An evaluation method for the geometric quality availability of optical remote sensing images

Abstract

The present invention discloses a method for evaluating the geometric quality availability of optical remote sensing images. The method specifically includes: parsing optical remote sensing image data to obtain the original multispectral image and the original panchromatic image, and obtaining a basic set of corresponding points through feature point extraction and feature registration; using an optimization strategy to optimize the basic set of corresponding points to obtain a first set of corresponding points; selecting corresponding point pairs from the first set of corresponding points based on geometric importance and distribution balance to obtain a second set of corresponding points; calculating the geometric accuracy for the second set of corresponding points; and performing availability determination based on the geometric accuracy to obtain the geometric quality availability. The present invention can comprehensively characterize the geometric positioning accuracy and geometric consistency of satellite remote sensing images, can quickly and automatically evaluate the geometric quality and availability of images, and supports subsequent image processing and product production.

Description

Technical Field

[0001] The present invention relates to the field of remote sensing image quality evaluation, and particularly to a method for evaluating the geometric quality availability of optical remote sensing images. Background Art

[0002] Due to the characteristics of high resolution and strong currency, satellite remote sensing data has been widely applied to fields such as resource investigation and environmental monitoring, and has become an important data source that cannot be obtained. However, due to the influence of a complex imaging environment, initial satellite remote sensing data may have geometric problems such as geometric misalignment of different spectral images. In the traditional satellite image processing process, geometric quality problems of different spectral images are usually found only during image fusion, making subsequent processing and application very passive. Therefore, it is necessary to determine whether the image data is available based on geometric quality evaluation when the original image is acquired.

[0003] Traditional satellite image geometric quality evaluation usually manually selects a certain number of uniformly distributed checkpoints on the image, and characterizes it by calculating the coordinate error between its coordinates on the image and the coordinates of its true geographical location, and statistically calculating the mean square error. This method can usually well characterize the geometric positioning accuracy or geometric registration accuracy of the image, but cannot characterize the geometric consistency within the image; at the same time, this method is manual interaction, with low automation and does not intuitively reflect the availability of the image. Summary of the Invention

[0004] Aiming at the problems existing in the current remote sensing image geometric quality evaluation method, such as incomplete geometric quality characterization, low automation level, and insufficient expression of image availability, the present invention proposes a method for evaluating the geometric quality availability of optical remote sensing images, which directly performs automated and comprehensive geometric quality evaluation on the original image data to obtain the availability evaluation result of the image data.

[0005] The present invention provides a method for evaluating the geometric quality availability of optical remote sensing images, and the method includes:

[0006] S1: Obtain optical remote sensing image data;

[0007] S2: Analyze the optical remote sensing image data to obtain the original multispectral image I1 and the original panchromatic image I2;

[0008] S3: Respectively obtain the feature points of I1 and I2, and perform feature registration on the feature points of I1 and I2 to form a plurality of corresponding point pairs, and all the corresponding point pairs form a basic corresponding point set;

[0009] S4: Use an optimization strategy to optimize the basic corresponding point set with the corresponding point pairs as units to obtain a first corresponding point set;

[0010] S5: Select corresponding point pairs for the first set of corresponding points based on geometric importance and distribution balance to obtain a second set of corresponding points;

[0011] S6: Calculate the geometric accuracy for the second set of corresponding points;

[0012] S7: Determine the usability of the optical remote sensing image data based on the geometric accuracy to obtain the geometric quality usability of the optical remote sensing image data.

[0013] In a specific embodiment of the present invention, step S4 includes:

[0014] S41: Select a corresponding point pair MP i from the basic set of corresponding points, where the MP i includes a feature point f i of I1 and a feature point g i of I2, where i is an integer greater than or equal to 1;

[0015] S42: Calculate the Euclidean distance between the feature points in I1 and f i , sort the results of the Euclidean distance in ascending order, select the first n feature points, and construct n - 1 triangles with f i as the fixed vertex and the n feature points respectively according to the proximity principle. The n - 1 triangles form the first geometric constraint set

[0016] S43: Calculate the Euclidean distance between the feature points in I2 and g i , sort the results of the Euclidean distance in ascending order, select the first n feature points, and construct n - 1 triangles with g i as the fixed vertex and the n feature points respectively according to the proximity principle. The n - 1 triangles form the second geometric constraint set

[0017] S44: Calculate the similarity of the corresponding triangles in and respectively to obtain n - 1 similarity results. Determine the similarity results according to the discrimination condition. If the similarity results meet the discrimination condition, add f i and g i to the first set of corresponding points, where the discrimination condition is that at least d of the n - 1 similarity results are greater than the first threshold;

[0018] S45: Repeat steps S41 to S44 to traverse all corresponding point pairs in the basic set of corresponding points to obtain the first set of corresponding points.

[0019] In a specific embodiment of the present invention, step S5 includes:

[0020] S51: Select m corresponding points within the corresponding points belonging to I1 in the first set of corresponding points based on geometric importance, and obtain m pairs of corresponding points from the first set of corresponding points according to the m corresponding points, and add the m pairs of corresponding points to the second set of corresponding points, where the m corresponding points are obtained by uniformly selecting in the vertical track direction, track direction, and diagonal direction of I1;

[0021] S52: Use the remaining pairs of corresponding points after removing the m pairs of corresponding points from the first set of corresponding points as the first remaining point set, and select k pairs of corresponding points from the first remaining point set based on distribution balance, where k > 4:

[0022] S521: Obtain four corner points of I1, and select four corresponding points from the first remaining point set according to the principle of proximity to the corner points, and add the corresponding pairs of corresponding points of the four corresponding points to the second set of corresponding points;

[0023] S522: Use the remaining pairs of corresponding points after removing the four pairs of corresponding points from the first remaining point set as the second remaining point set, cluster the corresponding points belonging to I1 in the second remaining point set according to the clustering algorithm to obtain k - 4 clustering centers, and add the pairs of corresponding points corresponding to the k - 4 clustering centers to the second set of corresponding points.

[0024] In a specific embodiment of the present invention, the geometric accuracy is calculated according to the geometric registration accuracy and geometric deformation accuracy.

[0025] In a specific embodiment of the present invention, the calculation method of the geometric deformation accuracy is:

[0026] Step 1: Establish a Delaunay triangulation De1 according to the corresponding points belonging to I1 in the second set of corresponding points;

[0027] Step 2: Obtain the corresponding points e i 1 of the triangle e i 1,1 , e i 1,2 , e i 1,3 in De1, and use the corresponding relationship to find the corresponding points e i 1,1 , e i 1,2 , e i 1,3 of the corresponding points on I2 as e i 2,1 , e i 2,2 , e i 2,3 , and use ei 2,1 , e i 2,2 , e i 2,3 Construct a triangle e i 2;

[0028] Step 3: Repeat Step 2 to traverse all the triangles in De1 and construct the Delaunay triangulation De2;

[0029] Step 4: Calculate the area change value between each pair of corresponding triangles in De1 and De2, and calculate the area standard deviation D based on the area change value area ;

[0030] Step 5: Calculate the angle change value between each pair of corresponding triangles in De1 and De2, and calculate the angle standard deviation D based on the angle change value angle ;

[0031] Step 6: Perform calculations on D area and D angle Based on the operation method to calculate the geometric deformation accuracy.

[0032] In a specific embodiment of the present invention, the calculation method of the geometric registration accuracy is as follows:

[0033] Construct a transformation matrix using the corresponding points in the second set of corresponding points;

[0034] Using the transformation matrix, transform the corresponding points belonging to I1 in the second set of corresponding points into the space of I2 to obtain the transformed corresponding points;

[0035] Calculate the distance between all the transformed corresponding points and the corresponding points belonging to I2 in the second set of corresponding points, and calculate the average value of the distances to obtain the geometric registration accuracy.

[0036] In a specific embodiment of the present invention, the operation method includes weighted calculation, multiplication calculation, and addition calculation.

[0037] In a specific embodiment of the present invention, the calculation method of "calculating according to the geometric registration accuracy and geometric deformation accuracy" includes weighted calculation, multiplication calculation, and addition calculation.

[0038] In a specific embodiment of the present invention, Step S7 includes:

[0039] S71: Normalize the geometric accuracy to a value within the range of 0 to 100;

[0040] S72: Set the usability threshold. When the value after normalizing the geometric accuracy is greater than the usability threshold, the geometric quality usability of the optical remote sensing image is 100%. Otherwise, the geometric quality usability of the optical remote sensing image is 0%.

[0041] In a specific embodiment of the present invention, step S3 includes:

[0042] S31: Extract the feature points of I1 based on the scale-invariant feature transform algorithm to obtain the first feature point set;

[0043] S32: Extract the feature points of I2 based on the scale-invariant feature transform algorithm to obtain the second feature point set;

[0044] S33: Perform feature point matching on the first feature point set and the second feature point set, and take the two successfully matched feature points as a homologous point pair. All the homologous point pairs form the basic homologous point set.

[0045] The beneficial effects of the present invention are as follows: The present invention discloses a method for evaluating the geometric quality usability of optical remote sensing images. In the first set of homologous points obtained through the optimization strategy of multi-triangle geometric constraints, not only whether the domain points are homologous points is considered, but also the geometric consistency between the domain points and the feature points is considered, which can quickly and accurately eliminate the incorrect homologous points and obtain the preferred homologous points. The geometric importance and distribution balance strategy selects homologous points from the optimized first set of homologous points to obtain the second set of homologous points, which can ensure that the selected homologous points have geometric importance, feature importance, and distribution balance, and can better evaluate the geometric quality of the image. The geometric accuracy is calculated through the geometric registration accuracy and the geometric deformation accuracy, and by constructing a Delaunay triangulation network, the standard variance of area change and the standard variance of angle transformation are used to quickly evaluate the geometric deformation accuracy. The present invention can quickly and automatically evaluate the geometric quality and usability of images, and support subsequent effective image processing and product production. Description of the Drawings

[0046] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0047] Figure 1 It is a schematic flowchart of the method according to the embodiment of the present invention. Detailed Embodiments

[0048] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. It should be noted that as long as there is no conflict, each embodiment in the present invention and each feature of each embodiment can be combined with each other, and the formed technical solutions are all within the protection scope of the present invention.

[0049] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present invention. As used herein, unless otherwise clearly specified in the context, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or combinations thereof.

[0050] Figure 1 The figure shows a flowchart of a method for evaluating the geometric quality availability of an optical remote sensing image provided by the present invention, including steps S1-S7.

[0051] S1: Obtain optical remote sensing image data.

[0052] Optical remote sensing image data is generally provided to users in the form of a tar package after rough processing, and its content includes image data of different bands, RPC files, thumbnails, etc.

[0053] S2: Analyze the optical remote sensing image data to obtain the original multispectral image I1 and the original panchromatic image I2.

[0054] Affected by the imaging sensor, the band image data included in the original optical remote sensing image data under different imaging sensors is different. Therefore, it is necessary to analyze the original remote sensing image data.

[0055] After analysis, the original multispectral image and the original panchromatic image are obtained. The original multispectral image is defined as I1, and the original panchromatic image is defined as I2.

[0056] S3: Respectively obtain the feature points of I1 and I2, and perform feature registration on the feature points of I1 and I2 to form a plurality of corresponding point pairs, and all the corresponding point pairs form a basic corresponding point set.

[0057] In order to evaluate the geometric quality of I1 and I2, first, the geometric similarity of the two scenes of images needs to be described at the feature level, and then the calculation and evaluation of the geometric registration accuracy and geometric deformation accuracy are carried out. In this example, the geometric similarity of I1 and I2 is described by corresponding points.

[0058] S31: Extract the feature points of I1 based on the Scale-Invariant Feature Transform (SIFT) algorithm to obtain the first set of feature points.

[0059] S32: Extract the feature points of I2 based on the Scale-Invariant Feature Transform (SIFT) algorithm to obtain the second set of feature points.

[0060] Specifically, in this example, the SIFT feature extraction method is used to extract the feature points of I1 and I2. The SIFT feature extraction method not only maintains invariance to rotation and brightness changes but also to scale scaling. Therefore, the feature points of I1 and I2 at different scales can be extracted, and the feature points at different scales can be feature-registered through feature point descriptors.

[0061] S33: Perform feature point matching on the first set of feature points and the second set of feature points, and take the two successfully matched feature points as a pair of homologous points. All pairs of homologous points form the basic set of homologous points.

[0062] The following specific example is used to illustrate step S33:

[0063] S331: For any feature point f1 in the first set of feature points, find the feature point g1 with the closest Euclidean distance to f1 and the second-closest feature point g2 in the second set of feature points. If the quotient of the distance between g1 and f1 and the distance between g2 and f1 is less than the threshold α1, then f1 and g1 are a pair of successfully matched homologous points.

[0064] Specifically, in this example, α1 is 0.4.

[0065] S332: Execute step S331 for all feature points in the first set of feature points to obtain the basic set of homologous points.

[0066] S4: Use an optimization strategy to optimize the basic set of homologous points in units of the pairs of homologous points to obtain the first set of homologous points.

[0067] Specifically, step S4 includes:

[0068] S41: Select a pair of homologous points MP i , the MP i includes the feature point f i of I1 and the feature point g i of I2, where i is an integer greater than or equal to 1.

[0069] S42: Calculate the Euclidean distance between the feature points in I1 and f i , sort the results of the Euclidean distance in ascending order, select the first n feature points, and use f iFor the fixed vertex, n - 1 triangles are constructed with n feature points according to the proximity principle, and these n - 1 triangles form the first geometric constraint set

[0070] In this example, taking f i as the fixed vertex, with the horizontal right direction of f i being 0 degrees to obtain the 0 - degree starting line f i -0, sort the n feature points in the counter - clockwise direction of the 0 - degree starting line f i -0, and construct triangles according to the proximity principle to form n - 1 triangles:

[0071] (1) Connect the j - th feature point with f i to get the connection line j - f i , where 1 ≤ j ≤ n.

[0072] (2) Obtain the included angle △j - f i -0 based on f i -0 and j - f i -0.

[0073] (3) Repeat steps (1) and (2), calculate the included angles of all n feature points with f i -0 in the counter - clockwise direction, and sort them in ascending order.

[0074] (4) Taking f i as the fixed vertex, construct triangles according to the sorting order and the proximity principle to form n - 1 triangles, and these n - 1 triangles form the first geometric constraint set

[0075] S43: Calculate the Euclidean distances between the feature points in I2 and g i , sort the results of the Euclidean distances in ascending order, select the first n feature points, and construct n - 1 triangles with g i as the fixed vertex and the n feature points according to the proximity principle, and these n - 1 triangles form the second geometric constraint set

[0076] In this example, taking g i as the fixed vertex, with the horizontal right direction of g i being 0 degrees to obtain the 0 - degree starting line g i -0, sort the n feature points in the counter - clockwise direction of the 0 - degree starting line g i -0, and construct triangles according to the proximity principle to form n - 1 triangles:

[0077] (1) Connect the j - th feature point with g i to get the connection line j - gi , where \(1\leq j\leq n\).

[0078] (2) According to \(g\) i -0 and \(j - g\) i obtain the included angle \(\triangle j - g\) i -0.

[0079] (3) Repeat steps (1) and (2), calculate the included angles between all \(n\) feature points and \(g\) i -0 in counterclockwise order, and sort them in ascending order.

[0080] (3) Using \(g\) i as the fixed vertex, construct triangles according to the sorted order and the proximity principle, forming \(n - 1\) triangles, and these \(n - 1\) triangles form the second geometric constraint set

[0081] It should be noted that the proximity principle means that the two most adjacent feature points in the sorted order are successively used to construct triangles with the fixed vertex, that is, among the \(n - 1\) formed triangles, the areas of each triangle are non - repeating.

[0082] S44: Calculate the similarities of the corresponding triangles in and respectively, obtain \(n - 1\) similarity results, and judge the similarity results according to the discrimination condition. If the similarity results meet the discrimination condition, then add \(f\) i and \(g\) i to the first set of same - name points, where the discrimination condition is that at least \(d\) of the \(n - 1\) similarity results are greater than the first threshold.

[0083] S45: Repeat steps S41 to S44, traverse all pairs of same - name points in the basic set of same - name points, and obtain the first set of same - name points.

[0084] Specifically, in this example, \(n = 5\), \(d = 3\), and the first threshold is \(0.9\).

[0085] The present invention can not only screen out wrong pairs of same - name points through the feature - point pairing rate by constructing triangle geometric constraints, but also screen out wrong pairs of same - name points according to geometric similarities such as angles and scales, and more efficiently eliminate wrong pairs of same - name points to obtain an optimized set of same - name points.

[0086] S5: Select pairs of same - name points from the first set of same - name points based on geometric importance and distribution balance to obtain the second set of same - name points.

[0087] The specific implementation steps of S5 include:

[0088] S51: Select m corresponding points within the corresponding points belonging to I1 in the first set of corresponding points based on geometric importance, and obtain m pairs of corresponding points from the first set of corresponding points according to the m corresponding points, and add the m pairs of corresponding points to the second set of corresponding points, where the m corresponding points are evenly selected according to the vertical track direction, track direction, and diagonal direction of I1.

[0089] Specifically, in a specific embodiment of the present invention:

[0090] (1) When evenly selecting m points according to the vertical track direction, track direction, and diagonal direction of I1, first, the center point of the I1 image can be determined according to the intersection point of the diagonals, and then the central corresponding point closest to the center point is obtained according to the nearest distance principle, and the central corresponding point belongs to the m corresponding points.

[0091] (2) Based on the central corresponding point, evenly select m - 1 corresponding points in the vertical track direction, track direction, and diagonal direction.

[0092] (3) Obtain m pairs of corresponding points from the first set of corresponding points according to the m corresponding points, and add the m pairs of corresponding points to the second set of corresponding points.

[0093] S52: Use the remaining pairs of corresponding points in the first set of corresponding points after removing the m pairs of corresponding points as the first remaining set of points, and select k pairs of corresponding points from the first remaining set of points based on distribution balance, where k > 4:

[0094] S521: Obtain the four corner points of I1, and select four corresponding points in the first remaining set of points according to the corner - proximity principle, and add the corresponding pairs of corresponding points of the four corresponding points to the second set of corresponding points.

[0095] It should be noted that the corner - proximity principle means finding four corresponding points in the first remaining set of points that are respectively the closest to the four corner points.

[0096] S522: Use the remaining pairs of corresponding points in the first remaining set of points after removing the four pairs of corresponding points as the second remaining set of points, cluster the corresponding points belonging to I1 in the second remaining set of points according to the clustering algorithm to obtain k - 4 clustering centers, and add the pairs of corresponding points corresponding to the k - 4 clustering centers to the second set of corresponding points.

[0097] Specifically, in this example, k is 15, and the corresponding points of 11 clustering centers can be obtained through the K - means clustering method.

[0098] S6: Calculate the geometric accuracy for the second set of corresponding points.

[0099] Specifically, the geometric accuracy is calculated based on the geometric registration accuracy and the geometric deformation accuracy.

[0100] The geometric registration accuracy characterizes the position accuracy in geometric quality, and the geometric deformation accuracy characterizes the geometric consistency accuracy in geometric quality.

[0101] In this embodiment, the calculation method of the geometric deformation accuracy is as follows:

[0102] Step 1: Establish a Delaunay triangulation De1 based on the corresponding points belonging to I1 in the second set of corresponding points.

[0103] Step 2: Obtain the corresponding points e i 1 of the triangle e i 1,1 in De1, and use the corresponding relationship to find the corresponding points e i 1,2 and e i 1,3 in I2, and use e i 1,1 and e i 1,2 and e i 1,3 to construct the triangle e i 2,1 and e i 2,2 and e i 2,3 to construct the triangle e i 2,1 and e i 2,2 and e i 2,3 to construct the triangle e i 2.

[0104] Step 3: Repeat Step 2 to traverse all the triangles in De1 and construct the Delaunay triangulation De2.

[0105] Step 4: Calculate the area change value between each pair of corresponding triangles in De1 and De2, and calculate the area standard deviation D area .

[0106] (1) Sequentially obtain the triangle e i 1 of De1 and the corresponding triangle e i 2 of De2, where 1 ≤ i ≤ (k + m - 1), m is the number of corresponding points extracted according to the geometric importance strategy, and k is the number of corresponding points extracted according to the distribution balance strategy.

[0107] (2) Calculate the area d i 1 of the triangle ei 1 and triangle e i The area d of 2 i 2, solve for the area change d i 1,2 :

[0108]

[0109] (3) Solve for the representation variance D of the area change area :

[0110]

[0111] Step Five: Calculate the angle change value between each pair of corresponding triangles in De1 and De2, and calculate the angle standard variance D based on the angle change value angle .

[0112] (1) Sequentially obtain the triangle e of De1 i 1, and the corresponding triangle e of De2 i 2, where 1 ≤ i ≤ (k + m - 1), m is the number of homologous points extracted according to the geometric importance strategy, and k is the number of homologous points extracted according to the distribution balance strategy;

[0113] (2) Calculate the average value of the differences between the three angles of triangle e i 1 and the three angles of triangle e i 2:

[0114]

[0115] Among them, a i 1,1 , a i 1,2 , a i 1,3 is the three angles of e i 1, a i 2,1 , a i 2,2 , a i 2,3 is the three angles of e i 2.

[0116] (3) Solve for the angle change standard variance D angle :

[0117]

[0118] Step Six: Calculate the geometric deformation accuracy based on the operation method for D area and D angle ​

[0119] The operation method may be weighted calculation, multiplication calculation, addition calculation or other operation methods.

[0120] Specifically, in this example, a weighted calculation method is adopted, that is:

[0121] D f = a×D area +(1 - a)D angle

[0122] where D f is the geometric deformation accuracy, a is the weight coefficient, and 0 < a < 1.

[0123] The calculation method of the geometric registration accuracy is as follows:

[0124] First, a transformation matrix is constructed using the corresponding points in the second set of corresponding points.

[0125] In this example, the transformation matrix is solved according to the affine transformation equation:

[0126] (1) Obtain the coordinates (x i , y j ) of the corresponding points on the I2 image, and the coordinates (X i , Y j ) of the corresponding points on the I1 image;

[0127] (2) Substitute (x i , y j ) and (X i , Y j ) into the affine transformation equation to solve for the affine transformation parameters A, B, C, D, E, F:

[0128] X = Ax + By + C

[0129] Y = Dx + Ey + F

[0130] By solving for A, B, C, D, E, F, an affine transformation model is obtained.

[0131] After that, the corresponding points belonging to I1 in the second set of corresponding points are transformed into the space of I2 to obtain the transformed corresponding points.

[0132] In this example, for the coordinates (x i , y j ) of the corresponding points on the I2 image, an affine transformation is performed by inputting the affine transformation model to obtain the affine point coordinates (x i ′, y j ′);

[0133] x′ = Ax + By + C

[0134] y′ = Dx + Ey + F

[0135] Finally, calculate the distances between all the transformed corresponding points and the corresponding points belonging to I2 in the second set of corresponding points, and calculate the average value of these distances to obtain the geometric registration accuracy.

[0136] In this example, calculate the Euclidean distance between the affine point coordinates (x i ′, y j ′) and (X i , Y j ) to obtain the corresponding point distance:

[0137]

[0138] where d(i,j) is the Euclidean distance between the affine point coordinates (x i ′, y j ′) and (X i , Y j ).

[0139] After obtaining the geometric registration accuracy and the geometric deformation accuracy, calculate the geometric registration accuracy and the geometric deformation accuracy to obtain the geometric accuracy.

[0140] Among them, the calculation method for obtaining the geometric accuracy can be weighted calculation, multiplication calculation, addition calculation or other calculation methods.

[0141] In this example, the multiplication calculation method is adopted, that is:

[0142] D geo = D f × D regis

[0143] where D geo is the geometric accuracy, D f is the geometric deformation accuracy, and D regis is the geometric registration accuracy.

[0144] S7: Based on the geometric accuracy, determine the usability of the optical remote sensing image data to obtain the geometric quality usability of the optical remote sensing image data.

[0145] S71: Normalize the geometric accuracy to a value in the range of 0 to 100;

[0146] S72: Set a usability threshold. When the normalized value of the geometric accuracy is greater than the usability threshold, the geometric quality usability of the optical remote sensing image is 100%, otherwise the geometric quality usability of the optical remote sensing image is 0%.

[0147] In this embodiment, the usability threshold is set to 75.

[0148] As described above, it is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A method for evaluating the geometric quality availability of optical remote sensing images, characterized in that The method includes: S1: Obtain optical remote sensing image data; S2: Analyze the optical remote sensing image data to obtain the original multispectral image I1 and the original panchromatic image I2; S3: Respectively obtain the feature points of I1 and I2, and perform feature registration on the feature points of I1 and I2 to form a plurality of corresponding point pairs, and all the corresponding point pairs form a basic corresponding point set; S4: Use an optimization strategy to optimize the basic corresponding point set in units of the corresponding point pairs to obtain a first corresponding point set; S5: Select corresponding point pairs from the first corresponding point set based on geometric importance and distribution balance to obtain a second corresponding point set; S6: Calculate the geometric accuracy for the second corresponding point set; S7: Based on the geometric accuracy, determine the usability of the optical remote sensing image data to obtain the geometric quality usability of the optical remote sensing image data; The S4 includes: S41: Select a corresponding point pair MPi from the basic corresponding point set, where the MPi includes the feature point fi of I1 and the feature point gi of I2, and i is an integer greater than or equal to 1; S42: Calculate the Euclidean distance between the feature points in I1 and fi, sort the results of the Euclidean distances from smallest to largest, select the top n feature points, and construct n - 1 triangles with fi as the fixed vertex and the n feature points respectively according to the proximity principle. The n - 1 triangles form the first geometric constraint set S43: Calculate the Euclidean distance between the feature points in I2 and gi, sort the results of the Euclidean distance from smallest to largest, select the first n feature points, and construct triangles with gi as the fixed vertex and the n feature points according to the proximity principle to form n - 1 triangles. The n - 1 triangles form the second geometric constraint set S44: Calculate respectively and the similarity of the corresponding triangles in, and obtain n - 1 similarity results. Determine the similarity results according to the discrimination condition. If the similarity results meet the discrimination condition, add fi and gi to the first set of same - name points, where the discrimination condition is that at least d similarity results among the n - 1 similarity results are greater than the first threshold; S45: Repeat steps S41 to S44 to traverse all the corresponding point pairs in the basic corresponding point set to obtain the first corresponding point set.

2. The geometric quality availability evaluation method of an optical remote sensing image according to claim 1, characterized in that Step S5 includes: S51: Select m corresponding points from the corresponding points of I1 in the first corresponding point set based on geometric importance, and obtain m corresponding point pairs from the first corresponding point set according to the m corresponding points, and add the m corresponding point pairs to the second corresponding point set, where the m corresponding points are uniformly selected according to the vertical orbit direction, orbit direction, and diagonal direction of I1; S52: Use the remaining corresponding point pairs after removing the m corresponding point pairs from the first corresponding point set as the first remaining point set, and select k corresponding point pairs from the first remaining point set based on distribution balance, where k > 4; The S52 includes: S521: Obtain the four corner points of I1, and select four corresponding points from the first remaining point set according to the principle of proximity to the corner points, and add the corresponding point pairs of the four corresponding points to the second corresponding point set; S522: Use the remaining corresponding point pairs after removing the four corresponding point pairs from the first remaining point set as the second remaining point set, cluster the corresponding points of I1 in the second remaining point set according to the clustering algorithm to obtain k - 4 clustering centers, and add the corresponding point pairs corresponding to the k - 4 clustering centers to the second corresponding point set.

3. The geometric quality availability evaluation method for optical remote sensing images according to claim 1, characterized in that The geometric accuracy is calculated based on the geometric registration accuracy and the geometric deformation accuracy.

4. The geometric quality availability evaluation method of an optical remote sensing image according to claim 3, characterized in that The calculation method of the geometric deformation accuracy is: Step 1: Establish a Delaunay triangulation De1 according to the corresponding points of I1 in the second corresponding point set; Step 2: Obtain the corresponding points ei1,1, ei1,2, ei1,3 of the triangle ei1 in De1, use the corresponding relationship to find the corresponding points ei2,1, ei2,2, ei2,3 of the corresponding points ei1,1, ei1,2, ei1,3 on I2, and construct a triangle ei2 using ei2,1, ei2,2, ei2,3; Step 3: Repeat Step 2 to traverse all the triangles in De1 and construct the Delaunay triangulation De2; Step 4: Calculate the area change value between each pair of corresponding triangles in De1 and De2, and calculate the area standard deviation Darea based on the area change value; Step 5: Calculate the angle change value between each pair of corresponding triangles in De1 and De2, and calculate the angle standard deviation Dangle based on the angle change value; Step 6: Calculate the geometric deformation accuracy by performing calculations on Darea and Dangle based on the operation method; 5. The geometric quality availability evaluation method for optical remote sensing images according to claim 3, characterized in that The calculation method of the geometric registration accuracy is as follows: Construct a transformation matrix using the corresponding points in the second corresponding point set; Using the transformation matrix, transform the corresponding points in I1 in the second corresponding point set into the space of I2 to obtain the transformed corresponding points; Calculate the distance between all the transformed corresponding points and the corresponding points in I2 in the second corresponding point set, and calculate the average value of the distances to obtain the geometric registration accuracy; 6. The geometric quality availability evaluation method of an optical remote sensing image according to claim 4, characterized in that The operation method includes weighted calculation, multiplication calculation, and addition calculation; 7. The geometric quality availability evaluation method for optical remote sensing images according to claim 3, characterized in that The calculation method of "calculating according to the geometric registration accuracy and the geometric deformation accuracy" includes weighted calculation, multiplication calculation, and addition calculation; 8. The geometric quality availability evaluation method for optical remote sensing images according to claim 1, characterized in that Step S7 includes: S71: Normalize the geometric accuracy to a value within the range of 0 to 100; S72: Set an availability threshold. When the normalized value of the geometric accuracy is greater than the availability threshold, the geometric quality availability of the optical remote sensing image is 100%, otherwise the geometric quality availability of the optical remote sensing image is 0%; 9. The geometric quality availability evaluation method for optical remote sensing images according to claim 1, characterized in that Step S3 includes: S31: Extract the feature points of I1 based on the scale-invariant feature transform algorithm to obtain the first feature point set; S32: Extract the feature points of I2 based on the scale-invariant feature transform algorithm to obtain the second feature point set; S33: Perform feature point matching on the first feature point set and the second feature point set, and use the two successfully matched feature points as a corresponding point pair. All the corresponding point pairs form the basic corresponding point set.

Images