Matching method and system for large-width satellite-borne remote sensing image

Abstract

The application discloses a wide-width spaceborne remote sensing image matching method and system, which realizes the matching of the spaceborne remote sensing image through four parts of image blocking, local image block matching, false matching point elimination and matching result output. The image blocking refers to extracting a limited number of local image blocks in the spaceborne remote sensing image to reduce the calculation amount in the image matching process; the local image block matching refers to matching the homonymic points between the image blocks by using the existing matching method; the false matching point elimination refers to eliminating the false matching points in the matching result after obtaining the preliminary matching points; and the matching result output refers to selectively outputting the points in the matching result according to the requirement. Based on the method, the matching algorithm in the computer vision field can be easily applied to the remote sensing image.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of remote sensing image processing, and is a matching method for large-size satellite-borne remote sensing images. BACKGROUND

[0002] All current image matching methods are designed for small-size images (generally 300-2000 in length of side), while satellite-borne remote sensing images are usually large in size in order to balance the width and resolution. For example, the GF7 forward-looking image is 31267 pixels in length and 30996 pixels in width. When the image matching method is directly applied to the satellite-borne remote sensing image, the following two problems exist: (1) the image data volume is too large, and a considerable amount of calculation is required for reading and calculating at one time, which is impossible for the memory of most computers; (2) the reliability of the matching algorithm is greatly reduced: the satellite-borne remote sensing image of such a large size can extract hundreds of millions of feature points, and the specificity of the feature points in a single image is reduced due to the increase in size, which leads to a sharp increase in the probability of false matching of such feature points, and the reliability of the image matching result is reduced. Therefore, it is necessary to design a matching framework for satellite-borne remote sensing images, so that the current matching method can be easily applied to satellite-borne remote sensing images. SUMMARY

[0003] The technical problem solved by the present application is to provide a matching framework for satellite-borne remote sensing images, so that the conventional matching algorithm can be well applied to the matching work of satellite-borne remote sensing images.

[0004] The technical solution for achieving the purpose of the present application is: a large-width satellite-borne remote sensing image matching method, comprising the following steps:

[0005] Step 1, block the image based on the image plane, i.e. block the image based on the image plane coordinates of the original image, and extract the image block on the reference image based on geographic information as the reference;

[0006] Step 2, match the extracted image block by using the matching method, and preliminarily obtain the homonymic points between the image blocks;

[0007] Step 3, use the local plus global false matching point elimination strategy to eliminate the false matching points in the matching result;

[0008] The local false matching point elimination is based on the affine model, and the global false matching point elimination is based on the fitting of the quadratic polynomial residual;

[0009] Step 4, selective output of the points in the matching result, i.e. only one point is selected from each image block participating in the matching to ensure the uniformity of the distribution of the whole scene image matching result.

[0010] Further, in step 1, if the width of the original image is w, the length is h, the number of image blocks in the width direction is n x , and the number of image blocks in the height direction is n y , assuming that the size of the image block participating in matching is s, the calculation method of the coordinates (x, y) of the sampling point is as follows:

[0011]

[0012] x i，j =S x +i*I x

[0013] y i，j =S y +j*I y

[0014] wherein (S x , S y ) is the coordinate of the starting sampling, i and j are the serial numbers of the horizontal sampling and the vertical sampling, i≤n x , j≤n y ; after obtaining the above image coordinates (x i，j , y i，j ), the image coordinates are converted into object coordinates (Lon i，j , Lat i，j ), so that the coordinate system is unified to the object coordinate system, and then the above object coordinates are taken as the center and s is taken as the side length to obtain the sampling block.

[0015] Further, the specific implementation of step 2 includes two steps, the first step is to eliminate the angle and resolution difference between the image blocks; the second step is to obtain the homonymic points of the image blocks after eliminating the angle and resolution difference through a matching algorithm; and the specific implementation includes the following sub-steps:

[0016] Step 21, image rotation resolution difference elimination: first, the minimum circumscribed rectangle of the intersection region of the two images is calculated to obtain four object coordinates A, B, C and D of the minimum circumscribed rectangle, then the image is converted with the center of the image block as the origin, and the above coordinates are all subtracted by the average value of the four corner points of the corresponding image, so as to obtain the conversion matrix H of the original image block to the reference image block with the center of the image block as the origin; in the subsequent matching process, the conversion matrix H is used to convert the original image block to obtain an image block consistent with the resolution and angle of the reference image block;

[0017] Step 22, after obtaining the image block consistent in angle and resolution, the image blocks are matched by using a matching method.

[0018] Further, the matching method is divided into feature matching and template matching; wherein, the template matching includes template extraction, template similarity calculation, the most similar template center and the center of the image to be matched, i.e. a pair of matching points; the feature matching includes feature point extraction, feature point description, feature point association, wherein the feature point association is judged according to the similarity between the feature point descriptors, and the most similar two points are a pair of matching points.

[0019] Further, the specific implementation mode of the RANSAC error matching point elimination based on the local affine model is as follows:

[0020] ①Randomly take points to solve the affine transformation matrix

[0021] Randomly select three pairs of homonymic points in the matching result point set X and Y in step 3, and solve the affine transformation matrix H based on the least square method. The solving method of the transformation matrix is as follows:

[0022] First, express the matching point pairs (x i , y i ) and (u i , v i ) as homogeneous coordinates and express them in matrix form, as shown in the following formula:

[0023]

[0024] Wherein, X and Y respectively represent the local matching point set in the original image and the reference image, and n represents the number of matching pairs;

[0025] Next, according to the formula H=Y T X(X T X) -1 Calculate the least square solution to obtain the conversion matrix H;

[0026] ②Predict the matching point position

[0027] Map all matching points X using the calculated affine transformation matrix H to obtain the predicted matching point position X';

[0028] ③Calculate the number of inliers

[0029] Calculate the Euclidean distance E between the predicted matching point position X' and the actual matching point X. If E is less than the threshold value T, it is determined as an inlier, and if E is greater than the threshold value T, it is considered as an outlier. The total number of inliers contained in X is counted;

[0030] ④Repeat steps ① to ③, and take the data set with the largest number of inliers as the best data set. The point pairs contained in the data set are the homonymic points correctly matched locally.

[0031] Further, the specific implementation of the global mismatched point elimination method based on quadratic polynomial residual fitting is as follows:

[0032] The fitting formula of the quadratic polynomial is as follows:

[0033] A = (X T X) -1 X T L

[0034] In the formula, A is the conversion matrix from the reference image matching point to the original image, X is the matrix composed of the reference image matching points, L is the matrix composed of the original image corresponding points, and the construction method of each matrix is as follows:

[0035] A = [a0 a1 a2... a 10 a 11 ] T

[0036]

[0037] L = [x'1 y'1... x'N y'N] k y' k ] T

[0038] In the formula, (x k , y k ) is the kth feature point on the reference image; (x' k , y' k ) is the corresponding point matched by the kth point of the reference image in the original image, and a0-a11 represents the conversion coefficients constituting the conversion matrix A.

[0039] After obtaining the conversion matrix A, the Euclidean distance between the homonymic point coordinates obtained based on the conversion matrix and the homonymic point coordinates matched by each point is calculated, and on this basis, the root mean square error of all distances is calculated. The points with a distance greater than m times the root mean square error are regarded as mismatched points and are eliminated.

[0040] Further, the calculation formula of the root mean square error R is as follows:

[0041]

[0042] In the formula, (x i0 , y i0 ) is the coordinate of the feature point in the reference image; d i is the Euclidean distance between the homonymic point coordinates obtained based on the conversion matrix and the homonymic point coordinates matched by the ith feature point of the reference image; (x, y) and (x', y') represent the homonymic point coordinates obtained based on the conversion matrix and the homonymic point coordinates matched by the ith feature point of the reference image, respectively.

[0043] determination d i The size relationship with m x R is determined, if d i ≥ mR, the i-th point is considered as a mismatch point, otherwise it is reserved; after judging all the feature points, the elimination of the mismatch points is completed.

[0044] Further, the specific implementation mode of step 4 is as follows:

[0045] The point position screening is performed on each image block, for a single image block, the center coordinates (C x , C y ) of the image block are calculated first, and the calculation formula of the center coordinates is as follows:

[0046] Max x = max (x1', x2', x3',..., x n ')

[0047] Min x = min (x1', x2', x3',..., x n ')

[0048] Max y = max (y1', y2', y3',..., y n ')

[0049] Min y = min (y1', y2', y3',..., y n ')

[0050]

[0051] In the formula, (x' i , y' i ) is the i-th coordinate of a single image block in the reference image, and n represents the number of matching points in the image block.

[0052] After obtaining the center point coordinates, the Euclidean distance between the matching points and (C x , C y ) is calculated, and the point with the minimum Euclidean distance is output as the final matching point of the image block.

[0053] The above operation is performed on each image block to obtain the matching points distributed uniformly in the image.

[0054] The application also provides a large-width satellite-borne remote sensing image matching system, comprising the following modules:

[0055] an image block module for dividing the image based on the image side, i.e., dividing the image based on the image side coordinates of the original image, and extracting the image block on the reference image based on geographic information;

[0056] an image block matching module for matching the extracted image block by using a matching method, and preliminarily obtaining the homonym points between the image blocks;

[0057] an error matching point elimination module for eliminating the error matching points in the matching result by using a local-global error matching point elimination strategy;

[0058] a local RANSAC error matching point elimination based on an affine model and a global error matching point elimination based on a quadratic polynomial residual fitting method;

[0059] a point position aggregation output module for selectively outputting the points in the matching result, i.e., selecting only one point for each image block participating in the matching, so as to ensure the uniformity of the distribution of the whole scene image matching result.

[0060] The innovation of the present application lies in that: (1) a set of matching method for remote sensing images is designed, and based on the method, the matching algorithm in the field of computer vision can be easily applied to remote sensing images.

[0061] (2) a set of error matching point elimination method for remote sensing images is designed, i.e., a local image block RANSAC error matching point elimination based on an affine transformation model + a secondary error matching point elimination based on geographic position information, and based on the strategy, the reliability of the output result can be effectively improved.

[0062] (3) a point position aggregation strategy is designed, which ensures the matching efficiency of remote sensing images while ensuring the uniformity of the distribution of the matching homonym points of remote sensing images. BRIEF DESCRIPTION OF DRAWINGS

[0063] Figure 1 is a flowchart of the embodiment of the present application.

[0064] Figure 2 is a schematic diagram of the homonym points matched in the GF7 satellite image data based on the embodiment of the present application. DETAILED DESCRIPTION

[0065] The specific embodiment of the present application will be described in detail below in combination with the drawings and embodiments.

[0066] The width of a spaceborne remote sensing image is wide, and the length and width of the image are up to ten thousand, and all the matching algorithms are designed for small image blocks, and there are few matching algorithms for such a large amount of data, so how to apply these algorithms to spaceborne remote sensing images is a problem worth solving. In view of the above problem, the application constructs a matching method and system for spaceborne remote sensing images, and the framework realizes the matching of spaceborne remote sensing images through four parts of image blocking, local image block matching, false matching point elimination and matching result output.

[0067] Referring to Figure 1 , the flow of the embodiment can be divided into the following steps, and the specific method, formula and flow of each step are as follows:

[0068] Step 1, the image is blocked based on the image, that is, the image is blocked based on the image coordinates of the original image, and the image block is extracted on the reference image based on geographic information. If the width of the original image is w, the length is h, the number of image blocks in the width direction is n x , and the number of image blocks in the height direction is ny, assuming that the size of the image block participating in matching is s, then the calculation method of the coordinate (x, y) of the sampling point is as follows:

[0069]

[0070] x i，j =S x +i*I x

[0071] y i，j =S y +j*I y

[0072] After obtaining the above image coordinates (x i，j , y i，j ), they are converted into object coordinates (Lon i，j , Lat i，j ), so that the coordinate system is unified to the object coordinate system, and then taking the series of coordinates as the center and s as the side length, the sampling block can be obtained, which lays a foundation for subsequent extraction of image blocks participating in matching, wherein (S x , S y ) are the coordinates of the initial sampling, which are generally taken i and j are the serial numbers of horizontal sampling and vertical sampling (i≤n x , j≤n y ).

[0073] Step 2, match the extracted image blocks using a matching method, and preliminarily obtain the homonymic points between the image blocks. This part mainly includes two steps. The first step is to eliminate the angle and resolution differences between the image blocks. The second step is to obtain the homonymic points of the image blocks after eliminating the angle and resolution differences by using a matching algorithm. The specific steps are as follows:

[0074] Step 21, eliminate the rotation and resolution differences between the images. First, calculate the minimum circumscribed rectangle of the intersection area of the two images to obtain the four object coordinates A, B, C, and D of the minimum circumscribed rectangle. Then, calculate the corresponding image coordinates A1, B1, C1, and D1 of the four object coordinates of the minimum circumscribed rectangle on the original image, and the corresponding four image coordinates A2, B2, C2, and D2 on the reference image. In order to make the calculated transformation matrix have the image block center as the origin, subtract the average of the four corner points of the corresponding image from the above coordinates, so as to obtain the transformation matrix H of the original image block to the reference image block with the image block center as the origin. In the subsequent matching process, the original image block is transformed using the transformation matrix H to obtain an image block with consistent resolution and angle with the reference image block.

[0075] Step 22, after obtaining the image blocks with consistent angle and resolution, match the image blocks using a matching method. The matching method can be mainly divided into feature matching and template matching. Among them, template matching includes template extraction, template similarity calculation, and the center of the most similar template and the center of the image to be matched as a pair of matching points; feature matching includes feature point extraction, feature point description, and feature point association, wherein feature point association is to judge according to the similarity between the feature point descriptors, and the most similar two points are a pair of matching points. Thus, the homonymic points between the local image blocks are matched.

[0076] Step 3, local to global false matching point elimination. After the above method, the homonymic points between different image blocks are obtained, but there are inevitably false matching points between these homonymic points. Therefore, a local + global false matching point elimination strategy is used to eliminate the false matching points in the matching results, and the elimination method is as follows.

[0077] Step 31, RANSAC false matching point elimination based on affine model. Although the width of the spaceborne remote sensing image is wide, the transformation relationship of the local part of the two images is relatively simple, and the RANSAC method based on affine model can be used to eliminate the false matching points. The specific steps include the following:

[0078] ① Randomly select three pairs of homonymic points from the matching result point sets X and Y to solve the affine transformation matrix H based on the least squares method. The solving method of the transformation matrix is as follows:

[0079] Randomly select three pairs of homonymic points from the matching result point sets X and Y to solve the affine transformation matrix H based on the least squares method. The solving method of the transformation matrix is as follows:

[0080] Firstly, the matching point pairs (x i , y i ) and (u i , v i ) are expressed in homogeneous coordinates and represented in matrix form as follows

[0081]

[0082] where X and Y represent the local matching point sets in the original image and the reference image respectively, and n represents the number of matching pairs.

[0083] Then, the least square solution is calculated according to the formula H = Y T X(X T X) -1 to obtain the conversion matrix H.

[0084] ②Predicting the matching point position

[0085] The affine transformation matrix H obtained by calculation is used to map all the matching points X to obtain the predicted matching point position X'.

[0086] ③Calculating the number of inliers

[0087] The Euclidean distance E between the predicted matching point position X' and the actual matching point X is calculated. If E is less than the threshold T (set to 3 in this algorithm), it is considered as an inlier, and if E is greater than the threshold T, it is considered as an outlier, and the total number of inliers contained in X is counted.

[0088] ④Repeat steps ① to ③, and take the data set with the largest number of inliers as the best data set, and the point pairs contained in this data set are the correctly matched homonymous points. In order to balance the robustness and efficiency of the algorithm, the number of loops is set to 5000 in this framework.

[0089] Through the above operations, the elimination of the mis-matching points in the local image block is realized.

[0090] Step 32, mis-matching point elimination method based on global quadratic polynomial residual fitting

[0091] After the local false matching points are removed, the accuracy of the matching results is greatly improved, but there are still some significant false matching points. In addition, due to the influence of the RANSAC random sampling mechanism, when the inlier rate in the local image block is low, the selected points are all false matching points. In order to reduce this situation as much as possible, we further use a global false matching point removal method based on quadratic polynomial residual fitting to remove the significant outliers, so as to improve the accuracy of the final matching results. Specifically, since the width of the satellite remote sensing image is wide, it is difficult to fit the transformation relationship between the images by using a simple affine model, so we use a quadratic polynomial to fit the transformation relationship between the images, and based on the fitting result, we remove the outliers in the matching results. The fitting formula of the quadratic polynomial is as follows:

[0092] A=(X T X) -1 X T L

[0093] In the formula, A is the conversion matrix from the reference image matching point to the original image, X is the matrix composed of the reference image matching points, L is the matrix composed of the corresponding points of the original image, and the construction method of each matrix is as follows:

[0094] A=[a0 a1 a2... a 10 a 11 ] T

[0095]

[0096] L=[x′1 y′1... x′ k y′ k ] T

[0097] In the formula, (x k , y k ) is the kth feature point on the reference image; (x′ k , y′ k ) is the corresponding point of the kth point on the reference image matched in the original image, and a0-a11 represents the conversion coefficients of the conversion matrix A.

[0098] After obtaining the conversion matrix A, the Euclidean distance between the coordinates of the same name point based on the conversion matrix and the coordinates of the same name point matched is calculated, and on this basis, the root mean square error of all distances is calculated. The points with a distance greater than 3 times (which can also be set according to specific needs) the root mean square error are removed as false matching points. The calculation formula of the root mean square error R is as follows:

[0099]

[0100] In the formula (x i0 y i0 ) represents the coordinates of the feature points in the reference image; d i Let (x, y) be the Euclidean distance between the coordinates of the original image corresponding point obtained from the transformation matrix of the i-th feature point of the reference image and the coordinates of the matched corresponding point; (x, y) and (x′, y′) represent the coordinates of the original image corresponding point obtained from the transformation matrix of the i-th feature point of the reference image and the coordinates of the matched corresponding point, respectively.

[0101] Determine d i The relationship between d and 3×R, if d i If the feature point is greater than or equal to 3R, then the i-th point is considered a mismatch; otherwise, it is retained. After judging all feature points in turn, the removal of mismatches is completed.

[0102] Step 4: Selectively output points from the matching results. Matching point filtering mainly refers to selecting only one point from each participating image block to ensure the uniformity of the matching results across the entire image. Specifically, this involves filtering points for each image block. Taking a single image block as an example, we first calculate the center coordinates (C...) of the image block. x C y The formula for calculating the center coordinates is as follows:

[0103] Max x =max(x1',x2',x3',...,x n ')

[0104] Min x =min(x1', x2', x3', ..., x n ')

[0105] Max y =max(y1',y2',y3',...,y n ')

[0106] Min y =min(y1',y2',y3',...,y n ')

[0107]

[0108] In the formula, (x' i y' i ) is the i-th coordinate of a single image block in the reference image, and n represents the number of matching points in the image block.

[0109] After obtaining the coordinates of the center point, we calculate the matching points and (C) one by one. x C y) as the final matching point of the image block. The above operation is sequentially performed on each image block to obtain matching points uniformly distributed in the image. Experiments show that the front and rear view image matching of GF7 satellite has a total of 164 feature points, and the matching result takes 5.393 seconds, as shown in the accompanying Figure 2 figure. As shown in the figure, the matching points are uniformly distributed in the whole image, which can meet the demand of most spaceborne remote sensing image matching.

[0110] The advantages of the present application are that the spaceborne remote sensing image is sampled and blocked to reduce the calculation amount in the matching process, and the local plus global false matching point elimination method is used to effectively eliminate false matching points in the spaceborne remote sensing image with large distortion. Finally, the matching point selection strategy effectively ensures the uniformity of the distribution of the image matching points.

[0111] On the other hand, the embodiment of the present application also provides a wide spaceborne remote sensing image matching system, comprising the following modules:

[0112] The image blocking module is used for blocking the image based on the image side, that is, blocking the image based on the image side coordinates of the original image, and extracting the image block on the reference image based on geographic information;

[0113] The image block matching module is used for matching the extracted image block by using a matching method, and preliminarily obtaining the homonymic points between the image blocks;

[0114] The false matching point elimination module is used for eliminating the false matching points in the matching result by using a local plus global false matching point elimination strategy;

[0115] The local false matching point elimination is based on the affine model RANSAC, and the global false matching point elimination is based on the quadratic polynomial residual fitting method;

[0116] The point position aggregation output module is used for selectively outputting the points in the matching result, that is, selecting only one point for each image block participating in the matching, so as to ensure the uniformity of the distribution of the whole scene image matching result.

[0117] The specific implementation modes of the modules are the same as those of the steps, and the present application will not be described.

[0118] The specific embodiments described herein are only illustrative of the spirit of the present application. Those skilled in the art of the present application can make various modifications or supplements to the described specific embodiments or replace them with similar ways, without deviating from the spirit of the present application or exceeding the scope defined by the appended claims.

Claims

1. A method for matching large-width spaceborne remote sensing images, characterized in that, The method comprises the following steps: Step 1, image block based on the image side, that is, based on the original image side coordinates of the image is divided into blocks, and the geographical information is extracted on the reference image based on the image block; Step 2, the matching method is used to match the image block, and the homonym point between the image blocks is obtained; Step 3, the error matching point in the matching result is removed by using the local and global error matching point removal strategy; Wherein the local RANSAC error matching point removal based on affine model, the global error matching point removal method based on quadratic polynomial residual fitting; Step 4, the point in the matching result is selectively output, that is, only one point is selected from each image block participating in the matching, so as to ensure the uniformity of the whole scene image matching result.

2. The method of claim 1, wherein the wide swath spaceborne remote sensing image matching method is characterized by: In step 1, if the width of the original image is w, the length is h, the number of image blocks in the width direction is n x , and the number of image blocks in the height direction is n y , assuming that the size of the image block participating in matching is s, the calculation method of the coordinates (x, y) of the sampling point is as follows: x i,j = S x + I*I x y i,j = S y + j * I y Among them, (S) x ,S y The initial sampling coordinates, where i and j are the indices of the horizontal and vertical sampling directions, i ≤ n. x ,j≤n y After obtaining the above image coordinates (x... i,j y i,j After that, convert it to object coordinates (Lon). i,j Lat i,j This shifts the coordinate system to the object coordinate system. Then, with the object coordinate system as the center and s as the side length, the sampling block can be obtained.

3. The method of claim 1, wherein the wide swath spaceborne remote sensing image matching method is characterized by: The specific implementation of step 2 includes two steps, the first step is to eliminate the angle and resolution difference between the image blocks; the second step is to obtain the homonym point of the image block after eliminating the angle and resolution difference by using the matching algorithm; it specifically includes the following sub steps: Step 21, image rotation resolution difference elimination: first, the minimum circumscribed rectangle of the intersection area of the two images is calculated to obtain the four object side coordinates A, B, C and D of the minimum circumscribed rectangle, then the corresponding image side coordinates A1, B1, C1 and D1 of the four object side coordinates of the minimum circumscribed rectangle on the original image and the corresponding four image side coordinates A2, B2, C2 and D2 on the reference image are calculated, the image is converted with the image block center as the origin, and the above coordinates are all subtracted by the average value of the four corner points of the corresponding image, so as to obtain the conversion matrix H of the original image block to the reference image block with the image block center as the origin; in the subsequent matching process, the conversion matrix H is used to convert the original image block to obtain the image block consistent with the resolution and angle of the reference image; Step 22, after obtaining the image block with consistent angle and resolution, the image block is matched by using the matching method.

4. The method of claim 3, wherein the wide swath spaceborne remote sensing image matching method is characterized by: The matching method includes feature matching and template matching; wherein, the template matching includes template extraction, template similarity calculation, the center of the most similar template and the center of the image to be matched, that is, a pair of matching points; the feature matching includes feature point extraction, feature point description and feature point association, wherein the feature point association is judged according to the similarity between the feature point descriptors, and the most similar two points are a pair of matching points.

5. The method of claim 1, wherein: The specific implementation mode of the local RANSAC error matching point removal based on affine model is as follows: ①Randomly select points to solve the affine transformation matrix Randomly select three pairs of homonym points from the matching result point set X and Y in step 3, solve the affine transformation matrix H based on the least square method, and the solving method of the transformation matrix is as follows: First, the matching point pairs (x i ,y i ) and (u i ,v i ) are expressed in homogeneous coordinates and in matrix form as follows: Wherein, X and Y represent the local matching point set in the original image and the reference image respectively, and n represents the number of matching pairs; Then according to the formula H=Y T X(X T X) -1 The least square solution is calculated to obtain the conversion matrix H; ②Predict the matching point position The calculated affine transformation matrix H is used to map all the matching points X to obtain the predicted matching point position X'; ③Calculate the number of inliers Calculate the Euclidean distance E between the predicted matching point position X' and the actual matching point X, if E is less than the threshold T, it is considered as an inlier, and if E is greater than the threshold T, it is considered as an outlier, and the total number of inliers contained in X is counted; ④ Repeat steps ① to ③, and take the data set with the most inner points as the best data set, and the points in the data set are the locally correctly matched homonymic points.

6. The method of claim 1, wherein: The specific implementation of the global residual fitting method based on a quadratic polynomial for removing mismatched points is as follows: The fitting formula of the quadratic polynomial is as follows: A = (X T X) -1 X T L In the formula, A is a conversion matrix from the reference image matching points to the original image, X is a matrix composed of the reference image matching points, L is a matrix composed of the original image corresponding points, and the construction method of each matrix is as follows: A = [a0a1a2...an]T 10 a 11 ] T L = [x'1 y'1... x'N y'N]T k y' k ] T where (x k ,y k ) is the kth feature point on the reference image; (x′ k ,y′ k ) is the kth point on the reference image matched to the corresponding point on the original image; and a0-a 11 represent the conversion coefficients constituting the conversion matrix A. After obtaining the conversion matrix A, the Euclidean distance between the coordinates of the homonymic points obtained based on the conversion matrix and the coordinates of the matched homonymic points is calculated, and on this basis, the root mean square error of all distances is calculated, and the points with a distance greater than m times the root mean square error are removed as mismatched points.

7. The method of claim 6, wherein: The calculation formula of the root mean square error R is as follows: where (x,y) is the coordinate of the feature point in the reference image; d i0 is the Euclidean distance between the original image homonymic point coordinate of the i i0 th feature point in the reference image and the matched homonymic point coordinate based on the conversion matrix; and (x,y) and (x′,y′) represent the homonymic point coordinate of the i i th feature point in the reference image based on the conversion matrix and the matched homonymic point coordinate, respectively. determination d i In relation to the size of m x R, if d i If d If d If d 8. The method of claim 1, wherein the wide swath spaceborne remote sensing image matching method is characterized by: The specific implementation of step 4 is as follows: The point position screening is performed on the image blocks one by one. For a single image block, the center coordinate (C x ,C y ) of the image block is calculated first. The calculation formula of the center coordinate is as follows: Max x = max(x1’, x2’, x3’, …, x n ’) Min x = min(x1’,x2’,x3’,…,x n ’) Max y = max(y1', y2', y3',..., y n ') Min y = min(y1’, y2’, y3’, …, y n ’) where (x i ,y i ) is the i-th coordinate of a single image block in the reference image, and n represents the number of matching points in the image block. After the coordinates of the center point are obtained, the Euclidean distances of the matching points and (C x ,C y ) are calculated one by one, and the point with the minimum Euclidean distance is output as the final matching point of the image block. The above operations are sequentially performed on each image block to obtain matching points uniformly distributed in the image.

9. A wide swath spaceborne remote sensing image matching system, characterized in that, The system comprises the following modules: An image blocking module, which is configured to block the image based on the image plane, i.e., block the image based on the image plane coordinates of the original image, and extract image blocks on the reference image based on geographic information; An image block matching module, which is configured to match the extracted image blocks by using a matching method, and preliminarily obtain homonymic points between the image blocks; An error matching point removal module, which is configured to remove error matching points in the matching result by using a local plus global error matching point removal strategy; The local error matching point removal strategy is based on an affine model RANSAC, and the global error matching point removal strategy is based on a residual fitting method of a quadratic polynomial; A point position aggregation output module, which is configured to selectively output points in the matching result, i.e., select only one point from each image block participating in the matching, so as to ensure the uniformity of the distribution of the whole scene image matching result.

Images