 0, Layer 2: ......

[0138] Group 1, Layer 0:

[0139] Group 1, Layer 1:

[0140] Group 1, Layer 2: ......

[0142] S322: Based on the image Gaussian blur coefficient, obtain the image Gaussian blur coefficient of each group in the multiple groups of Gaussian pyramids to obtain Gaussian blurred images of different scales in each layer of the multiple groups of Gaussian pyramids;

[0143] S33: Subtracting Gaussian blurred images of different scales in adjacent layers within each group of multiple Gaussian pyramids to obtain a difference image, thereby obtaining a Gaussian difference pyramid, namely, a DOG pyramid;

[0144] Obtaining differential image feature points according to the DOG pyramid; and the differential image feature points are local pixel extreme points of the differential image obtained by the DOG pyramid space;

[0145] and obtaining pixel fitting parameters for confirming the deformation of the solidified soil based on the local pixel extreme points, and determining a second reference sub-area image in the reference image of the second set of speckle image series according to the pixel fitting parameters;

[0146] The specific steps include:

[0147] S331: Subtracting Gaussian blurred images of different scales in adjacent layers within each of the multiple Gaussian pyramids to obtain a difference image, and then obtaining all the difference images in the multiple Gaussian pyramids group by group and layer by layer;

[0148] And build a Gaussian difference pyramid, namely DOG pyramid, based on all difference images;

[0149] Specifically, after creating multiple groups of Gaussian pyramids for the image, adjacent layers within each group are subtracted to obtain a Gaussian difference pyramid (DOG), which is the prerequisite for later detection of image pixel extreme points; the first layer of the first group of the DOG pyramid is obtained by subtracting the first layer of the first group of the multiple groups of Gaussian pyramids from the second layer of the first group, and so on, generating each difference image group by group and layer by layer, and all difference images constitute a difference pyramid;

[0150] S332: Obtaining differential image feature points according to the DOG pyramid; and the differential image feature points are local pixel extreme points of the differential image obtained from the DOG pyramid space;

[0151] Specifically, the differential image feature points are composed of local extreme points in the DOG space. In order to find the extreme points of the DOG function, each pixel point is compared with all its neighboring points to see whether it is larger or smaller than its neighboring points in the image domain and scale domain.

[0152] like Figure 2 As shown in the figure: When searching for extreme points in the Gaussian difference pyramid, in addition to considering points in the x and y directions, points in the σ direction must also be considered. Therefore, to determine whether a pixel is an extreme point, it is necessary to compare it with the 26 surrounding points.

[0153] Note: ① If each group of the Gaussian difference pyramid has 3 layers, the extreme points can only be found in the middle layer of images. The images at both ends are discontinuous and have no extreme points.

[0154] ② If the Gaussian difference pyramid has 5 layers per group, then the extreme points can only be found in the middle 3 layers of images; and so on...

[0155] In this embodiment, the points in the image are discrete points, and the real feature point may not be at a certain pixel position in the image. In this embodiment, the optimal extreme point is found by performing curve fitting on the DOG image to correct the values ​​of the extreme point x, y, and σ. That is, Taylor expansion is performed on the image at the extreme point, and then the derivative is taken. The place where the derivative is equal to 0 is the real extreme point.

[0156] And the formula for obtaining the local pixel extreme point is:

[0157]

[0158] Where: represents the local pixel extreme point of the differential image; D represents the DOG function about the feature point of the differential image; X represents the feature point of the differential image; σ represents the abbreviation of σ(O,S);

[0159] S333: Obtain pixel fitting parameters for confirming the deformation of solidified soil based on local pixel extreme points; and the pixel fitting parameter acquisition formula is:

[0160]

[0161] Where: represents the pixel fitting parameter used to confirm the deformation of solidified soil; T represents transposition;

[0162] Determining estimated positions of differential image feature points in the reference image of the second set of speckle image series based on the pixel fitting parameters, and confirming positions of corresponding reference sub-areas in the reference image based on the estimated positions using a SURF feature matching algorithm, thereby obtaining a second reference sub-area image; and the method of confirming positions of corresponding reference sub-areas in the reference image based on the estimated positions using the SURF feature matching algorithm is a well-known technical means and will not be described in detail herein.

[0163] S4: performing time-series matching on the first reference sub-area image and the second reference sub-area image to obtain the initial deformation displacement value of the solidified soil, so as to confirm the identified failure mode of the solidified soil under test according to the displacement deformation value of the solidified soil, specifically comprising the following steps:

[0164] S41: Based on the phase correlation method, the first reference sub-area image and the second reference sub-area image are time-series matched to obtain two sets of initial deformation displacement values ​​of the solidified soil;

[0165] In a specific embodiment, the method for obtaining two sets of initial deformation and displacement values ​​of the solidified soil is:

[0166] S411: Define f1(x,y) and f2(x,y) to represent two image signals of the first reference sub-area image and the second reference sub-area image, respectively, and f2(x,y) can be obtained by shifting f1(x,y) by (x0,y0) pixels.

[0167] Perform Fourier transform on the two speckle images before and after deformation of the speckle image series, and the expression is:

[0168] F1(u,v)=F2(u,v)exp[-j2π(ux0+vy0)] (6)

[0169] Where F1(u,v) is the Fourier transform of the speckle image f1(x,y) before deformation; F2(u,v) is the Fourier transform of the speckle image f2(x,y) after deformation; (x0,y0) is the relative translation of the pixel point from f1(x,y) to f2(x,y); u,v are the displacements of the pixel data point position (x0,y0) in the reference sub-area image;

[0170] S412: Obtain the normalized cross-power spectrum phase of the image according to formula (6), which is expressed as follows:

[0171]

[0172] Where: Conjugated with F2(u,v);

[0173] Perform inverse Fourier transform on the cross power spectrum phase to solve and obtain the specific value of the relative translation (x0, y0), which is expressed as

[0174] δ(x-x0,y-y0)=IFFT(exp[j2π(ux0+vy0)]) (8)

[0175] Where: IFFT represents the inverse Fourier transform; δ represents the impulse function used to obtain the initial deformation displacement value; in this embodiment, the impulse function has a value only at zero point, and all other points are zero. That is, the function has a value only when the value of (x, y) is (x0, y0), which is also the maximum value of the impulse function. It can be seen from equation (6) that the inverse Fourier transform of the cross-power spectrum is exactly the impulse function of the translation point (x0, y0). From this, the estimated value of the image motion can be obtained, that is, the initial deformation value. Then, the initial deformation parameter value is continuously iteratively optimized for the relevant standards, and finally the optimal parameter solution is obtained, which can determine the displacement experienced by the sub-region through its respective image series.

[0176] S42: using the inverse synthetic Gauss-Newton iteration method, solving and obtaining the zero-mean normalized minimum distance square criterion based on the two sets of initial deformation displacement values, and iteratively optimizing the zero-mean normalized minimum distance square criterion to obtain two sets of estimated displacements;

[0177] In a specific embodiment, Figure 4 As shown, the method for obtaining the two sets of estimated displacements is

[0178] Let the reference image be F, representing the speckle pattern captured at time t=0, and the reference sub-area on it be f, let the deformed image be G, and the deformed sub-area on it be g, as shown in the following example: Figure 3 As shown in the figure, the rectangular box in the upper left corner is the reference sub-area, and the parallelogram in the lower right corner is the deformation sub-area, where the center point of the reference sub-area f is x 0 =[x0,x0] T , other pixels on the reference sub-area are x i =[x i ,y i ] T , taking the center point x0 as a reference, the expressions of other pixel points are obtained as

[0179]

[0180] where Δx i x 0 Represents x 0 to x i Coordinate offset of

[0181] Set the deformation sub-area with x i The corresponding coordinate point is x i′ =[x′ i ,x′ i ] T , the center point of the deformed sub-area corresponding to the previous center point x0 is x d , and the expressions for other pixels are:

[0182]

[0183] where Δx i′ x 0 Represents x 0 to x i′ Coordinate offset, Δx i′ represents the displacement in the x direction, Δx i′ represents the displacement in the y direction;

[0184] Calculate Δx i′ x 0 , the result is the pixel coordinates that correspond one-to-one with the previous sub-area after deformation. To solve it, we need to introduce a concept in finite element analysis - shape function. For a region or a unit, its deformation can be represented by a mathematical model, and this mathematical model used to describe it is the shape function (the common affine transformation in image processing is actually a shape function). The higher the order of the shape function, the more complex the deformation described. Let the shape function relationship be W and the shape function parameter be P. The coordinate displacement represented by the shape function is expressed as

[0185] Δx i x 0 =W(Δx i x 0 ,P) (11)

[0186] The digital image itself is a sampling of light intensity, that is, it is a discrete point. Considering the matching accuracy, the deformed image is interpolated to achieve sub-pixel matching. After interpolation, the reference sub-area and the deformed sub-area are written in the form of a function, which is expressed as

[0187] f i =F(x 0 +Δx i x 0 ) (12)

[0188] g i =G(x 0 +W(Δx i x 0 ,P)) (13)

[0189] Establish shape functions and select shape function relationships of appropriate orders according to actual conditions. The most commonly used shape functions are:

[0190] 0th-order shape function: can only describe displacement changes, not stretching or twisting. Its expression is

[0191]

[0192] 1st order shape function: can describe the displacement and stretching of the subregion (affine transformation), but cannot describe the distortion. Its expression is

[0193]

[0194] 2nd order shape function: can describe the displacement, stretching and distortion of the sub-region, and its expression is

[0195]

[0196] The parameters that need to be solved for shape functions of different orders are expressed as follows:

[0197] P SF0 =[uv] T (17)

[0198] P SF1 =[uu x u y vv x v y ] T (18)

[0199] P SF2 =[uu x u y u xx u xy u yy vv x v y v xx v xy v yy ] T (19)

[0200] A mathematical model is constructed based on the shape function and the zero-mean normalized minimum distance square criterion, and the expression is:

[0201]

[0202] Among them, f i and g i represent the reference sub-region and the deformed sub-region after the reference sub-region is deformed, respectively; ZSSD represents the mean normalized minimum square distance standard; and Represent the mean grayscale values ​​of the reference image and the deformed image respectively; is the normalization function of the sub-area, C ZNSSD The smaller the value, the better the correlation;

[0203] According to the formula of ZNSSD, calculate P 0 Value, denoted as C ZNSSD (P 0 ), the optimal solution of shape function parameters under this mathematical model is C ZNSSD (P * )≤C ZNSSD (P 0 ), give P 0 An increment ΔP, so that C ZNSSD (P 0 +ΔP)≤C ZNSSD (P 0 ), for C ZNSSD (P 0 +ΔP) is the zero-mean normalized minimum distance square standard mathematical model of the reference sub-area shape function increment, and the mathematical model is subjected to a second-order Taylor expansion, and its expression is:

[0204]

[0205] Among them, P 0 represents the shape function, ΔP represents the shape function increment, represents the gradient of the zero-mean normalized minimum distance squared standard;

[0206] Since we are solving the maximum (minimum) value, that is, C ZNSSD (P 0 +ΔP) with respect to ΔP is 0, and the derivative of the formula after the second-order Taylor expansion is derived, and its expression is

[0207]

[0208] Solving the above derivative formula, we can get the expression of shape function increment as follows:

[0209]

[0210] Using the inverse synthetic Gauss-Newton iteration method to obtain a new zero-mean normalized minimum distance square standard mathematical model of the shape function parameter increment, and obtaining a specific value of the shape function increment according to the data model;

[0211] The new reference sub-area is found by using the inverse synthetic Gauss-Newton iteration method. The new reference sub-area is as follows:

[0212]

[0213] When P 0 = 0, it means that the new reference sub-area completely overlaps with the original reference sub-area. 0 The optimal solution is P * =0, based on the shape function parameters of the previous iteration obtained by the inverse synthesis Gauss-Newton iteration method, the zero-mean normalized minimum distance square standard mathematical model of the shape function parameter increment is constructed as follows:

[0214]

[0215] Among them, ΔP is the shape function parameter, P old represents the shape function parameters of the previous iteration obtained by the inverse synthetic Gauss-Newton iterative method;

[0216] The simplified formula for constructing the zero-mean normalized minimum distance square standard mathematical model for shape function parameter increment is:

[0217]

[0218] Formula (26) indicates that the mean grayscale value of the sub-area image will not change due to the change of the sub-area shape, that is, the average grayscale value of each value on the image does not change dramatically, requiring the image illumination to be uniform and the brightness to be consistent; Formula (27) indicates that the grayscale value dispersion of the sub-area will not change due to the change of the sub-area shape, that is, the speckle distribution on the image is as random as possible, requiring that the speckle spraying on the image does not appear as a large white or large black area;

[0219] Calculate the gradient of the zero-mean normalized minimum distance square standard mathematical model of the shape function parameter increment at 0, and its expression is

[0220]

[0221] The Hessian matrix of the zero-mean normalized minimum distance square standard mathematical model for calculating the shape function parameter increment at 0 is expressed as follows:

[0222]

[0223] According to the gradient formula, Hessian matrix formula and shape function increment formula, the specific value of the shape function increment is obtained, and its expression is:

[0224]

[0225] Using the increment to update the shape function parameters of the new reference sub-region and the deformed sub-region corresponding to the new reference sub-region, to obtain an accurate displacement;

[0226] The relationship between the new reference sub-area and the new deformation sub-area is expressed as follows:

[0227] W(Δx i x 0 ,P new )=W(W(Δx i x 0 ,ΔP) -1 ,P old ) (31)

[0228] By generalizing the coefficient matrix in the shape function to a diagonally dominant matrix, the update formula is

[0229] w(P new )=w(P old )w(ΔP) -1 (32)

[0230] Among them, w(P) is the augmented matrix of the coefficient matrix formed by the shape function parameter P. When the P new =0, the obtained ΔP is the precise displacement.

[0231] The inverse Gauss-Newton iteration method has a fast iteration speed, can improve operating efficiency, has good robustness, high precision, and low memory usage. In the inverse Gauss-Newton iteration method, the Hessian matrix is ​​used in the iteration. The quantities involved in the calculation are all known and will not change with the number of iterations, which greatly speeds up the calculation speed of each iteration.

[0232] S43: performing displacement field calculation on the two groups of estimated displacements to convert the image displacement into actual displacement;

[0233] The method for calculating the displacement field of the two sets of estimated displacements is a well-known technical means. For example, by calibrating the internal and external parameters of the two cameras and the intersection distortion parameters, the projection matrices of the two cameras and the two-dimensional coordinates of the corresponding matching points of the two sets of speckle images are obtained. The expression is:

[0234]

[0235] Where (u1, v1) is the pixel coordinate in one camera’s coordinate system, (u2, v2) is the pixel coordinate in another camera’s coordinate system, and m represents the projection matrix parameter.

[0236] Simplifying the projection matrix of the two cameras, its expression is

[0237]

[0238] Among them, x w ,y w , z w Represents the three-dimensional coordinates of the pixel point;

[0239] The simplified projection matrix is ​​calculated using the least squares method, and its expression is:

[0240] w=(A T *A) -1 *A T *B (36)

[0241] Among them, the matrix expressions of A and B are

[0242]

[0243] The least squares method is used to calculate the three-dimensional coordinate point [x0, y0, z0] before deformation and the three-dimensional coordinate point [x1, y1, z1] of the image after deformation. The actual displacement [u, v, w] is obtained by the coordinate difference between the three-dimensional coordinate points before and after deformation. This embodiment can reduce the influence of binocular sensor calibration error, algorithm matching error, and imaging noise by using this solution.

[0244] S44: determining an actual strain value of the solidified soil under test according to the actual displacement; the actual strain value includes a transverse strain value at the maximum crack of the solidified soil and an average transverse strain value of an area surrounding the maximum crack;

[0245] Specifically, the average strain calculation method based on Green's formula calculates the strain of the target point according to the displacement results of the pixels in the surrounding area, specifically including:

[0246] Let X be the three-dimensional coordinate of a point in the space under the reference image, and x be the coordinate of the point after deformation. Then the expression for obtaining the deformation gradient at the X position is:

[0247]

[0248] The expression of the average deformation gradient of the neighborhood is

[0249]

[0250] Where: S Ω Represents the neighborhood area, Ω represents the neighborhood, and according to Green's formula, the average deformation gradient formula is:

[0251]

[0252] Where: n Ω The normal vector representing the region boundary;

[0253] The average Green-Lagrange strain in the neighborhood is expressed as

[0254]

[0255] Where: I represents the unit matrix; E Ω represents the mean strain;

[0256] and obtaining a cracking coefficient of cracks after the solidified soil is damaged based on the actual strain value, and the cracking coefficient is used to characterize the cracking condition of the solidified soil and identify the failure mode;

[0257] In a specific embodiment, the formula for obtaining the cracking coefficient of the crack after the solidified soil is damaged in S44 is:

[0258]

[0259] Where: λ represents the cracking coefficient of the crack after the solidified soil is damaged; ε xmax Indicates the transverse strain value at the maximum crack; represents the average transverse strain value of the neighborhood around the maximum crack;

[0260] The crack opening coefficient identification method in this embodiment specifically includes:

[0261] Starting from the cracking of the solidified soil after being subjected to stress, the brittle failure and plastic failure of the solidified soil can be identified by analyzing the cracking of the cracks after the solidified soil is destroyed; Figure 5 As shown in the figure, from the deformation characteristics analysis of plastic failure and brittle failure of solidified soil, the failure process of solidified soil with plastic failure has obvious progressive characteristics. Although the strain at the maximum crack is large, a large plastic zone will be formed in the neighborhood around the crack due to plastic expansion, resulting in a large strain in the surrounding area. Therefore, the ratio of the strain at the maximum crack to the strain in the surrounding area will be relatively small. The failure process of solidified soil with brittle failure shows the characteristics of sudden cracking, and there is no obvious expansion deformation when it fails. Although the strain at the maximum crack is large, there is no obvious plastic zone in the neighborhood around the crack, resulting in a small strain in the surrounding area. Therefore, the ratio of the strain at the maximum crack to the strain in the surrounding area will be relatively large. This ratio is defined as the cracking coefficient of the crack.

[0262] Therefore, based on the advantage of digital speckle technology in being able to measure the full-field lateral strain of the solidified soil, the cracking coefficient of the cracks after the solidified soil is calculated to characterize the cracking situation of this embodiment and identify the failure mode. xmax The calculation can be used to obtain the full-field x-direction strain at 3 / 5T0 through the 3D-DIC measurement system, and the maximum strain value in the full-field x-direction strain can be screened out. xmax ;for To calculate the strain field data in the x direction at the time 3 / 5T0, the strain data in the neighborhood of the maximum strain value are extracted by the maximum area search algorithm (see Table 1 for MATLAB code). Finally, the average value of all the extracted strain data is selected through box plot statistics.

[0263] Table 1. Maximum value field retrieval algorithm code

[0264]

[0265] A solidified soil failure pattern recognition system based on digital speckle includes an image acquisition module, an image preprocessing module, a reference sub-area image acquisition module, an image data processing module, and a solidified soil failure pattern recognition module;

[0266] An image acquisition module is used to acquire two sets of speckle image series of the solidified soil under test;

[0267] An image preprocessing module is used to perform grayscale conversion, interpolation, and denoising on the two sets of speckle image series in sequence to obtain two sets of preprocessed image series;

[0268] a reference sub-region image acquisition module, configured to decompose the reference images of the first set of pre-processed image series into first reference sub-region images using a sub-region matching model; and simultaneously perform stereo matching on the second set of pre-processed image series based on a feature algorithm of scale-invariant feature transformation to obtain second reference sub-region images;

[0269] An image data processing module is used to perform time-series matching on the first reference sub-area image and the second reference sub-area image based on a phase correlation method to obtain two sets of initial deformation and displacement values ​​of the solidified soil;

[0270] The inverse synthetic Gauss-Newton iteration method is used to solve and obtain the zero-mean normalized minimum distance square standard based on the two sets of initial deformation displacement values, and the zero-mean normalized minimum distance square standard is iteratively optimized to obtain two sets of estimated displacements;

[0271] The displacement fields of the two sets of estimated displacements are calculated to convert the image displacements into actual displacements;

[0272] The solidified soil failure mode identification module is used to determine the actual strain value of the solidified soil under test based on the actual displacement; and obtain the cracking coefficient of the cracks after the solidified soil is destroyed based on the actual strain value, and the cracking coefficient is used to characterize the identification failure mode of the cracking condition of the solidified soil;

[0273] Compared with the prior art, the method and system described in this embodiment have the following beneficial effects:

[0274] Stereo matching is performed on the second set of preprocessed image series using a feature algorithm based on scale-invariant feature transformation to obtain a second reference sub-area image. Time-series matching is performed on the first reference sub-area image and the second reference sub-area image to obtain the initial deformation displacement value of the solidified soil. The identified failure mode of the solidified soil under test is confirmed based on the displacement deformation value of the solidified soil. The present invention obtains the deformation result of the solidified soil after failure through high-precision detection of digital speckle, and quantitatively analyzes the failure mode of the solidified soil to achieve more accurate identification, providing a solid technical guarantee for the safety management and life prediction of engineering structures.

[0275] Case Study:

[0276] Taking ordinary Portland cement-stabilized soil (OPC) at 3d and 28d age with cement content of 3%, 9% and 15% as an example, the failure modes of ordinary Portland cement-stabilized soil under uniaxial compression are as follows: Figure 6 As shown in the figure, the failure morphology of the plastic failure solidified soil specimens is mostly manifested as volume expansion cracking, while the failure morphology of the brittle failure solidified soil specimens is that there is no obvious expansion deformation on the failure surface.

[0277] The stress-strain curve of ordinary Portland cement-stabilized soil under uniaxial compression conditions is as follows: Figures 7 and 8 As shown, the stress-strain curve of the plastically failed solidified soil specimen exhibits a relatively flat curve characteristic, while the brittlely failed solidified soil specimen exhibits a relatively steep curve characteristic. The points on the stress-strain curve represent the displacement and strain field results of the solidified soil surface measured using digital speckle patterning at different times. T0 represents the time from the start of the test to the peak stress of the solidified soil. The purpose is to ensure that the results of each subsequent set of digital speckle patterning measurements are compared under the same conditions. Through analysis of the failure morphology and stress-strain curves, it is known that the failure mode of the solidified soil with a 3% cement content is plastic failure, and the failure mode of the solidified soil with a 15% cement content is brittle failure. However, the solidified soil with a 9% cement content is in a state intermediate between the two failure modes and cannot be accurately identified. Therefore, it is necessary to use digital speckle patterning technology, namely the method and system proposed in this embodiment, to measure the cracking coefficient of ordinary Portland cement solidified soil for judgment. The steps for analyzing its failure mode using digital speckle patterning software are as follows:

[0278] Step 1: Open the exe file through MATLAB and the entire interface layout of the software will be displayed, as shown below:

[0279] Step 2: Based on the solidified soil speckle image captured by the imaging device (if the image format is already a grayscale image, skip this step), click the Import RGBA / RGB Image button. A window will pop up for the user to batch select the images to be converted. After the selection is completed, a pop-up window will prompt the user that the import is successful. Then click the Convert button. When the conversion is successful, a pop-up window will prompt the user. Finally, click the Save Grayscale Image button, which will provide the user with a save path, as shown in the figure:

[0280] Step 3: After the image is converted to the format required for analysis, the next step is to perform DIC analysis. Click the Select Analysis Area button, and a window will pop up to import the converted image. The user can select the analysis area. Then click the Import Calibration Data button, and a window will pop up to select the user's calibrated data (two clicks are required) to prepare for subsequent analysis, as shown in the figure:

[0281] Step 4: First, the user needs to select the sub-area size, sub-area step size, sub-area shape, and shape function order. The sub-area size is generally around 41, which gives a better displacement result accuracy. Figure 9 As shown in the figure, the sub-area step size needs to be selected according to the speckle distribution. If the speckle distribution is scattered, a larger step size is required. If the speckle distribution is dense, a smaller step size is required. The sub-area shape is divided into circular and square. The shape function order is divided into 0th order (describing the displacement change of the object's rigid body), 1st order (describing the displacement and stretching change of the object), and 2nd order (describing the displacement, stretching and distortion change of the object). Users can choose according to the deformation tested. After entering the above data, the information box on the right will display the data information. Finally, you only need to select different interpolation analysis methods to analyze and obtain the displacement and deformation results. Figures 10 and 11 As shown:

[0282] Step 5: After calculating the x-direction strain field of ordinary Portland cement-stabilized soil at 5 / 3T0 (cracks are obvious at 3 / 5T0) through digital speckle measurement results, the strain data of the neighborhood of the maximum strain value is extracted by the maximum area search algorithm to make box plot statistics, as shown in the following figure: Figure 12 shown.

[0283] The mean of each group is selected from the box plot as The maximum transverse strain value was found by screening the x-direction strain field. Finally, the cracking coefficient calculation formula was used to obtain the results shown in Table 2. By analyzing the cracking coefficient calculation results listed in Table 2, it was found that the cracking coefficient at the crack of ordinary Portland cement-stabilized soil with a cement content of 3% was significantly smaller than the cracking coefficient at the crack of ordinary Portland cement-stabilized soil with a cement content of 15%. In theory, the cracking behavior of ordinary Portland cement-stabilized soil undergoing plastic failure and brittle failure is significantly different, and this test data precisely verifies this theory.

[0284] Table 2. Cracking coefficients at cracks in ordinary Portland cement-stabilized soil

[0285]

[0286] Further analysis of the six data sets in Table 2 shows that the cracking coefficients of the 3-day-old ordinary Portland cement-stabilized soil with 3% and 9% cement content, and the 28-day-old ordinary Portland cement-stabilized soil with 3% cement content, are relatively small, indicating cracking and dehiscence of the soil under plastic failure, and can be classified as falling within the plastic deformation range. Conversely, the cracking coefficients of the 28-day-old ordinary Portland cement-stabilized soil with 9% and 15% cement content, and the 3-day-old ordinary Portland cement-stabilized soil with 15% cement content, are relatively large, exhibiting typical cracking and dehiscence of the soil under brittle failure, and can therefore be classified as falling within the brittle deformation range. To establish an effective failure mode identification standard, this example also takes the closest cracking coefficients from the two ranges and averages them to 9.69% as the recommended value for transitioning the soil from plastic failure to brittle failure. When the cracking coefficient is less than 9.69%, it is identified as plastic failure, and when the cracking coefficient is greater than 9.69%, it is identified as brittle failure.

[0287] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for identifying failure patterns of stabilized soil based on digital speckle, characterized in that: The specific steps include: S1: Obtain two sets of left and right speckle image series of the solidified soil under test through the calibrated camera; The two sets of speckle image series are subjected to grayscale conversion, interpolation and denoising respectively to obtain two sets of pre-processed image series; The left pre-processed image series is defined as the first set of pre-processed image series; The right pre-processed image series is defined as the second set of pre-processed image series; S2: decomposing the reference image of the first set of pre-processed image series into first reference sub-region images using a preset sub-region matching algorithm; S3: Stereo matching is performed on the second set of pre-processed image series based on a feature algorithm based on scale-invariant feature transformation to obtain a second reference sub-area image, specifically comprising the following steps: S31: constructing multiple groups of Gaussian pyramids for scale space division according to the second group of preprocessed image series; S32: Obtaining Gaussian blurred images of different scales in each layer structure of the multiple groups of Gaussian pyramids; S33: subtracting the Gaussian blurred images of different scales in adjacent layers within each group of the multiple Gaussian pyramids to obtain a difference image, thereby obtaining a Gaussian difference pyramid, namely, a DOG pyramid; Obtaining differential image feature points according to the DOG pyramid; and the differential image feature points are local pixel extreme points of the differential image obtained by the DOG pyramid space; and obtaining pixel fitting parameters for confirming the deformation of the solidified soil based on the local pixel extreme points, and determining a second reference sub-area image in the reference image of the second set of speckle image series according to the pixel fitting parameters; S4: performing time-series matching on the first reference sub-area image and the second reference sub-area image to obtain the initial deformation displacement value of the solidified soil, so as to confirm the identified failure mode of the solidified soil under test according to the displacement deformation value of the solidified soil.

2. The method for identifying failure patterns of stabilized soil based on digital speckle according to claim 1, characterized in that: The method for constructing multiple groups of Gaussian pyramids for scale space partitioning in S31 specifically includes the following steps: S311: obtaining the number of groups of Gaussian pyramids according to the second pre-processed image series, and encoding the number of groups of Gaussian pyramids to obtain a group sequence table of multiple Gaussian pyramids; And the formula for obtaining the number of Gaussian pyramid groups is: O=[log2(min(W,H))]-t (1) Where: O represents the number of groups of Gaussian pyramids; W, H represent the width and height of the second group of preprocessed images respectively; t represents the image size of the minimum resolution image; S312: Obtain multi-layer image data of any Gaussian pyramid group Q in a list of multiple Gaussian pyramid groups; The method for acquiring multi-layer image data includes: S3121: Define the total number of layers of the Gaussian pyramid group Q to be L, and the initial image set of each layer is an empty set; After doubling the size of the images in the second pre-processed image series in the Qth group, the images are used as the image data of the first layer of the Qth group of multiple Gaussian pyramids; The images in the first layer of the Qth group of multiple Gaussian pyramids are subjected to Gaussian convolution as the image data of the second layer of the Qth group; And the function expression of Gaussian convolution is Where: G(x,y) represents the Gaussian function; σ represents the scale space factor in the SIFT operator; x,y represent the position points of the image pixels in the second set of preprocessed image series; x0,y0 represent the position points of the image pixels after expansion in the second set of preprocessed image series; S3122: Update the scale space factor to k*σ to obtain a new Gaussian convolution function; The image of the second layer of the Q group is subjected to Gaussian convolution by the new Gaussian convolution function and used as the image data of the third layer of the Q group; S3123: Update the scale space factor to K again 2 *σ, to obtain the updated Gaussian convolution function; The image data of the third layer of the Q group is subjected to the updated Gaussian convolution function and Gaussian convolution is used as the image data of the fourth layer of the Q group; ...... Update and obtain the scale space factor K of the Gaussian convolution function in the Lth layer of the Qth group (L-1 )σ, to obtain the Gaussian convolution function of the Lth layer of the Qth group; The image of the L-1th layer of the Qth group is subjected to the Gaussian convolution function of the Lth layer of the Qth group, and the resulting image is used as the image data of the Lth layer of the Qth group; S313: Repeat step S312 to obtain O groups of Gaussian pyramids, each group including L layers of image data, totaling O*L images to achieve the construction of multiple groups of Gaussian pyramids.

3. The method for identifying failure patterns of stabilized soil based on digital speckle according to claim 2, characterized in that: The method for obtaining Gaussian blurred images of different scales in each layer structure of multiple groups of Gaussian pyramids in S32 includes: S321: Obtain the total number of layers S of each group in multiple Gaussian pyramids; Calculate and obtain the image Gaussian blur coefficient of each layer in the same group; And the formula for obtaining the image Gaussian blur coefficient is: Where: σ(O,S) represents the Gaussian blur coefficient of the image; σ0 represents the initial scale of the image, that is, the initial value of the Gaussian blur; S i Indicates the layer number code; S indicates the total number of layers in each group; S322: Based on the image Gaussian blur coefficient, obtain the image Gaussian blur coefficient of each group in the multiple groups of Gaussian pyramids to obtain Gaussian blurred images of different scales in each layer of the multiple groups of Gaussian pyramids.

4. The method for identifying failure patterns of stabilized soil based on digital speckle according to claim 3 is characterized in that: The S33 specifically includes the following steps: S331: Subtracting Gaussian blurred images of different scales in adjacent layers within each of the multiple Gaussian pyramids to obtain a difference image, and then obtaining all the difference images in the multiple Gaussian pyramids group by group and layer by layer; And build a Gaussian difference pyramid, namely DOG pyramid, based on all difference images; S332: Obtaining differential image feature points according to the DOG pyramid; and the differential image feature points are local pixel extreme points of the differential image obtained from the DOG pyramid space; And the formula for obtaining the local pixel extreme point is: Where: represents the local pixel extreme point of the differential image; D represents the DOG function about the feature point of the differential image; X represents the feature point of the differential image; σ represents the abbreviation of σ(O,S); S333: Obtaining pixel fitting parameters for confirming the deformation of the solidified soil based on the local pixel extreme points; And the formula for obtaining pixel fitting parameters is: Where: represents the pixel fitting parameter used to confirm the deformation of solidified soil; T represents transposition; The estimated positions of the differential image feature points in the reference image of the second set of speckle image series are determined according to the pixel fitting parameters, and the positions of the corresponding reference sub-areas in the reference image are confirmed based on the estimated positions based on the SURF feature matching algorithm, thereby obtaining a second reference sub-area image.

5. The method for identifying failure patterns of stabilized soil based on digital speckle according to claim 4 is characterized in that: The S4 specifically includes the following steps: S41: Based on the phase correlation method, the first reference sub-area image and the second reference sub-area image are time-series matched to obtain two sets of initial deformation displacement values ​​of the solidified soil; S42: using the inverse synthetic Gauss-Newton iteration method, solving and obtaining the zero-mean normalized minimum distance square criterion based on the two sets of initial deformation displacement values, and iteratively optimizing the zero-mean normalized minimum distance square criterion to obtain two sets of estimated displacements; S43: performing displacement field calculation on the two groups of estimated displacements to convert the image displacement into actual displacement; S44: determining the actual strain value of the solidified soil under test according to the actual displacement; The cracking coefficient of the cracks after the solidified soil is destroyed is obtained based on the actual strain value, and the cracking coefficient is used to characterize the cracking situation of the solidified soil and identify the failure mode.

6. The method for identifying failure patterns of stabilized soil based on digital speckle according to claim 5, characterized in that: The formula for obtaining the cracking coefficient of the crack after the solidified soil is damaged in S44 is: Where: λ represents the cracking coefficient of the crack after the solidified soil is damaged; ε xmax Indicates the transverse strain value at the maximum crack; The average transverse strain value of the neighborhood around the maximum crack is the actual strain value of the tested solidified soil.

7. A system based on the method for identifying failure patterns of stabilized soil based on digital speckle patterns according to any one of claims 1 to 6, characterized in that: It includes an image acquisition module, an image preprocessing module, a reference sub-area image acquisition module, an image data processing module and a solidified soil failure pattern recognition module; An image acquisition module is used to acquire two sets of speckle image series of the solidified soil under test; An image preprocessing module is used to perform grayscale conversion, interpolation, and denoising on the two sets of speckle image series in sequence to obtain two sets of preprocessed image series; a reference sub-region image acquisition module, configured to decompose the reference images of the first set of pre-processed image series into first reference sub-region images using a preset sub-region matching algorithm; and simultaneously perform stereo matching on the second set of pre-processed image series based on a feature algorithm of scale-invariant feature transformation to obtain second reference sub-region images; An image data processing module is used to perform time-series matching on the first reference sub-area image and the second reference sub-area image based on a phase correlation method to obtain two sets of initial deformation and displacement values ​​of the solidified soil; The inverse synthetic Gauss-Newton iteration method is used to solve and obtain the zero-mean normalized minimum distance square standard based on the two sets of initial deformation displacement values, and the zero-mean normalized minimum distance square standard is iteratively optimized to obtain two sets of estimated displacements; The displacement fields of the two sets of estimated displacements are calculated to convert the image displacements into actual displacements; The solidified soil failure mode recognition module is used to determine the actual strain value of the solidified soil under test based on the actual displacement; The cracking coefficient of the cracks after the solidified soil is destroyed is obtained based on the actual strain value, and the cracking coefficient is used to characterize the cracking situation of the solidified soil and identify the failure mode.

Images