Slope deformation monitoring method, system, medium and equipment

Abstract

The invention discloses a slope deformation monitoring method and system, a medium and equipment, and relates to the field of slope deformation monitoring, and the method comprises the steps: obtaining a slope image sequence with a fixed time interval; extracting features of the first image, tracking the features of the first image in the to-be-detected image by using a Lucas-Kanade optical flow method, estimating a view field translation amount and a rotation angle of the to-be-detected image relative to the first image, and performing view field motion correction on the to-be-detected image; performing semantic segmentation on the first image and the corrected to-be-detected image to obtain an image pair with a slope region mask; obtaining a feature point matching set of the image pair based on LoFTR; performing feature point mismatching elimination on the feature point matching set by using a two-stage mismatching elimination method to obtain a matching point pair set M [beta]; and calculating the displacement of each point pair in the M [beta], and obtaining a sub-pixel-level slope deformation field through Gaussian interpolation. The scheme is low in cost, small in calculation amount and high in precision.

Description

Technical Field

[0001] The present invention relates to the technical field of slope deformation monitoring, and in particular to a slope deformation monitoring method, system, medium and equipment. Background Art

[0002] During the development of mineral resources, operations such as rock excavation, blasting and underground mining will cause serious damage to the original stress balance of the entire mountain and slope. A large amount of data has proved that slope deformation monitoring is an important means of early warning for such geological disasters.

[0003] The high and steep slopes formed by open-pit mining change the state of occurrence of rock and soil, and the cavities formed by underground mining cause displacement, cracks and loosening of the upper mountain. These changes are accompanied by slope deformation information. Slope deformation can be divided into three main stages: deceleration creep, stable creep and accelerated creep. Before the accelerated creep stage, the deformation process of the slope is relatively slow, the deformation amplitude is small, and it is usually difficult to be detected by the human eye. However, once the accelerated creep stage is entered, slope disasters often occur and emergency measures need to be taken. As the number of slopes increases year by year around the world, casualties and property losses caused by slope collapse have become more and more serious. Therefore, establishing an accurate slope deformation monitoring system is of vital significance for preventing slope disasters.

[0004] Traditional slope monitoring usually uses total stations, levels, crack meters, GNSS displacement meters, fixed inclinometers and other equipment. These traditional methods have the following disadvantages: First, if full coverage monitoring is to be achieved, a large number of high-precision sensors must be deployed on the slope, which leads to high costs; second, technical personnel are required to enter the site to complete the deployment of monitoring points, which poses a great safety risk; third, the deployment time is long. It takes at least 3 to 4 days to complete the deployment of monitoring equipment on a slope, which makes it difficult to meet the requirements of rapid deployment.

[0005] The key technologies for existing slope deformation monitoring include: "3S" technology: including remote sensing (RS), geographic information system (GIS) and global positioning system (GPS) technology, but these monitoring methods are easily affected by terrain and meteorological conditions. The continuous monitoring of slopes is disturbed by meteorological conditions and satellite periodicity, and the equipment is expensive.

[0006] 3D laser scanning technology: by scanning, the position coordinates of a large number of dense points on the slope surface are obtained, and a 3D model is quickly constructed to reflect the displacement changes of the slope. However, the amount of data obtained by 3D laser scanning technology is usually large, which is difficult to process and costly; the scanning process is easily affected by the surrounding environment and introduces noise.

[0007] Synthetic Aperture Interferometric Radar Technology (InSAR): By transmitting wave signals with a radar, it reflects the deformation information of the slope, and has the advantages of high monitoring accuracy and real-time monitoring. The InSAR technology is affected by coherence and atmospheric delay, resulting in deviations. The data volume obtained by the SAR system is large and the equipment cost is high, which is not very suitable for some small projects. Summary of the Invention

[0008] The purpose of the present invention is to: in order to solve the problems of high cost and large data volume in the existing technology of slope deformation monitoring, a slope deformation monitoring method is proposed, including the following steps:

[0009] S1. Collect video data of the slope to be monitored, and extract frames to obtain an image sequence N = {p0, p1,..., p n};

[0010] S2. Extract features f0 from the first image p0, and use the Lucas-Kanade optical flow algorithm to track features f0 in the image to be detected p i (0 < i ≤ n), estimate the field-of-view translation amount and rotation angle of p i relative to p0, and perform field-of-view motion correction on p i ;

[0011] S3. Respectively perform semantic segmentation on p0 and the p i after field-of-view motion correction, extract the bare land in the image, and obtain an image pair (q0, q i ) with a slope area mask;

[0012] S4. Based on the LoFTR model, obtain the feature point matching result M i of the image pair (q0, q f );

[0013] S5. Remove the mis-matched feature points from M f , and obtain the final set of matching point pairs M β ;

[0014] S6. Calculate the displacement amount of each point pair in M β , and obtain the sub-pixel level slope deformation field through Gaussian interpolation.

[0015] Further, S2 is specifically:

[0016] S21. Convert p0 into a grayscale image, and use the feature point description vector generation function to obtain the feature point set f0;

[0017] S22. Convert p i into a grayscale image, and use the Lucas-Kanade optical flow algorithm to obtain the corresponding position set f of f0 in p i ini ;

[0018] S23, according to f0, f i Calculate from p0 to p i The Euclidean transformation is used to analyze the field of view displacement and rotation angle;

[0019] S24, according to the field of view displacement and rotation angle, p i Perform field of view displacement correction and rotation angle correction in sequence.

[0020] Furthermore, a two-stage mismatch elimination algorithm is used to f To eliminate feature point mismatches, the first stage uses the neighborhood information algorithm to f Eliminate feature point mismatches to obtain a matching point pair set M α In the second stage, the similar triangle algorithm is used to α Eliminate feature point mismatches to obtain the final matching point pair set M β .

[0021] Furthermore, the neighborhood information algorithm is used to f Eliminate feature point mismatches to obtain a matching point pair set M α , as follows:

[0022] M f A fixed-size neighborhood is established with the two points of a matching point pair as the center, and the number of matching points in the neighborhood does not exceed 20; the matching points in the neighborhood are calculated (R l , S l ) and the average distance from the center point If R l and S l The distance between Greater than or Then (R l , S l ) are regarded as mismatched point pairs and removed, and finally the set of matching point pairs M after the neighborhood information algorithm is removed is obtained. α .

[0023] Furthermore, we use the similar triangle algorithm to find α Eliminate feature point mismatches to obtain the final matching point pair set M β , as follows:

[0024] Select M α The matching point pair with the smallest Euclidean distance (R l , S l ), and then randomly select another two pairs with (R l , S l ) adjacent points (Rl-1 , S l-1 )、(R l+1 , S l+1 ), forming two triangles ΔR l-1 R l R l+1 , ΔS l-1 S l S l+1 , verify whether it meets the similar triangle conditions. If so, then (R l , S l )、(R l-1 , S l-1 ) as the reference point pair. If it does not meet the requirement, reselect (R l , S l ) adjacent point pairs, repeat the above process; the point pair to be detected (R m , S m ) and the reference point pair (R l , S l )、(R l-1 , S l-1 ) form two triangles ΔR l-1 R l R m , ΔS l-1 S l S m , if similar, then it is considered that (R m , S m ) is correct, if not similar, remove it (R m , S m ), and finally obtain the set of matching point pairs M after the similar triangle algorithm eliminates them β .

[0025] Furthermore, S6 is specifically:

[0026] S61, M β The local feature extraction is performed with the matching point in the middle as the center, and the β The image block feature vector F is centered at the matching point β ;

[0027] S62, Get M β The feature descriptor vector D corresponding to the matching point β ;

[0028] S63, F β and D β Cross product to obtain the image feature vector weighted by the feature descriptor;

[0029] S64, the weighted image feature vector is subjected to SoftArgMax operation to obtain M βThe displacement of the matching point pairs;

[0030] S65. Perform Gaussian interpolation on the displacement to obtain the slope deformation field at the sub-pixel level.

[0031] The present invention also proposes a slope deformation monitoring system, including:

[0032] A data acquisition module, configured to collect video data of the slope to be monitored, and extract frames to obtain an image sequence N = {p0, p1,..., p n};

[0033] A field-of-view motion correction module, configured to extract features f0 from the first image p0, and use the Lucas-Kanade optical flow algorithm to track features f0 in the image to be detected p i (0 < i ≤ n), estimate the field-of-view translation amount and rotation angle of p i relative to p0, and perform field-of-view motion correction on p i ;

[0034] A semantic segmentation module, configured to perform semantic segmentation on p0 and the p after field-of-view motion correction respectively, extract the bare land in the image, and obtain an image pair (q0, q i ) with a slope area mask; i );

[0035] A feature point matching module, configured to obtain the feature point matching result M of the image pair (q0, q i ) based on the LoFTR model f ;

[0036] A feature point false matching elimination module, configured to eliminate false matches of feature points in M f , and obtain the final set of matching point pairs M β ;

[0037] A slope deformation field acquisition module, configured to calculate the displacement of each point pair in M β , and obtain the slope deformation field at the sub-pixel level through Gaussian interpolation.

[0038] The present invention also proposes a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above-mentioned slope deformation monitoring method is implemented.

[0039] The present invention also proposes an electronic device, including a processor and a memory, the processor is connected to the memory, wherein the memory is used to store a computer program, the computer program includes computer-readable instructions, and the processor is configured to call the computer-readable instructions to execute the above-mentioned slope deformation monitoring method.

[0040] The beneficial effects brought by the technical solution provided by the present invention are:

[0041] The present invention first estimates the camera field of view motion and corrects the position of the image, eliminating the influence of slight camera motion on the slope displacement calculation results during data collection; semantic segmentation is performed on the collected image to eliminate the interference of non-target objects such as vegetation and falling rocks on subsequent point matching, and the feature point matching range is narrowed to improve the matching accuracy and reduce the amount of calculation; the LoFTR method is used for feature point matching, and a coarse-to-fine matching strategy is used, which is robust to the time-varying lighting conditions in the outdoor environment, achieving high efficiency and high-precision matching; the two-stage mismatch elimination algorithm can effectively eliminate mismatches that do not meet the local structural similarity, further improving the matching accuracy. The scheme of the present invention uses ordinary consumer-grade cameras or industrial cameras as monitoring equipment, which has the advantages of low cost, fast deployment speed, wide coverage, low cost, small amount of calculation, and high result accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 is a flow chart of a slope deformation monitoring method according to an embodiment of the present invention;

[0043] Figure 2 The embodiment of the present invention uses the neighborhood information algorithm to f Eliminate feature point mismatches to obtain a matching point pair set M α Flowchart of

[0044] Figure 3 The embodiment of the present invention uses a similar triangle algorithm to calculate M α Eliminate feature point mismatches to obtain the final matching point pair set M β Flowchart of

[0045] Figure 4 It is a block diagram of an electronic device in an exemplary embodiment of the present invention. DETAILED DESCRIPTION

[0046] To make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be further described below with reference to the accompanying drawings.

[0047] Example 1: A flow chart of a slope deformation monitoring method according to an embodiment of the present invention is shown in FIG. Figure 1 , specifically including the following steps:

[0048] S1. Collect video data of the slope to be monitored, extract frames to obtain an image sequence N={p0, p1, ..., p n}.

[0049] In the embodiment of the present invention, an industrial camera is used as a monitoring device, which is installed at a position where the camera's field of view is sufficient to cover the slope to be detected, video data is collected, and the video is frame-extracted at a fixed frequency to obtain images at fixed time intervals.

[0050] S2. Extract the feature f0 from the first image p0, and use the Lucas-Kanade optical flow algorithm to track the feature f0 in the image p i (0 < i ≤ n), estimate the field-of-view translation amount and rotation angle of p i relative to p0, and perform field-of-view motion correction on p i . Specifically:

[0051] S21. Convert p0 into a grayscale image, and use the feature point description vector generation function to obtain a set of feature points f0 that are easy to track. The f0 described by the feature points contains the position information of the feature points. The feature point description vector generation function is usually used for generating feature descriptors in computer vision, such as the SIFT (Scale-Invariant Feature Transform) algorithm.

[0052] S22. Convert p i into a grayscale image, and use the Lucas-Kanade optical flow algorithm to obtain the corresponding position set f i of f0 in p i .

[0053] S23. Use the estimateRigidTransform function in OpenCV to calculate the Euclidean transformation from p0 to p i according to f0 and f i , and parse the field-of-view displacement amount and rotation angle from it.

[0054] S24. According to the field-of-view displacement amount and rotation angle, perform field-of-view displacement amount correction and rotation angle correction on p i in sequence.

[0055] S3. Respectively perform semantic segmentation on p0 and the p i after field-of-view motion correction, extract the bare land in the image, and obtain an image pair (q0, q i ) with slope area masks.

[0056] Use the basic image segmentation model to perform semantic segmentation on p0 and p i , accurately understand the image content in each field, and achieve accurate segmentation of the slope and the surrounding environment. Exclude the interference of the surrounding environment, reduce the area range to be monitored, and speed up the subsequent calculation speed.

[0057] S4. Based on the LoFTR model, obtain the image pair (q0, q i) feature point matching result M f .

[0058] The LoFTR model is a local image feature matching method. LoFTR first establishes pixel-level dense matching at a coarse level, and then refines good matching at a fine level. The self-attention layer and cross-attention layer are used in the Transformer to obtain feature descriptors conditioned on two images. The LoFTR model includes a local feature extraction module, a global feature matching mapping module, a coarse matching module, and a coarse-to-fine transition matching module.

[0059] The local feature extraction module first uses the feature pyramid network (FPN) based on the frozen visual large model network weights to extract p0 and p i Extract features and generate global level feature maps and and fine-grained level feature maps and The image size of the global level feature map is 1 / 8 of the original input image size, and the fine-grained level feature map refers to the image size of the original input image p0 and p i 1 / 2 the size.

[0060] The global feature matching mapping module converts the global feature map and Flatten into a one-dimensional vector and add position encoding, input into the LoFTR module to obtain local features related to position and context and The LoFTR module consists of 4 self-attention layers and 4 cross-attention layers interlaced

[0061] The coarse matching module uses a double softmax operator, according to Calculate the score matrix between and Then according to The matching confidence between points j and k is calculated, and finally matching point pairs with a confidence higher than 0.2 are selected, and mutual nearest neighbor (MNN) constraints are performed to filter out abnormal global image matching and obtain global matching image feature point pairs Matching c .

[0062] From coarse to fine transition matching module, for Matching c Each coarse matching feature point pair (R l , S l ), first, the fine-grained level feature map and Locate its position (R′ l , S′ l), then cut two sets of local windows of size w×w, and use the LoFTR module to multiply the cut features in each window by N f times to generate R′ l and S′ l Two transformed local feature maps centered on and Will The center vector of Compare the correlation of all vectors in and calculate R′ l Each pixel in the neighborhood and S′ l By calculating the expectation of the probability distribution, we can obtain S′ l In p i The fine-grained matching result R′1 on the l , S′ l ) Collect and obtain the final matching point pair M f .

[0063] S5, for M f Eliminate feature point mismatches to obtain the final matching point pair set M β .

[0064] A two-stage false match elimination algorithm is used to f To eliminate feature point mismatches, the first stage uses the neighborhood information algorithm to f Eliminate feature point mismatches to obtain a matching point pair set M α In the second stage, the similar triangle algorithm is used to α Eliminate feature point mismatches to obtain the final matching point pair set M β .

[0065] Use the neighborhood information algorithm to f Eliminate feature point mismatches to obtain a matching point pair set M α The flow chart is as follows Figure 2 As shown, the details are as follows:

[0066] M f A matching point pair (R o , S o ) of two points R o and S o A fixed-size neighborhood is established for each center, and the number of matching points in the neighborhood does not exceed 20; the matching points in the neighborhood are calculated (R l , S l ) and the center point (R o , S o ) (R l With R oThe average distance of (S l With S o The average distance of l and S l The distance between Greater than or Then (R l , S l ) are considered as mismatched points and removed until M is calculated. f All matching points in the final result are the set of matching point pairs M after the neighborhood information algorithm eliminates them. α .

[0067] Use the similar triangle algorithm to α Eliminate feature point mismatches to obtain the final matching point pair set M β The flow chart is as follows Figure 3 As shown, the details are as follows:

[0068] Select M α The matching point pair with the smallest Euclidean distance (R l , S l ), and then randomly select another two pairs with (R l , S l ) adjacent points (R l-1 , S l-1 )、(R l+1 , S l+1 ), forming two triangles ΔR l-1 R l R l+1 , ΔS l-1 S l S l+1 , verify whether it meets the similar triangle conditions. If so, then (R l , S l )、(R l-1 , S l-1 ) as the reference point pair. If it does not meet the requirement, reselect (R l , S l ) adjacent point pairs, repeat the above process; the point pair to be detected (R m , S m ) and the reference point pair (R l , S l )、(R l-1 , S l-1 ) form two triangles ΔR l-1 R l R m , ΔS l-1 S l S m , if similar, then it is considered that (Rm , S m ) is correct. If they are not similar, they are excluded (R m , S m ), and finally, the set M of matching point pairs after excluding similar triangles by the algorithm is obtained β .

[0069] S6. Calculate the displacement of each point pair in M β , and obtain the slope deformation field at the sub-pixel level through Gaussian interpolation. Specifically:

[0070] S61. Perform local feature extraction with the matching points in M β as the center to obtain the image patch feature vector F β with the matching points in M β as the center.

[0071] S62. Obtain the feature descriptor vector D β corresponding to the matching points in M β .

[0072] S63. Cross-multiply F β and D β to obtain the image feature vector weighted by the feature descriptor.

[0073] S64. Perform the SoftArgMax operation on the weighted image feature vector to obtain the displacement of the matching point pairs in M β .

[0074] S65. Perform Gaussian interpolation on the displacement to obtain the slope deformation field at the sub-pixel level.

[0075] Example 2: In an exemplary embodiment, a slope deformation monitoring system is proposed, including:

[0076] A data acquisition module for collecting video data of the slope to be monitored and extracting frames to obtain an image sequence N = {p0, p1,..., p n};

[0077] A field-of-view motion correction module for extracting the feature f0 from the first image p0 and using the Lucas-Kanade optical flow algorithm to track the feature f0 in the image p i (0 < i ≤ n) to estimate the field-of-view translation and rotation angle of p i relative to p0, and perform field-of-view motion correction on p i ;

[0078] A semantic segmentation module for respectively performing semantic segmentation on p0 and the p i after field-of-view motion correction, extracting the bare land in the image, and obtaining an image pair (q0, qi );

[0079] Feature point matching module, used to obtain image pairs (q0, q i ) feature point matching result M f ;

[0080] Feature point mismatch elimination module is used to f Eliminate feature point mismatches to obtain the final matching point pair set M β ;

[0081] Slope deformation field acquisition module, used to calculate M β The displacement of each point pair in is used to obtain the sub-pixel slope deformation field through Gaussian interpolation.

[0082] In an exemplary embodiment, a computer-readable storage medium is included, the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned slope deformation monitoring method is implemented.

[0083] See also Figure 4 In an exemplary embodiment, an electronic device is also included, including at least one processor, at least one memory, and at least one communication bus.

[0084] Wherein, a computer program is stored in the memory, and the computer program includes computer-readable instructions. The processor calls the computer-readable instructions stored in the memory through the communication bus to execute the above-mentioned slope deformation monitoring method.

[0085] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A slope deformation monitoring method, characterized in that: The following steps are involved: S1. Collect video data of the slope to be monitored, extract frames to obtain an image sequence N={p0, p1, ..., p n }; S2. Extract the feature f0 from the first image p0, and use the Lucas-Kanade optical flow algorithm to track the feature f0 in the images to be detected p i (0 < i ≤ n), estimate the field-of-view translation and rotation angle of p i relative to p0, and perform field-of-view motion correction on p i ; S3, respectively p0 and p after field motion correction i Perform semantic segmentation to extract the exposed land in the image and obtain the image pair with slope area mask (q0,q i ); S4. Based on the LoFTR model, we obtain the image pair (q0,q i ) feature point matching result M f ; S5, for M f Eliminate feature point mismatches to obtain the final matching point pair set M β ; S6. Calculate M β The displacement of each point pair in is used to obtain the sub-pixel slope deformation field through Gaussian interpolation.

2. A slope deformation monitoring method according to claim 1, characterized in that: S2 is specifically: S21, convert p0 into a grayscale image, use the feature point description vector generation function, and obtain the feature point set f0; S22, p i Convert to grayscale image and use Lucas-Kanade optical flow algorithm to get f0 at p i The corresponding position set f in i ; S23, according to f0, f i Calculate from p0 to p i The Euclidean transformation is used to analyze the field of view displacement and rotation angle; S24, according to the field of view displacement and rotation angle, p i Perform field of view displacement correction and rotation angle correction in sequence.

3. A slope deformation monitoring method according to claim 1, characterized in that: A two-stage false match elimination algorithm is used to f To eliminate feature point mismatches, the first stage uses the neighborhood information algorithm to f Eliminate feature point mismatches to obtain a matching point pair set M α In the second stage, the similar triangle algorithm is used to α Eliminate feature point mismatches to obtain the final matching point pair set M β .

4. A slope deformation monitoring method according to claim 3, characterized in that: Use the neighborhood information algorithm to f Eliminate feature point mismatches to obtain a matching point pair set M α , as follows: M f A fixed-size neighborhood is established with the two points of a matching point pair as the center, and the number of matching points in the neighborhood does not exceed 20; the matching points in the neighborhood are calculated (R l ,S l ) and the average distance from the center point If R l and S l The distance between Greater than or Then (R l ,S l ) are regarded as mismatched point pairs and removed, and finally the set of matching point pairs M after the neighborhood information algorithm is removed is obtained. α .

5. A slope deformation monitoring method according to claim 3, characterized in that: Use the similar triangle algorithm to α Eliminate feature point mismatches to obtain the final matching point pair set M β , as follows: Select M α The matching point pair with the smallest Euclidean distance (R l ,S l ), and then randomly select another two pairs with (R l ,S l ) adjacent points (R l-1 ,S l-1 )、(R l+1 ,S l+1 ), forming two triangles ΔR l-1 R l R l+1 , ΔS l-1 S l S l+1 , verify whether it meets the similar triangle conditions. If so, then (R l ,S l )、(R l-1 ,S l-1 ) as the reference point pair. If it does not meet the requirement, reselect (R l ,S l ) adjacent point pairs, repeat the above process; the point pair to be detected (R m ,S m ) and the reference point pair (R l ,S l )、(R l-1 ,S l-1 ) form two triangles ΔR l-1 R l R m , ΔS l-1 S l S m , if similar, then it is considered that (R m ,S m ) is correct, if not similar, remove it (R m ,S m ), and finally obtain the set of matching point pairs M after the similar triangle algorithm eliminates them β .

6. A slope deformation monitoring method according to claim 1, characterized in that: S6 is specifically: S61, M β The local feature extraction is performed with the matching point in the middle as the center, and the β The image block feature vector F is centered at the matching point β ; S62, Get M β The feature descriptor vector D corresponding to the matching point β ; S63, F β and D β Cross product to obtain the image feature vector weighted by the feature descriptor; S64, the weighted image feature vector is subjected to SoftArgMax operation to obtain M β The displacement of the matching point pairs in ; S65. Perform Gaussian interpolation on the displacement to obtain a slope deformation field at a sub-pixel level.

7. A slope deformation monitoring system, characterized in that: include: The data acquisition module is used to collect the video data of the slope to be monitored, and extract frames to obtain an image sequence N={p0, p1, ..., p n }; A field-of-view motion correction module, which is used to extract feature f0 from the first image p0 and track feature f0 in the images to be detected p i (where 0 < i ≤ n) using the Lucas-Kanade optical flow algorithm, estimate the field-of-view translation amount and rotation angle of p i relative to p0, and perform field-of-view motion correction on p i ; The semantic segmentation module is used to separate p0 and p after the field of view motion correction. i Perform semantic segmentation to extract the exposed land in the image and obtain the image pair with slope area mask (q0,q i ); Feature point matching module is used to obtain image pairs (q0,q i ) feature point matching result M f ; Feature point mismatch elimination module is used to f Eliminate feature point mismatches to obtain the final matching point pair set M β ; Slope deformation field acquisition module, used to calculate M β The displacement of each point pair in is used to obtain the sub-pixel slope deformation field through Gaussian interpolation.

8. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

9. An electronic device, characterized in that: The method comprises a processor and a memory, wherein the processor and the memory are interconnected, wherein the memory is used to store a computer program, the computer program comprises computer-readable instructions, and the processor is configured to call the computer-readable instructions to execute the method according to any one of claims 1 to 6.

Images