Self-adaptive discontinuous displacement digital image calculation method

Abstract

The invention relates to a self-adaptive discontinuous displacement digital image calculation method, which comprises the following steps of: S1, obtaining image subsets and correlation coefficient values and displacement values of the image subsets through a standard DIC method; and S2, obtaining a correlation coefficient threshold value, comparing the correlation coefficient threshold value with the correlation coefficient value of the image subsets, if the correlation coefficient value of the image subsets is greater than the correlation coefficient threshold value, dividing the corresponding image subsets into a to-be-segmented subset queue, and if the correlation coefficient value of the image subsets is smaller than the correlation coefficient threshold value, ending screening of the image subsets. And S3, carrying out adaptive discontinuous subset tracking algorithm processing on the image subsets in the to-be-segmented subset queue one by one until the to-be-segmented subset queue is empty or the elements in the queue remain unchanged after the to-be-segmented subset queue is subjected to circulation for a set number of times. Compared with the prior art, the non-continuous displacement field in the image can be calculated, and the measurement precision and integrity of the non-continuous displacement field are remarkably improved.

Description

Technical Field

[0001] The invention relates to the field of image processing, and in particular to an adaptive discontinuous displacement digital image calculation method. Background Art

[0002] There are often a large number of crack textures in rocks. Under the action of external loads, the cracks will aggregate, diffuse, and penetrate, eventually leading to material damage and morphological destruction. Rocks are important materials for large-scale projects such as mining, underground structures, and water conservancy. Their deformation and destruction will threaten the safety of the project. Therefore, it is of great significance to use effective means to accurately locate the location of rock cracks and study the expansion mechanism of rock cracks.

[0003] At present, there are many observation techniques to comprehensively study crack propagation, such as magnifying glasses, optical microscopes, high-speed cameras, etc., but these methods rely on human observation and it is difficult to quantitatively obtain the displacement and strain field of the rock surface. As an optical testing technology, digital image correlation technology is simple to implement and has wide applicability. It can accurately measure the full-field displacement and strain field at the micro and macro scales. The DIC method can better study the mechanism of rock crack propagation, which is of great significance to understanding the mechanism of rock fracture instability.

[0004] The key to calculating subset displacement in the traditional DIC method is to accurately find the deformed subset corresponding to the reference subset. This method uses a first-order shape function to describe the subset deformation. However, there will be obvious discontinuity in the subset deformation near the crack, that is, there will be displacement discontinuity. If the first-order continuous shape function is continued to be used to describe the deformation of the subset in these areas, the correlation between the reference subset and the deformed subset will be significantly reduced, resulting in significant errors in the displacement measurement results. In addition, when using IC-GN iteration for sub-pixel matching, due to ignoring the sudden change of subset deformation, errors or even non-convergence will occur during the iteration process, making it impossible to obtain an accurate displacement field with sub-pixel accuracy.

[0005] Secondly, since the traditional DIC method is based on the assumption of continuous deformation, its algorithm is path-dependent. In this method, the first calculation point uses the method of solving the maximum value of the global correlation coefficient to obtain the initial guess, and the initial guess of the subsequent calculation point is determined by the displacement value of the previous calculation point. When the calculation path passes through cracks, low speckle quality, and areas with large deformation, the deformation subset on one side of the crack has a low correlation with the reference subset and cannot be used as the initial guess of the subset on the other side. Therefore, it is impossible to obtain an accurate initial guess value. When solving the displacement value with sub-pixel accuracy later, an inaccurate initial guess will cause the calculation to contain a large error. Using a displacement value with a high error as the initial guess of the next calculation point will lead to the accumulation of calculation errors, and even the displacement value cannot be calculated, which will affect the accuracy of the DIC calculation of displacement and strain field.

[0006] In summary, the traditional DIC method has limitations such as lack of accuracy and completeness when dealing with discontinuous deformation measurements near cracks. Summary of the invention

[0007] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and to provide an adaptive discontinuous displacement digital image calculation method to calculate the discontinuous displacement field of the image and significantly improve the measurement accuracy and integrity of the discontinuous displacement field.

[0008] The purpose of the present invention can be achieved by the following technical solutions:

[0009] An adaptive discontinuous displacement digital image calculation method, the method comprising:

[0010] Step 1, using the DIC method, calculate the correlation coefficient value and displacement value of the image to obtain the image subset and the correlation coefficient value and displacement value of each image subset;

[0011] Step 2, obtaining a correlation coefficient threshold and comparing it with the correlation coefficient value of the image subset. If the correlation coefficient value of the image subset is greater than the correlation coefficient threshold, the corresponding image subset is divided into a queue of subsets to be segmented;

[0012] Step 3, taking out a subset to be segmented from the queue of subsets to be segmented and calculating its correlation coefficient matrix;

[0013] Step 4, obtaining a binarization threshold, performing binarization processing and logistic regression segmentation on the correlation coefficient matrix, and obtaining a segmentation line, a segmented subset, and a main part of the segmented subset;

[0014] Step 5: Substitute the coordinates of the center point and the adjacent points of the segmented subset into the segmentation line equation for calculation to obtain the orientation values ​​between the center point and each adjacent point and the segmentation line. If the product of the orientation value of the center point and any adjacent point is greater than zero and the correlation coefficient value of the image subset where this adjacent point is located is less than the correlation coefficient threshold, proceed to the next step, otherwise return to step 3;

[0015] Step 6, perform sub-pixel matching on the main part of the segmented subset to obtain the deformation vector and correlation coefficient value of the main part with sub-pixel accuracy. If the correlation coefficient value of the segmented subset is less than the correlation coefficient threshold at this time, the subset is deleted from the queue; if the correlation coefficient value of the subset after sub-pixel matching is still greater than the correlation coefficient threshold, the binarization threshold is adaptively adjusted and the process returns to step 4;

[0016] Step 7: If the queue of subsets to be segmented is empty or the elements in the queue remain unchanged after the set number of cycles, the loop ends and the discontinuous displacement field of the image is obtained based on the remaining subsets in the queue of subsets to be segmented, otherwise return to step 3.

[0017] Furthermore, the formula for calculating the correlation coefficient value includes:

[0018]

[0019] Among them, S is a subset, C LS is the correlation coefficient value, i and j are the indices of each pixel in the subset, f(x i ,y j ) and g(x i ′,y j ′) represent the gray value of each pixel in the subset before and after deformation, respectively, f m and g m is the average gray value of the subset.

[0020] Furthermore, if the correlation coefficient value of the image subset is less than the correlation coefficient threshold, it is determined that the image subset does not contain non-continuous digital images and belongs to a continuous displacement field.

[0021] Furthermore, the process of calculating the correlation coefficient matrix includes:

[0022] A subset to be segmented is taken out from the queue of subsets to be segmented, and an image subset with a correlation coefficient lower than a correlation coefficient threshold is searched within a range of 1 to 2 pixels around the subset to be segmented, and the searched image subset is taken as a valid subset. If it cannot be found, the processing of the subset to be segmented is stopped, and a new subset to be segmented is taken out from the queue of subsets to be segmented;

[0023] The subset to be segmented is further segmented into small subsets, and the correlation between the small subset and the effective subset is calculated with the displacement value of the effective subset as a reference to obtain the correlation coefficient value of the small subset;

[0024] The correlation coefficient values ​​of each small subset are arranged according to the positions of the corresponding pixels in the image to construct a correlation coefficient matrix.

[0025] Furthermore, the process of performing the binarization processing and logistic regression segmentation includes:

[0026] Get the binary threshold. If the value in the correlation coefficient matrix is ​​greater than the binary threshold, it is set to 0; if it is less than the binary threshold, it is set to 1 to obtain a binary matrix.

[0027] The straight line equation is used as the fitting curve of the binary matrix dividing line, and the Sigmoid function is used as the prediction function to solve the straight line equation and obtain the prediction function equation;

[0028] Construct a logarithmic loss function, combine the logarithmic loss function with the Newton-Raphson method, iteratively solve the prediction function equation, and obtain the segmentation line and its parameters;

[0029] Use the segmentation line to segment the subset to be segmented, divide the subset to be segmented into two parts, and obtain the segmented subset;

[0030] Of the two parts of the subset to be segmented, the part where the center point of the subset is located is taken as the main part of the subset to be segmented.

[0031] Furthermore, the prediction function equation is:

[0032]

[0033] Among them, i and j are the coordinates of the pixel points in the subset to be segmented, θ0, θ1 and θ2 are the parameters of the straight line equation, and g(i,j) is the gray value of each pixel point in the subset to be segmented;

[0034] The logarithmic loss function is:

[0035]

[0036] Among them, m is the number of pixels in the segmentation subset, y i,j is the actual value of the pixel [i, j] in the binary matrix, and l(θ) is the logarithmic loss function.

[0037] Furthermore, the adjacent points of the segmented subset are the center points of each image subset within a range of 1 to 2 pixels around the center point of the segmented subset, and the number of adjacent points is no less than 8.

[0038] Furthermore, the calculation formula of the orientation value between the center point and the adjacent point and the segmentation line is:

[0039] h cir =θ0+θ1×radius+θ2×radius,

[0040] h xy =θ0+θ1×y+θ2×x,

[0041] Among them, h cir is the azimuth value of the center point, h xy is the azimuth value of the adjacent point, θ0, θ1 and θ2 are the parameters of the segmentation line, radius is the subset radius, and x and y are the horizontal and vertical coordinates of the adjacent point.

[0042] Furthermore, the process of performing sub-pixel matching on the main part of the segmented subset includes:

[0043] Substitute the main part of the segmented subset and the adjacent points in the main part of the segmented subset into the inverse combined Gauss-Newton iterative function to solve, and obtain a deformation vector with sub-pixel accuracy for the main part, which includes a correlation coefficient value with sub-pixel accuracy for the main part.

[0044] Furthermore, the formula for adaptively adjusting the binarization threshold includes:

[0045]

[0046] B t =B t0 +0.05×sig,

[0047] Among them, B t0 is the initial binarization threshold, B t is the binarization threshold after adaptive adjustment, and sig is the adaptive parameter.

[0048] Compared with the prior art, the present invention has the following beneficial effects:

[0049] 1. Based on the standard DIC method, the present invention also adopts an adaptive non-continuous subset tracking algorithm to process non-continuous digital images, which can obtain a more accurate crack displacement field and locate defects in the image more accurately; in the algorithm, a queue of subsets to be segmented is created, and the subsets containing non-continuous images are further processed to improve the accuracy of displacement measurement; a correlation coefficient matrix is ​​constructed to effectively improve the recognition and processing capabilities of discontinuous deformation areas near cracks, and also provides reliable technical support for accurate measurement and analysis of cracks and their surrounding deformations; the correlation coefficient matrix is ​​binarized to segment the subsets to be segmented, and a straight line equation is used as a pseudo-line of the dividing line in the process of segmenting the subsets. The combined curve reduces the solution parameters and makes the calculation more concise; the deformation vector is solved by sub-pixel matching on the main part after the subset segmentation, and the correlation coefficient value is compared with the correlation coefficient threshold again to further screen out the subsets with discontinuous areas, ensuring the integrity of the calculation of the discontinuous displacement field of the subset; the binarization processing threshold of the subsets with discontinuous areas after sub-pixel matching is adaptively adjusted until there is no discontinuous area in the main part of the subset, which effectively avoids the high correlation coefficient and non-convergence of iteration when using Gauss-Newton iteration for sub-pixel matching, and can obtain a continuous deformation displacement field of the main part, further improving the accuracy of displacement calculation of the continuous deformation area;

[0050] 2. The present invention has high computational efficiency, and the algorithm can implement multi-threaded optimization. In practical applications, it can create a queue of subsets to be segmented equal to the number of each computing thread, and assign an adaptive binarization processing threshold to each queue, which significantly improves the computational efficiency of the DIC algorithm and enables it to process large amounts of data faster;

[0051] 3. In actual use, a camera is used to continuously photograph the surface of a rock sample to record its surface changes under different loading conditions. The real-time photographed pictures are imported into the software and the displacement measurement and calculation are performed using the present invention, so that real-time crack displacement field changes can be obtained, which further helps the efficiency and accuracy of crack analysis;

[0052] 4. The present invention can measure the discontinuous displacement field of various rock surfaces and calculate the fracture parameters of cracks through the measured accurate displacement field, which is crucial for understanding the fracture behavior of materials under stress and provides important quantitative indicators for predicting the fracture performance of materials; in addition, the present invention can also be used in the fields of slope monitoring of indoor models, tunnel surrounding rock monitoring, etc., and also has a driving role in the engineering field;

[0053] 5. In the present invention, the subsets stored in the queue for segmentation are traversed multiple times, and the subsets in the queue are continuously processed in a loop, so that even if a subset is surrounded by invalid points and cannot be processed, it can be skipped first, and the next subset is measured and processed. In the next round of processing, the subsequently processed subset is used as a valid point to provide processing assistance to the former, avoiding the phenomenon in the prior art that the displacement measurement is incomplete due to the inability to obtain the initial guess of the subset. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 It is a flow chart of the present invention;

[0055] Figure 2 It is a flow chart of the adaptive non-continuous subset tracking algorithm of the present invention;

[0056] Figure 3 A schematic diagram of the process of calculating the correlation coefficient matrix of the present invention;

[0057] Figure 4 This is a schematic diagram of searching for effective points in the present invention. DETAILED DESCRIPTION

[0058] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0059] Example 1

[0060] This embodiment intends to disclose an adaptive non-continuous displacement digital image calculation method. In this embodiment, the method is intended to be used on an image containing cracks to identify cracks in the image. The method process is as follows: Figure 1 As shown, including:

[0061] Step S1, using a standard DIC method, the image is processed to obtain an initial displacement field for distinguishing whether there is a crack in the image ROI area, and image subsets and correlation coefficient values ​​and displacement values ​​of each image subset are obtained.

[0062] The standard DIC method for image processing includes:

[0063] Preprocess the image, including denoising and contrast enhancement, to improve image quality and reduce errors in subsequent calculations;

[0064] The image is divided into several image subsets, each of which contains a certain number of pixels. These subsets will serve as the basic calculation units of the DIC method.

[0065] Calculate the correlation coefficient value, displacement value, strain value, strain tensor and other parameters of the image subset;

[0066] The crack orientation is centrally identified based on the strain value, the displacement discontinuity region is identified based on the displacement value, and the crack tip orientation is identified based on the strain tensor;

[0067] Combining the above identification operations, the initial displacement field is obtained.

[0068] The formula for calculating the correlation coefficient value includes:

[0069]

[0070] Among them, S is a subset, C LS is the correlation coefficient value, i and j are the indices of each pixel in the subset, f(x i ,y j ) and g(x i ′,y j ′) represent the gray value of each pixel in the subset before and after deformation, respectively, f m and g m is the average gray value of the subset.

[0071] Step S2, obtain the correlation coefficient threshold and compare it with the correlation coefficient value of the image subset. If the correlation coefficient value of the image subset is greater than the correlation coefficient threshold, the corresponding image subset is divided into the queue of subsets to be segmented. If the correlation coefficient value of the image subset is less than the correlation coefficient threshold, it is determined that the image subset does not contain non-continuous digital images and belongs to a continuous displacement field, and the screening of this image subset is ended.

[0072] The queue of subsets to be segmented is a subset of images that may contain cracks.

[0073] Step S3, the image subsets in the queue of the segmented subsets are processed one by one as follows: Figure 2The adaptive non-continuous subset tracking algorithm shown is processed until the queue of subsets to be segmented is empty or the elements in the queue remain unchanged after ten cycles.

[0074] As stated Figure 2 The detailed process of the adaptive non-continuous subset tracking algorithm shown is as follows:

[0075] Step S301, taking out a subset to be segmented from a queue of subsets to be segmented.

[0076] Step S302, finding the valid subset closest to the subset to be segmented, specifically includes:

[0077] An image subset with a correlation coefficient lower than the correlation coefficient threshold is searched within a range of 1 to 2 pixels around the subset to be segmented, and the searched image subset is taken as a valid subset. If no subset is found, the process returns to step S301.

[0078] Step S303, calculating the correlation coefficient matrix of the subset to be segmented.

[0079] The process of calculating the correlation coefficient matrix is ​​as follows Figure 3 Shown include:

[0080] like Figure 3 As shown in (b), the subset to be segmented is further segmented into small subsets, and the displacement value of the effective subset is used as a reference to calculate the correlation between the small subset and the effective subset to obtain the correlation coefficient value of the small subset;

[0081] Arrange the correlation coefficient values ​​of each small subset according to the position of the corresponding pixel points in the image, such as Figure 3 As shown in (c), the correlation coefficient matrix is ​​constructed.

[0082] Step S304, obtaining a binarization threshold, and performing binarization and logistic regression segmentation on the correlation coefficient matrix to obtain a segmentation line, a segmented subset, and a main part of the segmented subset.

[0083] The process of binarization and logistic regression segmentation includes:

[0084] Get the binary threshold. If the value in the correlation coefficient matrix is ​​greater than the binary threshold, it is set to 0; if it is less than the binary threshold, it is set to 1 to obtain a binary matrix.

[0085] The straight line equation is used as the fitting curve of the binary matrix dividing line, and the Sigmoid function is used as the prediction function to solve the straight line equation and obtain the prediction function equation;

[0086] Construct a logarithmic loss function, combine the logarithmic loss function with the Newton-Raphson method, iteratively solve the prediction function equation, and obtain the segmentation line and its parameters;

[0087] Use the segmentation line to segment the subset to be segmented, divide the subset to be segmented into two parts, and obtain the segmented subset;

[0088] After the subset is divided into two parts, the subset on the side where the center point is located, that is, the main part, is further analyzed. The main part needs to be stored in the computer's memory in the next step to solve the deformation vector of the main part.

[0089] The prediction function equation is:

[0090]

[0091] Among them, i and j are the coordinates of the pixel points in the subset to be segmented, θ0, θ1 and θ2 are the parameters of the straight line equation, and g(i,j) is the gray value of each pixel point in the subset to be segmented;

[0092] The logarithmic loss function is:

[0093]

[0094] Among them, m is the number of pixels in the segmentation subset, y i,j is the actual value of the pixel [i, j] in the binary matrix, and l(θ) is the logarithmic loss function.

[0095] Step S305, searching for a valid point on the same side as the center point, if there is a valid point, performing sub-pixel matching on the main part of the segmented subset, otherwise returning to step S301.

[0096] The process of finding a valid point on the same side as the center point includes:

[0097] The adjacent points of the segmented subset are the center points of the image subset within 1 to 2 pixels around the center point of the segmented subset. After the subset is segmented, the adjacent points on the same side of the segmentation line as the center point of the main part are used as the initial guess of the main part.

[0098] Use as Figure 4 The search path sequence shown is to substitute the coordinates of the adjacent points of the segmented subset into the segmentation line equation one by one for calculation, and obtain the azimuth value between the adjacent points and the segmentation line;

[0099] Substitute the coordinates of the center point of the segmented subset into the segmentation line equation for calculation to obtain the orientation value between the center point and the segmentation line;

[0100] The calculation formula for the orientation value between the center point and the adjacent points and the dividing line is:

[0101] h cir =θ0+θ1×radius+θ2×radius,

[0102] hxy =θ0+θ1×y+θ2×x,

[0103] Among them, h cir is the azimuth value of the center point, h xy is the azimuth value of the adjacent point, θ0, θ1 and θ2 are the parameters of the segmentation line, radius is the subset radius, x and y are the horizontal and vertical coordinates of the adjacent point;

[0104] If the product of the orientation value of the center point and a certain adjacent point is greater than zero, and the correlation coefficient value of the image subset where the adjacent point is located is less than the correlation coefficient threshold, then the adjacent point is judged to be a valid point on the same side as the center point.

[0105] Step S306, perform sub-pixel matching on the main part of the segmented subset to obtain the deformation vector and correlation coefficient value of the main part with sub-pixel accuracy. If the correlation coefficient value of the segmented subset is less than the correlation coefficient threshold at this time, the subset is deleted from the queue; if the correlation coefficient value of the subset after sub-pixel matching is still greater than the correlation coefficient threshold, the binarization threshold is adaptively adjusted, and it is determined whether the binarization threshold is adaptively adjusted 10 times. If so, jump to step S207, otherwise, the binarization threshold is adaptively adjusted and return to step S304.

[0106] The process of sub-pixel matching of the main part of the segmented subset includes:

[0107] Substitute the main part of the segmented subset and the adjacent points in the main part of the segmented subset into the inverse combined Gauss-Newton iterative function to solve, and obtain a deformation vector with sub-pixel accuracy for the main part, which includes a correlation coefficient value with sub-pixel accuracy for the main part.

[0108] The formula for adaptively adjusting the binarization threshold includes:

[0109]

[0110] B t =B t0 +0.05×sig,

[0111] Among them, B t0 is the initial binarization threshold, B t is the binarization threshold after adaptive adjustment, and sig is the adaptive parameter.

[0112] According to the sub-pixel matching results, automatically adjusting the binarization processing threshold 3 to 5 times will usually produce a suitable threshold to exclude cracks in the main part, but for the sake of safety it is set here to 10 times. In another embodiment, the number of adjustments will be 5 times.

[0113] Step S307, determine whether this segmented subset is the last subset in the queue of subsets to be segmented. If not, return to step S301. If so, determine whether the queue of subsets to be segmented is empty or has gone through ten cycles and the elements in the queue remain unchanged. If so, end the algorithm processing and output the queue of subsets to be segmented. Otherwise, return to step S301.

[0114] Step S4, obtaining a discontinuous displacement field of the image according to the remaining subsets in the queue of subsets to be segmented, and then obtaining cracks in the image, after which the cracks can be further analyzed manually.

[0115] Example 2

[0116] Based on Example 1, this embodiment provides an electronic device, including: one or more processors and a memory, wherein the memory stores one or more programs, and the one or more programs include instructions for executing the aforementioned adaptive discontinuous displacement digital image calculation method.

[0117] At the hardware level, the electronic device includes a processor, an internal bus, a network interface, a memory, and a non-volatile memory, and may also include hardware required for other services. The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it to implement the above-mentioned adaptive discontinuous displacement digital image calculation method. Of course, in addition to the software implementation, the present invention does not exclude other implementation methods, such as logic devices or a combination of software and hardware, etc., that is to say, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0118] The memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0119] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0120] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. An adaptive non-continuous displacement digital image calculation method, characterized in that: The method comprises: Step 1, using the DIC method, calculate the correlation coefficient value and displacement value of the image to obtain the image subset and the correlation coefficient value and displacement value of each image subset; Step 2, obtaining a correlation coefficient threshold and comparing it with the correlation coefficient value of the image subset. If the correlation coefficient value of the image subset is greater than the correlation coefficient threshold, the corresponding image subset is divided into a queue of subsets to be segmented; Step 3, taking out a subset to be segmented from the queue of subsets to be segmented and calculating its correlation coefficient matrix; Step 4, obtaining a binarization threshold, performing binarization processing and logistic regression segmentation on the correlation coefficient matrix, and obtaining a segmentation line, a segmented subset, and a main part of the segmented subset; Step 5: Substitute the coordinates of the center point and the adjacent points of the segmented subset into the segmentation line equation for calculation to obtain the orientation values ​​between the center point and each adjacent point and the segmentation line. If the product of the orientation value of the center point and any adjacent point is greater than zero and the correlation coefficient value of the image subset where this adjacent point is located is less than the correlation coefficient threshold, proceed to the next step, otherwise return to step 3; Step 6, perform sub-pixel matching on the main part of the segmented subset to obtain the deformation vector and correlation coefficient value of the main part with sub-pixel accuracy. If the correlation coefficient value of the segmented subset is less than the correlation coefficient threshold at this time, the subset is deleted from the queue; if the correlation coefficient value of the subset after sub-pixel matching is still greater than the correlation coefficient threshold, the binarization threshold is adaptively adjusted and the process returns to step 4; Step 7: If the queue of subsets to be segmented is empty or the elements in the queue remain unchanged after the set number of cycles, the loop is terminated to obtain the discontinuous displacement field of the image, otherwise, return to step 3.

2. The adaptive non-continuous displacement digital image calculation method according to claim 1, characterized in that: The formula for calculating the correlation coefficient value includes: Among them, S is a subset, C LS is the correlation coefficient value, i and j are the indices of each pixel in the subset, f(x i ,y j ) and g(x i ′,y j ′) represent the gray value of each pixel in the subset before and after deformation, respectively, f m and g m is the average gray value of the subset.

3. The adaptive discontinuous displacement digital image calculation method according to claim 1, characterized in that: If the correlation coefficient value of the image subset is less than the correlation coefficient threshold, it is determined that the image subset does not contain non-continuous digital images and belongs to a continuous displacement field.

4. The adaptive non-continuous displacement digital image calculation method according to claim 1, characterized in that: The process of calculating the correlation coefficient matrix includes: A subset to be segmented is taken out from the queue of subsets to be segmented, and an image subset with a correlation coefficient lower than a correlation coefficient threshold is searched within a range of 1 to 2 pixels around the subset to be segmented, and the searched image subset is taken as a valid subset. If it cannot be found, the processing of the subset to be segmented is stopped, and a new subset to be segmented is taken out from the queue of subsets to be segmented; The subset to be segmented is further segmented into small subsets, and the correlation between the small subset and the effective subset is calculated with the displacement value of the effective subset as a reference to obtain the correlation coefficient value of the small subset; The correlation coefficient values ​​of each small subset are arranged according to the positions of the corresponding pixels in the image to construct a correlation coefficient matrix.

5. The adaptive non-continuous displacement digital image calculation method according to claim 1, characterized in that: The process of performing the binarization processing and logistic regression segmentation includes: Get the binary threshold. If the value in the correlation coefficient matrix is ​​greater than the binary threshold, it is set to 0; if it is less than the binary threshold, it is set to 1 to obtain a binary matrix. The straight line equation is used as the fitting curve of the binary matrix dividing line, and the Sigmoid function is used as the prediction function to solve the straight line equation and obtain the prediction function equation; Construct a logarithmic loss function, combine the logarithmic loss function with the Newton-Raphson method, iteratively solve the prediction function equation, and obtain the segmentation line and its parameters; Use the segmentation line to segment the subset to be segmented, divide the subset to be segmented into two parts, and obtain the segmented subset; Of the two parts of the subset to be segmented, the part where the center point of the subset is located is taken as the main part of the segmented subset.

6. The adaptive non-continuous displacement digital image calculation method according to claim 5, characterized in that: The prediction function equation includes: Among them, i and j are the coordinates of the pixel points in the subset to be segmented, θ0, θ1 and θ2 are the parameters of the straight line equation, and g(i,j) is the gray value of each pixel point in the subset to be segmented; The logarithmic loss function is: Among them, m is the number of pixels in the segmentation subset, y i,j is the actual value of the pixel [i, j] in the binary matrix, and l(θ) is the logarithmic loss function.

7. The adaptive non-continuous displacement digital image calculation method according to claim 1, characterized in that: The adjacent points of the segmented subset are the center points of each image subset within a range of 1 to 2 pixels around the center point of the segmented subset, and the number of adjacent points is no less than 8.

8. The adaptive non-continuous displacement digital image calculation method according to claim 1, characterized in that: The calculation formula of the orientation value between the center point and the adjacent point and the segmentation line is: h cir =θ0+θ1×radius+θ2×radius, h xy =θ0+θ1×y+θ2×x, Among them, h cir is the azimuth value of the center point, h xy is the azimuth value of the adjacent point, θ0, θ1 and θ2 are the parameters of the segmentation line, radius is the subset radius, and x and y are the horizontal and vertical coordinates of the adjacent point.

9. The adaptive non-continuous displacement digital image calculation method according to claim 1, characterized in that: The process of performing sub-pixel matching on the main part of the segmented subset comprises: Substitute the main part of the segmented subset and the adjacent points in the main part of the segmented subset into the inverse combined Gauss-Newton iterative function to solve, and obtain a deformation vector with sub-pixel accuracy for the main part, which includes a correlation coefficient value with sub-pixel accuracy for the main part.

10. The adaptive non-continuous displacement digital image calculation method according to claim 1, characterized in that: The formula for adaptively adjusting the binarization threshold includes: B t =B t0 +0.05×sig, Among them, B t0 is the initial binarization threshold, B t is the binarization threshold after adaptive adjustment, and sig is the adaptive parameter.

Images