Remote sensing image cataloging data similar repeat cleaning method

Abstract

The application discloses a remote sensing image cataloging data similar repetition cleaning method, comprising: obtaining remote sensing image cataloging database information; the remote sensing image cataloging database information comprises attribute information of remote sensing image; time information and orbit number in the remote sensing image cataloging database information are used for screening operation, and first similar image data set is obtained; clustering and feature matching operation are carried out on the first similar image data set, and matching image data set is obtained; quality evaluation and cleaning operation are carried out on the matching image data set, and remote sensing image cataloging data similar repetition cleaning result is obtained; the remote sensing image cataloging data similar repetition cleaning result comprises remote sensing image information and score information. The method solves the problem of image repetition production of section cataloging and scene cataloging data caused by image product production problems, repeated receiving and processing of original data, and improves the detection accuracy.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of image catalog data processing, and particularly relates to a remote sensing image catalog data similar repetition cleaning method. BACKGROUND

[0002] For current remote sensing image catalog data, there is a problem of low-quality storage of scene image data caused by repeated production of section catalog and scene catalog data due to repeated reception, processing or production of original data, and further caused by low efficiency and inability to accurately obtain the query of catalog data records and production of data products. Therefore, it is necessary to perform repeated verification on the catalog data, mark the repeated data, and retain the original data to realize cleaning of similar repeated data.

[0003] Scholars at home and abroad have made a lot of research on similar repetition cleaning of catalog data. The current similar repetition cleaning method mainly includes the following types: (1) similar repetition data detection is determined by analyzing the database table, such as the clustering algorithm of DBSCAN, which realizes similar repetition detection on data. This kind of method is simple and efficient, but when the database table records are redundant and the corresponding entity images are not similar repetition, this kind of method does not have the condition of direct deletion. For example, due to production processing, the images produced before and after contain different regions of interest, or the images produced before and after lack different blocks, etc. In the above cases, directly deleting similar repeated records will cause the loss of useful data, etc. (2) Similar repetition detection is performed on images by image matching algorithm. Image matching is a basic problem of digital photogrammetry and computer vision. At present, image matching algorithm mainly includes three types: 1) feature space selection; 2) similarity measurement; 3) search method, etc. Among them, the feature matching algorithm is widely used in image matching. However, the amount of data in the remote sensing image catalog database is large, and when the feature matching detection is directly performed on the entity image data in the database, there is a problem of slow processing speed and low efficiency. SUMMARY

[0004] In view of the problems of limited similar repetition detection accuracy, slow processing speed and low efficiency of the existing similar repetition cleaning method of image catalog data, the present application discloses a remote sensing image catalog data similar repetition cleaning method, which adopts an improved DBSCAN similar repetition record detection combined with image matching to clean the similar repetition of data in the catalog database table, which can not only reduce the range of data cleaning, but also accurately detect the similar repetition of entity image data. Finally, the final cleaning strategy is determined through image quality evaluation, so as to realize efficient and accurate cleaning.

[0005] The present application discloses a remote sensing image catalog data similar repetition cleaning method, comprising:

[0006] S1, obtaining remote sensing image catalog database information; the remote sensing image catalog database information includes attribute information of remote sensing images; the attribute information includes sequence information, time information, orbit circle number, star type information, load information, latitude and longitude information, and storage location information;

[0007] S2, using the time information and the orbit circle number in the remote sensing image catalog database information to perform a screening operation on the remote sensing image catalog database information to obtain a first similar image data set;

[0008] S3, performing clustering and feature matching operations on the first similar image data set to obtain a matching image data set;

[0009] S4, performing quality evaluation and cleaning operations on the matching image data set to obtain a remote sensing image catalog data similar repetition cleaning result; the remote sensing image catalog data similar repetition cleaning result includes remote sensing image information and score information.

[0010] The use of the time information and the orbit circle number in the remote sensing image catalog database information to perform a screening operation on the remote sensing image catalog database information to obtain a first similar image data set, comprising:

[0011] S21, using the time information and the orbit circle number in the remote sensing image catalog database information to perform a screening operation on the remote sensing image catalog database information to obtain sequence information corresponding to the same time information and orbit circle number; using the same time information and orbit circle number to generate a space-time label;

[0012] S22, determining relevant storage location information according to the sequence information obtained in step S21;

[0013] S23, obtaining corresponding remote sensing image information according to the relevant storage location information;

[0014] S24, using the space-time label to label the corresponding remote sensing image information to obtain space-time labeled image data;

[0015] S25, using all space-time labeled image data to construct a first similar image data set.

[0016] The clustering and feature matching operations on the first similar image data set to obtain a matching image data set, comprising:

[0017] S31, performing a clustering operation on the first similar image data set to obtain a second similar image data set;

[0018] S32, performing a feature matching operation on the second similar image dataset to obtain a matching image dataset.

[0019] The clustering operation on the first similar image dataset to obtain a second similar image dataset comprises:

[0020] S311, obtaining attribute information of remote sensing images with the same spatio-temporal label in the first similar image dataset;

[0021] S312, removing time information and track number from the attribute information, and constructing clustering attribute information using the obtained attribute information;

[0022] S313, performing quantization processing on the clustering attribute information to obtain a clustering attribute vector of the remote sensing image;

[0023] S314, performing clustering processing on each group of remote sensing images with the same spatio-temporal label using the clustering attribute vector to obtain clustering information;

[0024] S315, generating a clustering label using the clustering information;

[0025] S316, performing weighted similarity calculation processing on the clustering attribute vector of remote sensing images with the same clustering label to obtain a classification label;

[0026] S317, performing aggregation processing on remote sensing images with the same classification label to obtain a second similar image dataset;

[0027] The weighted similarity calculation processing on the clustering attribute vector of remote sensing images with the same clustering label to obtain a classification label comprises:

[0028] S3161, performing similarity calculation on the clustering attribute vector of a combination of two remote sensing images with the same clustering label to obtain a similarity value;

[0029] S3162, if the similarity value is less than or equal to a preset first similarity discrimination threshold, the combination of the two remote sensing images is not processed;

[0030] If the similarity value is greater than the preset first similarity discrimination threshold, S3163 is executed;

[0031] S3163, performing weighted similarity calculation on the clustering attribute vector of a combination of two remote sensing images with the same clustering label to obtain a weighted similarity value;

[0032] S3164, if the weighted similarity value is less than or equal to a preset second similarity discrimination threshold, the combination of the two remote sensing images is not processed;

[0033] If the weighted similarity value is greater than a preset second similarity discrimination threshold, the same classification label is assigned to the remote sensing images in the combination of the two remote sensing images.

[0034] The feature matching operation on the second similar image dataset to obtain a matched image dataset comprises:

[0035] S321, performing a feature point detection operation on all remote sensing images of the second similar image dataset to obtain first feature point information;

[0036] S322, determining first detection range information using the position information of the first feature points and preset first detection region scale information;

[0037] S323, performing a feature point extraction operation on the first detection range to obtain second feature point information;

[0038] S324, performing a feature matching operation on the second feature points of the remote sensing images with the same classification label to obtain matched feature point pair information;

[0039] S325, generating a matched label using the matched feature point pair information;

[0040] S326, performing projection matching detection processing on the remote sensing images with the same matched label to obtain matched image pairs;

[0041] S327, performing aggregation processing on all matched image pairs to obtain a matched image dataset.

[0042] The feature point detection operation on all remote sensing images of the second similar image dataset to obtain first feature point information comprises:

[0043] S3211, for each remote sensing image of the second similar image dataset, calculating a filter matrix M for a pixel point with an image position coordinate (x, y); the calculation expression of the filter matrix is:

[0044]

[0045] wherein, I x is a derivative value obtained by performing derivation on the remote sensing image at the pixel point with the image position coordinate (x, y) along the x-axis direction of the image, I y is a derivative value obtained by performing derivation on the remote sensing image at the pixel point with the image position coordinate (x, y) along the y-axis direction of the image, ω(x, y) is a filter function of the pixel point with the image position coordinate (x, y), and L is a neighboring region of the pixel point with the image position coordinate (x, y).

[0046] S3212, a response function F(x, y) of the pixel point is calculated by using the filter matrix M; the calculation expression of the response function F(x, y) is:

[0047] F(x, y) = det(M) - Ktrace 2 (M),

[0048] Wherein, det(M) and trace(M) are the determinant value and trace value of the filter matrix M respectively, and K is a preset parameter;

[0049] S3213, for the response function of all pixel points of the remote sensing image, whether greater than a preset response threshold is judged, to obtain a response function judgment result;

[0050] For the response function of the response function judgment result, no processing is performed;

[0051] The pixel point corresponding to the response function of the response function judgment result is determined as the first feature point of the remote sensing image, and the first feature point information is obtained.

[0052] The first detection range is subjected to a feature point extraction operation to obtain second feature point information, comprising:

[0053] S3231, using scale invariant feature descriptor, the pixel points in the first detection range are subjected to a feature point extraction operation to obtain second feature points and corresponding feature descriptor information;

[0054] S3232, the second feature points and the corresponding feature descriptor information are subjected to a fusion operation to obtain second feature point information;

[0055] The second feature points of the remote sensing image with the same classification label are subjected to a feature matching operation to obtain matching feature point pair information, comprising:

[0056] S3241, for any two remote sensing images with the same classification label, for each second feature point of the first remote sensing image, the Euclidean distance between the second feature point and all second feature points of the second remote sensing image is calculated in sequence, and the first smallest Euclidean distance and the second smallest Euclidean distance are searched from all Euclidean distances;

[0057] S3242, the ratio of the first smallest Euclidean distance and the second smallest Euclidean distance is calculated, if the ratio is less than a set ratio threshold, it is determined that the second feature point of the first remote sensing image and the second feature point of the second remote sensing image with the smallest Euclidean distance are matching feature point pairs; if the ratio is not less than the set ratio threshold, no operation is performed;

[0058] S3243, performing the operations of S3241 and S3242 on all combinations of two remote sensing images with the same classification label to obtain matching feature point pair information of all remote sensing images with the same classification label; the matching feature point pair information includes second feature point information contained in the matching feature point pair.

[0059] The projection matching detection processing on the remote sensing images with the same matching label to obtain the matching image pair includes:

[0060] S3261, taking each two remote sensing images with the same matching label as a first remote sensing image and a second remote sensing image respectively;

[0061] S3262, using a minimum matching error model to process the second feature points of the matching feature point pairs commonly contained in the first remote sensing image and the second remote sensing image to obtain a projection transformation matrix;

[0062] S3263, using the projection transformation matrix to perform projection transformation processing on the first remote sensing image to obtain a projection transformation feature point;

[0063] S3264, calculating a matching similarity value of the second feature points of the matching feature point pairs in the second remote sensing image and the projection transformation feature point;

[0064] S3265, determining whether the matching similarity value is less than a preset matching similarity threshold value; if the matching similarity value is less than the preset matching similarity threshold value, determining that the first remote sensing image and the second remote sensing image are a matching image pair; if the matching similarity value is not less than the preset matching similarity threshold value, determining that the first remote sensing image and the second remote sensing image are not a matching image pair.

[0065] The quality evaluation and cleaning operation on the matching image data set to obtain a similar repetition cleaning result of the remote sensing image cataloging data includes:

[0066] S401, obtaining mean error information of each pixel of each remote sensing image of the matching image data set; the mean error of the pixel is a difference between longitude and latitude information of the pixel of the remote sensing image and longitude and latitude information of a real object corresponding to the pixel of the remote sensing image;

[0067] S402, performing mean value calculation on the mean errors of all pixels of the remote sensing image to obtain a mean value of the mean errors of the remote sensing image;

[0068] S403, determining whether the mean value of the mean errors of the remote sensing image is less than or equal to a preset mean error threshold to obtain a mean error determination result of the remote sensing image;

[0069] If the mean error discrimination result is yes, it is determined that the remote sensing image satisfies the position accuracy qualified condition;

[0070] If the mean error discrimination result is no, it is determined that the remote sensing image does not satisfy the position accuracy qualified condition;

[0071] S404, according to the mean value of the mean error of the remote sensing image, the position accuracy quality score of the remote sensing image is calculated, and the calculation expression is:

[0072] Position accuracy quality score = 60 + 40 x (preset mean error threshold - mean error mean value) / 0.7 x preset mean error threshold;

[0073] S405, the color model information of each remote sensing image of the matching image data set is obtained, if the color mode information of the remote sensing image is full spectrum mode, it is determined that the remote sensing image satisfies the color mode qualified condition;

[0074] If the color mode information of the remote sensing image is not full spectrum mode, it is determined that the remote sensing image does not satisfy the color mode qualified condition;

[0075] For the remote sensing image satisfying the color mode qualified condition, the color mode quality score of the remote sensing image is determined as 100;

[0076] For the remote sensing image not satisfying the color mode qualified condition, the color mode quality score of the remote sensing image is determined as 0;

[0077] The color mode information of the remote sensing image includes full spectrum mode, panchromatic mode and infrared mode.

[0078] S406, the color error pixel number value of each remote sensing image of the matching image data set is obtained, if the color error pixel number value is less than or equal to the preset allowed total color error pixel number value, it is determined that the remote sensing image satisfies the color error qualified condition; if the color error pixel number value is greater than the preset allowed total color error pixel number value, it is determined that the remote sensing image does not satisfy the color error qualified condition;

[0079] S407, the color error quality score of the remote sensing image is calculated by using the color error pixel number value of the remote sensing image; the calculation expression is:

[0080] Color error quality score = 60 + 40 x (preset allowed total color error pixel number value - color error pixel number value) / preset allowed total color error pixel number value;

[0081] S408, acquiring an image noise pixel number value of each remote sensing image of the matching image data set, if the image noise pixel number value is less than or equal to a preset allowed total image noise pixel number value, determining that the remote sensing image satisfies an image noise qualified condition; if the image noise pixel number value is greater than the preset allowed total image noise pixel number value, determining that the remote sensing image does not satisfy the image noise qualified condition;

[0082] S409, calculating an image noise quality score of the remote sensing image by using the image noise pixel number value of the remote sensing image; the calculation expression is:

[0083] image noise quality score = 60 + 40 x (pre-set allowed total image noise pixel number value-image noise pixel number value) / pre-set allowed total image noise pixel number value;

[0084] S410, acquiring an information loss pixel number value of each remote sensing image of the matching image data set, if the information loss pixel number value is less than or equal to a preset allowed total information loss pixel number value, determining that the remote sensing image satisfies an information loss qualified condition; if the information loss pixel number value is greater than the preset allowed total information loss pixel number value, determining that the remote sensing image does not satisfy the information loss qualified condition;

[0085] S411, calculating an information loss quality score of the remote sensing image by using the information loss pixel number value of the remote sensing image; the calculation expression is:

[0086] information loss quality score = 60 + 40 x (pre-set allowed total information loss pixel number value-information loss pixel number value) / pre-set allowed total information loss pixel number value;

[0087] S412, for each remote sensing image of the matching image data set, determining a minimum value of all type quality scores of the remote sensing image as a total quality score of the remote sensing image;

[0088] S413, for each remote sensing image of the matching image data set, performing total quality score discrimination, if the total quality score of the remote sensing image is greater than or equal to 90 points, judging that a quality level of the remote sensing image is excellent;

[0089] if the total quality score of the remote sensing image is greater than or equal to 75 points and less than 90 points, judging that the quality level of the remote sensing image is good;

[0090] if the total quality score of the remote sensing image is greater than or equal to 60 points and less than 75 points, judging that the quality level of the remote sensing image is qualified;

[0091] If the remote sensing image does not satisfy any one of the quality element qualified conditions, or the total quality score of the remote sensing image is less than 60 points, it is judged that the quality level of the remote sensing image is unqualified;The quality element qualified condition includes a position accuracy qualified condition, a color mode qualified condition, a color error qualified condition, an image noise qualified condition and an information loss qualified condition.

[0092] S414, the matching image data set is cleaned, if the quality level of all remote sensing images of the matching image pair of the matching image data set is the same, one remote sensing image in the matching image pair is reserved;If the quality level of the remote sensing image of the matching image data set is unqualified, the remote sensing image is deleted;If the quality level of all remote sensing images of the matching image pair of the matching image data set is not the same and is not unqualified, the remote sensing image is reserved and labeled according to the total quality score of the remote sensing image, and the score information of the remote sensing image is obtained;

[0093] S415, the reserved remote sensing image and the score information of the remote sensing image of the matching image data set are used to construct the remote sensing image catalog data similar repetition cleaning result.

[0094] The beneficial effects of the present application are:

[0095] The present application designs a remote sensing image catalog data similar repetition cleaning method, adopts the combination of similar repetition record detection and image matching to detect the similar repetition of catalog data to ensure the accuracy of detection, finally determines the final cleaning strategy through quality evaluation, solves the problems of image repetition production of section catalog and scene catalog data caused by image product production problems, repeated reception and processing of original data, and improves the accuracy of detection. BRIEF DESCRIPTION OF DRAWINGS

[0096] Figure 1 A flowchart of a remote sensing image catalog data similar repetition cleaning method in the embodiment of the present application is shown;

[0097] Figure 2 A catalog database table record of similar repetition in the embodiment of the present application is shown.

[0098] Figure 3 A catalog image of similar repetition in the embodiment of the present application is shown. DETAILED DESCRIPTION

[0099] Exemplary embodiments of the present application will be described herein below with reference to the accompanying drawings. While exemplary embodiments of the present application are shown in the drawings, it is understood that the present application can be embodied in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the application to those skilled in the art.

[0100] Figure 1 A flowchart of a remote sensing image cataloging data similar repetition cleaning method in an embodiment of the present application is shown; Figure 2 A similar repetition cataloging database table record in an embodiment of the present application is shown. Figure 3 A similar repetition cataloging image map in an embodiment of the present application is shown.

[0101] Current database table-based cataloging data similar repetition cleaning can cause image information loss, false deletion, etc. And image entity-based similar repetition detection can cause slow processing speed, low efficiency, etc.

[0102] The present embodiment discloses a remote sensing image cataloging data similar repetition cleaning method, comprising:

[0103] S1, obtaining remote sensing image cataloging database information; the remote sensing image cataloging database information includes attribute information of remote sensing images; the attribute information includes sequence information, time information, orbit circle number, star type information, load information, latitude and longitude information, and storage location information;

[0104] S2, using the time information and the orbit circle number in the remote sensing image cataloging database information to perform a screening operation on the remote sensing image cataloging database information, to obtain a first similar image data set;

[0105] S3, performing clustering and feature matching operations on the first similar image data set to obtain a matched image data set;

[0106] S4, performing quality evaluation and cleaning operations on the matched image data set to obtain a remote sensing image cataloging data similar repetition cleaning result; the remote sensing image cataloging data similar repetition cleaning result includes remote sensing image information and score information.

[0107] The use of the time information and the orbit circle number in the remote sensing image cataloging database information to perform a screening operation on the remote sensing image cataloging database information to obtain a first similar image data set, comprises:

[0108] S21, filtering the remote sensing image catalog database information according to the time information and the orbit circle number in the remote sensing image catalog database information, to obtain sequence information corresponding to the same time information and orbit circle number; and generating a space-time label according to the same time information and orbit circle number;

[0109] S22, determining relevant storage location information according to the sequence information obtained in step S21;

[0110] S23, obtaining corresponding remote sensing image information according to the relevant storage location information;

[0111] S24, labeling the corresponding remote sensing image information according to the space-time label, to obtain space-time labeled image data;

[0112] S25, constructing a first similar image data set by using all the space-time labeled image data.

[0113] The clustering and feature matching operation on the first similar image data set to obtain a matching image data set comprises:

[0114] S31, performing a clustering operation on the first similar image data set to obtain a second similar image data set;

[0115] S32, performing a feature matching operation on the second similar image data set to obtain a matching image data set.

[0116] The clustering operation on the first similar image data set to obtain a second similar image data set comprises:

[0117] S311, obtaining attribute information of remote sensing images with the same space-time label in the first similar image data set;

[0118] S312, removing the time information and the orbit circle number in the attribute information, and constructing clustering attribute information by using the obtained attribute information;

[0119] S313, performing quantization processing on the clustering attribute information to obtain a clustering attribute vector of the remote sensing image;

[0120] The S313 is to respectively convert the sequence information, the star type information, the load information, the latitude and longitude information, and the storage location information into corresponding quantization data, and construct a clustering attribute vector by using the quantization data;

[0121] S314, performing clustering processing on each group of remote sensing images with the same space-time label by using the clustering attribute vector thereof to obtain clustering information;

[0122] The S314 comprises:

[0123] S3141, for each group of remote sensing images with the same spatio-temporal label, calculate the relative distance between any two vectors of the clustering attribute vectors; select the two vectors with the smallest relative distance as a class vector;

[0124] The relative distance can be Euclidean distance, Manhattan distance, Chebyshev distance, Chi-square distance, etc.

[0125] S3142, use the center point of the two vectors in the class vector as a vector representing the class of the vector;

[0126] S3143, calculate the distance between all other clustering attribute vectors and the vector representing the class of the vector, and if the distance is less than or equal to a preset distance threshold, the vector is classified into the class of the vector;

[0127] S3144, for the vectors with a distance greater than or equal to the preset distance threshold, repeat the judgment of S3141 to S3143 until all vectors are classified into the class of the vector;

[0128] S3145, use the class vector information of all clustering attribute vectors to construct clustering information;

[0129] S315, use the clustering information to generate a clustering label;

[0130] S316, use the clustering attribute vectors of remote sensing images with the same clustering label to perform weighted similarity calculation processing to obtain a classification label;

[0131] S317, aggregate remote sensing images with the same classification label to obtain a second similar image dataset;

[0132] The weighted similarity calculation processing using the clustering attribute vectors of remote sensing images with the same clustering label to obtain a classification label comprises:

[0133] S3161, for a combination of two remote sensing images with the same clustering label, perform similarity calculation on the clustering attribute vectors to obtain a similarity value;

[0134] The similarity calculation on the clustering attribute vectors in S3161 can use cosine similarity or distance similarity.

[0135] For clustering attribute vectors a and b of two remote sensing images, a=(a1, a2, …, a5) and b=(b1, b2, …, b5), the cosine similarity calculation formula is:

[0136]

[0137] wherein, cos(a, b) represents the cosine similarity value of a and b;

[0138] Correspondingly, the calculation formula of the distance similarity is:

[0139]

[0140] wherein, d(a, b) represents the distance similarity value of a and b;

[0141] S3162, if the similarity value is less than or equal to a preset first similarity discrimination threshold, the combination of the two remote sensing images is not processed;

[0142] If the similarity value is greater than the preset first similarity discrimination threshold, S3163 is executed;

[0143] The preset first similarity discrimination threshold can be 5% or 10% of the average value of all similarities.

[0144] S3163, for the combination of the two remote sensing images with the same clustering label, the weighted similarity calculation is performed on the clustering attribute vector to obtain a weighted similarity value;

[0145] In the S3163, the weighted similarity calculation on the clustering attribute vector can use cosine weighted similarity or distance weighted similarity.

[0146] For the clustering attribute vectors a and b of the two remote sensing images, a=(a1, a2, …, a5) and b=(b1, b2, …, b5), the calculation formula of the cosine weighted similarity is:

[0147]

[0148] wherein, cos(a, b) represents the cosine weighted similarity value of a and b, c i is the weighted value of the i-th component of the clustering attribute vector; correspondingly, the calculation formula of the distance weighted similarity is:

[0149]

[0150] wherein, d(a, b) represents the distance weighted similarity value of a and b; the weighted value is determined according to the importance degree of each component of the clustering attribute vector, and the value can be c=[c1, c2, c3, c4, c5]=[0.05, 0.25, 0.25, 0.4, 0.05];

[0151] S3164, if the weighted similarity value is less than or equal to a preset second similarity discrimination threshold, the combination of the two remote sensing images is not processed;

[0152] If the weighted similarity value is greater than a preset second similarity discrimination threshold, the same classification label is assigned to the remote sensing images in the combination of the two remote sensing images.

[0153] The preset second similarity discrimination threshold can be 5% or 10% of the average value of all weighted similarities.

[0154] The classification labels of all remote sensing images are the classification labels obtained in S316.

[0155] The step S316 is performed on all combinations of two remote sensing images with the same clustering label. If a remote sensing image is assigned two or more classification labels, one of the classification labels can be selected, and other remote sensing images related to the remote sensing image are assigned the classification label.

[0156] The feature matching operation on the second similar image dataset to obtain the matched image dataset includes:

[0157] S321, performing a feature point detection operation on all remote sensing images of the second similar image dataset to obtain first feature point information;

[0158] S322, determining first detection range information using position information of the first feature point and preset first detection region scale information; the first detection range can be a rectangular or circular region centered on the first feature point and having a side length or diameter equal to the first detection region scale information. The first detection region scale information can be obtained by multiplying the resolution of the remote sensing image by a coefficient k, and the value range of the coefficient k is [5, 10].

[0159] S323, performing a feature point extraction operation on the first detection range to obtain second feature point information;

[0160] S324, performing a feature matching operation on the second feature points of the remote sensing images with the same classification label to obtain matched feature point pair information;

[0161] S325, generating matched labels using the matched feature point pair information; a corresponding matched label is generated for each second feature point in each matched feature point pair; the matched labels of the second feature points in each matched feature point pair are the same; the matched labels of the second feature points in different matched feature point pairs are different.

[0162] S326, performing projection matching detection processing on the remote sensing images with the same matching label to obtain a matching image pair; the matching image pair includes a plurality of mutually matching remote sensing images; the matching label of the remote sensing image is the matching label of the second feature point in the matching feature point pair contained in the remote sensing image.

[0163] S327, performing aggregation processing on all the matching image pairs to obtain a matching image dataset.

[0164] The S327 is to construct the matching image dataset by using all the matching image pairs.

[0165] The feature point detection operation on all the remote sensing images of the second similar image dataset to obtain the first feature point information includes:

[0166] S3211, for each remote sensing image of the second similar image dataset, a filter matrix M of a pixel point with an image position coordinate (x, y) is calculated; the calculation expression of the filter matrix is:

[0167]

[0168] wherein, I x is a derivative value obtained by performing derivation on the remote sensing image at the pixel point with the image position coordinate (x, y) along the x-axis direction of the image, I y is a derivative value obtained by performing derivation on the remote sensing image at the pixel point with the image position coordinate (x, y) along the y-axis direction of the image, ω(x, y) is a filter function of the pixel point with the image position coordinate (x, y), and L is a neighboring region of the pixel point with the image position coordinate (x, y).

[0169] The neighboring region of the pixel point with the image position coordinate (x, y) can be a rectangular region obtained by extending left and right along the x-axis and up and down along the y-axis with the pixel point as the center. The rectangular region can have 10 pixel points in the x-axis direction and 8 pixel points in the y-axis direction. The filter function can be a two-dimensional Gaussian function.

[0170] S3212, a response function F(x, y) of the pixel point is calculated by using the filter matrix M; the calculation expression of the response function F(x, y) is:

[0171] F(x, y) = det(M) - Ktrace 2 (M),

[0172] wherein, det(M) and trace(M) are the determinant value and the trace value of the filter matrix M respectively, and K is a preset parameter.

[0173] S3213, judging whether the response function of all pixel points of the remote sensing image is greater than a preset response threshold, to obtain a response function judgment result;

[0174] The response function with the response function judgment result being no is not processed.

[0175] The pixel point corresponding to the response function with the response function judgment result being yes is determined as a first feature point of the remote sensing image, to obtain first feature point information.

[0176] The first detection range is subjected to a feature point extraction operation to obtain second feature point information, including:

[0177] S3231, using a scale-invariant feature descriptor to perform a feature point extraction operation on the pixel points of the first detection range, to obtain second feature points and corresponding feature descriptor information;

[0178] S3232, performing a fusion operation on the second feature points and the corresponding feature descriptor information, to obtain second feature point information;

[0179] The second feature point information includes second feature points and corresponding feature descriptor information; the feature point extraction operation on the pixel points of the first detection range can be implemented by using a SURF algorithm, an ORB algorithm or a FAST algorithm.

[0180] The feature descriptor corresponding to the second feature point is determined by the feature point extraction algorithm used, and can be a SURF feature descriptor, an ORB feature descriptor or a FAST feature descriptor.

[0181] The second feature points of the remote sensing images with the same classification label are subjected to a feature matching operation to obtain matching feature point pair information, including:

[0182] S3241, for any combination of two remote sensing images with the same classification label, the Euclidean distance between each second feature point of the first remote sensing image and all second feature points of the second remote sensing image is calculated in sequence to search for a first smallest Euclidean distance and a second smallest Euclidean distance from all Euclidean distances; the combination of the two remote sensing images includes the first remote sensing image and the second remote sensing image.

[0183] S3242, calculating a ratio of the first smallest Euclidean distance and the second smallest Euclidean distance, if the ratio is less than a set ratio threshold, determining that the second feature point of the first remote sensing image and the second feature point of the second remote sensing image with the smallest Euclidean distance are a matched feature point pair; if the ratio is not less than the set ratio threshold, not performing an operation; the set ratio threshold can be 0.5;

[0184] S3243, performing the operations of S3241 and S3242 on all combinations of two remote sensing images with the same classification label to obtain matched feature point pair information of all remote sensing images with the same classification label. The matched feature point pair information includes second feature point information contained in the matched feature point pair.

[0185] The projection matching detection processing on the remote sensing images with the same matching label to obtain a matched image pair includes:

[0186] S3261, taking each two remote sensing images with the same matching label as a first remote sensing image and a second remote sensing image respectively;

[0187] S3262, using a minimum matching error model to process the second feature points of the matched feature point pairs commonly contained in the first remote sensing image and the second remote sensing image to obtain a projection transformation matrix;

[0188] The S3262 includes:

[0189] The minimum matching error model is established and its expression is:

[0190]

[0191] Wherein, n is the number of the matched feature point pairs commonly contained in the first remote sensing image and the second remote sensing image, (x i ,y i ) is the image position coordinates of the i-th first feature point of the matched feature point pair in the first remote sensing image, is the image position coordinates of the i-th second feature point of the matched feature point pair in the second remote sensing image, λ is an additional coefficient of the minimum matching error model, a, b, c and d are all solving variables of the minimum matching error model, and J is a minimum matching error function.

[0192] When J takes the minimum value, the corresponding values of the four variables a, b, c and d are a0, b0, c0 and d0 respectively, and a projection transformation matrix T is established and its expression is:

[0193]

[0194] Thus, the projection transformation matrix is obtained.

[0195] The common containing refers to that two second feature points of the matching feature point pair belong to the first remote sensing image and the second remote sensing image respectively.

[0196] S3263, using the projection transformation matrix, performing projection transformation processing on the first remote sensing image to obtain a projection transformation feature point;

[0197] The S3263 comprises:

[0198] Using the projection transformation matrix, multiplying the image position coordinates (x i ,y i ) of the i-th first feature point of the matching feature point pair in the first remote sensing image respectively to obtain the i-th projection transformation feature point, and the image position coordinates of the i-th projection transformation feature point are (a0x i +c0y i ,b0x i +d0y i ).

[0199] S3264, calculating the matching similarity value of the second feature point of the matching feature point pair in the second remote sensing image and the projection transformation feature point;

[0200] The calculation process of the S3264 is expressed as:

[0201]

[0202] Wherein, d1 is the matching similarity value.

[0203] S3265, determining whether the matching similarity value is less than a preset matching similarity threshold value, if the matching similarity value is less than the preset matching similarity threshold value, determining that the first remote sensing image and the second remote sensing image are a matching image pair; if the matching similarity value is not less than the preset matching similarity threshold value, determining that the first remote sensing image and the second remote sensing image are not a matching image pair.

[0204] The preset matching similarity threshold value can be 5% or 10% of the average value of the absolute values of the position coordinates of all second feature points of the matching feature point pair.

[0205] The quality evaluation and cleaning operation on the matching image data set to obtain the remote sensing image cataloging data similarity and repetition cleaning result comprises:

[0206] S401, for each remote sensing image of the matching image data set, obtaining the mean error information of each pixel thereof; the mean error of the pixel is the difference between the latitude and longitude information of the pixel of the remote sensing image and the latitude and longitude information of the real object corresponding to the pixel of the remote sensing image.

[0207] S402, mean error of all pixels of the remote sensing image is calculated to obtain the mean error of the remote sensing image;

[0208] S403, whether the mean error of the remote sensing image is less than or equal to the preset mean error threshold is judged to obtain the mean error discrimination result of the remote sensing image;

[0209] If the mean error discrimination result is yes, it is determined that the remote sensing image meets the position accuracy qualified condition;

[0210] If the mean error discrimination result is no, it is determined that the remote sensing image does not meet the position accuracy qualified condition;

[0211] S404, the position accuracy quality score of the remote sensing image is calculated according to the mean error of the remote sensing image, and the calculation expression is:

[0212] Position accuracy quality score = 60 + 40 × (preset mean error threshold - mean error) / 0.7 × preset mean error threshold;

[0213] S405, color model information of each remote sensing image of the matching image data set is obtained, if the color mode information of the remote sensing image is full spectrum mode, it is determined that the remote sensing image meets the color mode qualified condition;

[0214] If the color mode information of the remote sensing image is not full spectrum mode, it is determined that the remote sensing image does not meet the color mode qualified condition;

[0215] For the remote sensing image meeting the color mode qualified condition, the color mode quality score of the remote sensing image is determined as 100;

[0216] For the remote sensing image not meeting the color mode qualified condition, the color mode quality score of the remote sensing image is determined as 0;

[0217] The color mode information of the remote sensing image includes full spectrum mode, panchromatic mode and infrared mode.

[0218] S406, color error pixel number value of each remote sensing image of the matching image data set is obtained, if the color error pixel number value is less than or equal to the preset allowed total color error pixel number value, it is determined that the remote sensing image meets the color error qualified condition; if the color error pixel number value is greater than the preset allowed total color error pixel number value, it is determined that the remote sensing image does not meet the color error qualified condition;

[0219] S407, the color error quality score of the remote sensing image is calculated by using the color error pixel number value of the remote sensing image; the calculation expression is:

[0220] Color error quality score = 60 + 40 x (preset allowed total color error pixel value - color error pixel value) / preset allowed total color error pixel value;

[0221] S408, obtaining the image noise pixel value of each remote sensing image of the matching image dataset, if the image noise pixel value is less than or equal to the preset allowed total image noise pixel value, determining that the remote sensing image satisfies the image noise qualified condition; if the image noise pixel value is greater than the preset allowed total image noise pixel value, determining that the remote sensing image does not satisfy the image noise qualified condition;

[0222] S409, calculating the image noise quality score of the remote sensing image by using the image noise pixel value of the remote sensing image; the calculation expression is:

[0223] Image noise quality score = 60 + 40 x (preset allowed total image noise pixel value - image noise pixel value) / preset allowed total image noise pixel value;

[0224] S410, obtaining the information loss pixel value of each remote sensing image of the matching image dataset, if the information loss pixel value is less than or equal to the preset allowed total information loss pixel value, determining that the remote sensing image satisfies the information loss qualified condition; if the information loss pixel value is greater than the preset allowed total information loss pixel value, determining that the remote sensing image does not satisfy the information loss qualified condition;

[0225] S411, calculating the information loss quality score of the remote sensing image by using the information loss pixel value of the remote sensing image; the calculation expression is:

[0226] Information loss quality score = 60 + 40 x (preset allowed total information loss pixel value - information loss pixel value) / preset allowed total information loss pixel value;

[0227] S412, for each remote sensing image of the matching image dataset, determining the minimum value of all type quality scores of the remote sensing image as the total quality score of the remote sensing image;

[0228] S413, for each remote sensing image of the matching image dataset, performing total quality score discrimination, if the total quality score of the remote sensing image is greater than or equal to 90 points, judging that the quality level of the remote sensing image is excellent;

[0229] if the total quality score of the remote sensing image is greater than or equal to 75 points and less than 90 points, judging that the quality level of the remote sensing image is good;

[0230] If the total quality score of the remote sensing image is greater than or equal to 60 points and less than 75 points, the quality level of the remote sensing image is determined to be qualified.

[0231] If the remote sensing image does not satisfy any one of the qualified conditions of the quality elements, or the total quality score of the remote sensing image is less than 60 points, the quality level of the remote sensing image is determined to be unqualified.

[0232] S414, the matching image data set is cleaned, if the quality level of all remote sensing images of the matching image pair of the matching image data set is the same, one remote sensing image in the matching image pair is reserved, if the quality level of the remote sensing image of the matching image data set is unqualified, the remote sensing image is deleted, if the quality level of all remote sensing images of the matching image pair of the matching image data set is not the same and is not unqualified, the remote sensing image is reserved and labeled according to the total quality score of the remote sensing image, and the score information of the remote sensing image is obtained; the score information of the remote sensing image is the total quality score of the remote sensing image.

[0233] S415, the remote sensing image and the score information of the remote sensing image reserved by the matching image data set are used to construct the cataloging data similarity and repetition cleaning result of the remote sensing image.

[0234] The preset allowed total pixel number value in S406, S408 and S410 can be 1% of the total pixel number of the remote sensing image.

[0235] The preset mean error threshold can be 2.2m to 2.6m.

[0236] The present application designs a remote sensing image cataloging data similarity and repetition cleaning method and system, adopts the combination of similarity and repetition record detection and image matching to detect the cataloging data similarity and repetition to ensure the accuracy of detection, finally determines the final cleaning strategy through quality evaluation, solves the problems of image repetition production of section cataloging and scene cataloging data caused by image product production problems, repeated reception and processing of original data, and improves the accuracy of detection.

[0237] The above is only an embodiment of the present application and is not used to limit the present application. The present application can have various changes and variations for those skilled in the art. Any modification, equivalent replacement, improvement, etc. within the spirit and principles of the present application shall be included in the scope of claims of the present application.

Claims

1. A method for cleaning similar duplicates of remote sensing imagery cataloging data, characterized in that, The method comprises the following steps: S1, obtaining remote sensing image catalog database information; the remote sensing image catalog database information comprises attribute information of remote sensing images; the attribute information comprises sequence information, time information, orbit circle number, star type information, load information, latitude and longitude information, and storage location information; S2, performing filtering operation on the remote sensing image catalog database information by using the time information and the orbit circle number in the remote sensing image catalog database information, to obtain a first similar image data set, which comprises the following steps: S21, performing filtering operation on the remote sensing image catalog database information by using the time information and the orbit circle number in the remote sensing image catalog database information, to obtain sequence information corresponding to the same time information and orbit circle number; generating a space-time label by using the same time information and orbit circle number; S22, determining relevant storage location information according to the sequence information obtained in step S21; S23, obtaining corresponding remote sensing image information according to the relevant storage location information; S24, labeling the corresponding remote sensing image information by using the space-time label, to obtain space-time labeled image data; S25, constructing the first similar image data set by using all the space-time labeled image data; S3, performing clustering and feature matching operation on the first similar image data set, to obtain a matching image data set; S4, performing quality evaluation and cleaning operation on the matching image data set, to obtain remote sensing image catalog data similar repetition cleaning result; the remote sensing image catalog data similar repetition cleaning result comprises remote sensing image information and score information.

2. The method of claim 1, wherein the remote sensing image cataloging data similarity de-duplication method is characterized by, The operation of performing clustering and feature matching operation on the first similar image data set to obtain a matching image data set comprises the following steps: S31, performing clustering operation on the first similar image data set to obtain a second similar image data set; S32, performing feature matching operation on the second similar image data set to obtain a matching image data set.

3. The method for cleaning up similar and duplicate remote sensing image cataloging data as described in claim 2, characterized in that, The operation of performing clustering operation on the first similar image data set to obtain a second similar image data set comprises the following steps: S311, obtaining attribute information of remote sensing images with the same space-time label in the first similar image data set; S312, removing the time information and the orbit circle number in the attribute information, and constructing clustering attribute information by using the obtained attribute information; S313, performing quantization processing on the clustering attribute information to obtain a clustering attribute vector of the remote sensing image; S314, performing clustering processing on each group of remote sensing images with the same space-time label by using the clustering attribute vector thereof, to obtain clustering information; S315, generating a clustering label by using the clustering information; S316, performing weighted similarity calculation processing on the clustering attribute vectors of remote sensing images with the same clustering label, to obtain a classification label; S317, performing aggregation processing on remote sensing images with the same classification label, to obtain a second similar image data set.

4. The method of claim 3, wherein the step of cleaning the similar duplicates comprises the step of: 4.

1. removing the similar duplicates from the cataloged data. The operation of performing weighted similarity calculation processing on the clustering attribute vectors of remote sensing images with the same clustering label to obtain a classification label comprises the following steps: ​ S3161, performing similarity calculation on the clustering attribute vectors of the combination of the two remote sensing images with the same clustering label to obtain a similarity value; S3162, if the similarity value is less than or equal to a preset first similarity discrimination threshold, not processing the combination of the two remote sensing images; if the similarity value is greater than the preset first similarity discrimination threshold, performing S3163; S3163, performing weighted similarity calculation on the clustering attribute vectors of the combination of the two remote sensing images with the same clustering label to obtain a weighted similarity value; S3164, if the weighted similarity value is less than or equal to a preset second similarity discrimination threshold, not processing the combination of the two remote sensing images; if the weighted similarity value is greater than the preset second similarity discrimination threshold, assigning the same classification label to the remote sensing images in the combination of the two remote sensing images.

5. The method for cleaning up similar and duplicate remote sensing image cataloging data as described in claim 2, characterized in that, The feature matching operation on the second similar image dataset to obtain the matched image dataset comprises: S321, performing feature point detection operation on all remote sensing images of the second similar image dataset to obtain first feature point information; S322, determining first detection range information by using position information of the first feature point and preset first detection region scale information; S323, performing feature point extraction operation on the first detection range to obtain second feature point information; S324, performing feature matching operation on the second feature points of the remote sensing images with the same classification label to obtain matched feature point pair information; S325, generating a matched label by using the matched feature point pair information; S326, performing projection matching detection processing on the remote sensing images with the same matched label to obtain matched image pairs; S327, performing aggregation processing on all matched image pairs to obtain a matched image dataset.

6. The method of claim 5, wherein the step of cleaning the similar duplicates comprises the step of: 5 deleting the similar duplicates from the remote sensing image catalog data. 6 The feature point detection operation on all remote sensing images of the second similar image dataset to obtain the first feature point information comprises: S3211, for each remote sensing image of the second similar image dataset, calculating a filter matrix M of a pixel point with an image position coordinate (x, y) of the remote sensing image; a calculation expression of the filter matrix is: wherein I x is a derivative value obtained by derivation along the x-axis direction of the image for a pixel point with image position coordinates (x, y) of the remote sensing image, I y is a derivative value obtained by derivation along the y-axis direction of the image for a pixel point with image position coordinates (x, y) of the remote sensing image, ω(x, y) is a filter function of the pixel point with image position coordinates (x, y), and L is a neighboring region of the pixel point with image position coordinates (x, y). S3212, calculating a response function F(x, y) of the pixel point by using the filter matrix M; a calculation expression of the response function F(x, y) is: F(x, y) = det(M) - Ktrace 2 (M), wherein, det(M) and trace(M) are respectively a determinant value and a trace value of the filter matrix M, and K is a preset parameter; S3213, judging whether the response function of all pixel points of the remote sensing image is greater than a preset response threshold to obtain a response function judgment result; not processing the response function with the response function judgment result being no; determining the pixel point corresponding to the response function with the response function judgment result being yes as a first feature point of the remote sensing image to obtain first feature point information.

7. The method for cleaning up similar and duplicate remote sensing image cataloging data as described in claim 5, characterized in that, The feature point extraction operation on the first detection range to obtain the second feature point information comprises: S3231, performing feature point extraction on the pixels in the first detection range by using a scale-invariant feature descriptor to obtain second feature points and corresponding feature descriptor information; S3232, performing fusion on the second feature points and the corresponding feature descriptor information to obtain second feature point information.

8. The method of claim 5, wherein the step of cleaning the similar duplicates of the remote sensing image cataloging data is characterized by, The projection matching detection processing on the remote sensing images with the same matching label to obtain a matching image pair comprises: S3261, taking each two remote sensing images with the same matching label as a first remote sensing image and a second remote sensing image respectively; S3262, processing the second feature points of a matching feature point pair commonly contained in the first remote sensing image and the second remote sensing image by using a minimum matching error model to obtain a projection transformation matrix; S3263, performing projection transformation processing on the first remote sensing image by using the projection transformation matrix to obtain a projection transformation feature point; S3264, calculating a matching similarity value of the second feature point of the matching feature point pair in the second remote sensing image and the projection transformation feature point; S3265, determining whether the matching similarity value is less than a preset matching similarity threshold value, if the matching similarity value is less than the preset matching similarity threshold value, determining that the first remote sensing image and the second remote sensing image are a matching image pair, and if the matching similarity value is not less than the preset matching similarity threshold value, determining that the first remote sensing image and the second remote sensing image are not a matching image pair.

9. The method of claim 1, wherein the remote sensing image cataloging data similarity de-duplication method is characterized by, The quality evaluation and cleaning operation on the matching image data set to obtain a similar repetition cleaning result of the remote sensing image cataloging data comprises: S401, obtaining the mean error information of each pixel of each remote sensing image in the matching image data set, wherein the mean error of the pixel is the difference between the longitude and latitude information of the pixel of the remote sensing image and the longitude and latitude information of the real object corresponding to the pixel of the remote sensing image; S402, performing mean value calculation on the mean errors of all pixels of the remote sensing image to obtain the mean value of the mean errors of the remote sensing image; S403, determining whether the mean value of the mean errors of the remote sensing image is less than or equal to a preset mean error threshold to obtain a mean error discrimination result of the remote sensing image; if the mean error discrimination result is yes, determining that the remote sensing image satisfies the position accuracy qualified condition; if the mean error discrimination result is no, determining that the remote sensing image does not satisfy the position accuracy qualified condition; S404, calculating the position accuracy quality score of the remote sensing image according to the mean value of the mean errors of the remote sensing image, and the calculation expression of the position accuracy quality score is: Position accuracy quality score = 60 + 40 × (preset mean error threshold - mean value of mean errors) / 0.7 × preset mean error threshold; S405, obtaining the color model information of each remote sensing image in the matching image data set, if the color mode information of the remote sensing image is a full spectrum mode, determining that the remote sensing image satisfies the color mode qualified condition; if the color mode information of the remote sensing image is not a full spectrum mode, determining that the remote sensing image does not satisfy the color mode qualified condition; The color mode quality score of the remote sensing image meeting the color mode qualified condition is determined as 100; The color mode quality score of the remote sensing image not meeting the color mode qualified condition is determined as 0; The color mode information of the remote sensing image includes full spectrum mode, panchromatic mode and infrared mode; S406, the color error pixel number value of each remote sensing image of the matching image data set is obtained, if the color error pixel number value is less than or equal to the preset allowed total color error pixel number value, it is determined that the remote sensing image meets the color error qualified condition; if the color error pixel number value is greater than the preset allowed total color error pixel number value, it is determined that the remote sensing image does not meet the color error qualified condition; S407, the color error quality score of the remote sensing image is calculated by using the color error pixel number value of the remote sensing image; the calculation expression of the color error quality score is: Color error quality score = 60 + 40 × (preset allowed total color error pixel number value - color error pixel number value) / preset allowed total color error pixel number value; S408, the image noise pixel number value of each remote sensing image of the matching image data set is obtained, if the image noise pixel number value is less than or equal to the preset allowed total image noise pixel number value, it is determined that the remote sensing image meets the image noise qualified condition; if the image noise pixel number value is greater than the preset allowed total image noise pixel number value, it is determined that the remote sensing image does not meet the image noise qualified condition; S409, the image noise quality score of the remote sensing image is calculated by using the image noise pixel number value of the remote sensing image; the calculation expression of the image noise quality score is: Image noise quality score = 60 + 40 × (preset allowed total image noise pixel number value - image noise pixel number value) / preset allowed total image noise pixel number value; S410, the information loss pixel number value of each remote sensing image of the matching image data set is obtained, if the information loss pixel number value is less than or equal to the preset allowed total information loss pixel number value, it is determined that the remote sensing image meets the information loss qualified condition; if the information loss pixel number value is greater than the preset allowed total information loss pixel number value, it is determined that the remote sensing image does not meet the information loss qualified condition; S411, the information loss quality score of the remote sensing image is calculated by using the information loss pixel number value of the remote sensing image; the calculation expression of the information loss quality score is: Information loss quality score = 60 + 40 × (preset allowed total information loss pixel number value - information loss pixel number value) / preset allowed total information loss pixel number value; S412, for each remote sensing image of the matching image data set, the minimum value of all type quality scores of the remote sensing image is determined as the total quality score of the remote sensing image; S413, for each remote sensing image of the matching image data set, total quality score discrimination is performed, if the total quality score of the remote sensing image is greater than or equal to 90 points, it is judged that the quality level of the remote sensing image is excellent. If the total quality score of the remote sensing image is greater than or equal to 75 and less than 90, the quality level of the remote sensing image is determined to be good; If the total quality score of the remote sensing image is greater than or equal to 60 and less than 75, the quality level of the remote sensing image is determined to be qualified; If the remote sensing image does not satisfy any one of the qualified conditions of the quality elements, or the total quality score of the remote sensing image is less than 60, the quality level of the remote sensing image is determined to be unqualified; the qualified conditions of the quality elements include a qualified condition of position accuracy, a qualified condition of color mode, a qualified condition of color error, a qualified condition of image noise, and a qualified condition of information loss; S414, the matching image data set is cleaned, if the quality levels of all remote sensing images of the matching image pair of the matching image data set are the same, one remote sensing image in the matching image pair is reserved; if the quality level of the remote sensing image of the matching image data set is unqualified, the remote sensing image is deleted; if the quality levels of all remote sensing images of the matching image pair of the matching image data set are not the same and are not unqualified, the remote sensing images are reserved and labeled according to the total quality scores of the remote sensing images, and the scoring information of the remote sensing images is obtained; S415, the remote sensing image cataloging data similarity and repetition cleaning result is constructed by using the reserved remote sensing images and the scoring information of the remote sensing images of the matching image data set.

Images