High-resolution satellite remote sensing image registration method combined with geographic information

Abstract

The present invention relates to a high-resolution satellite remote sensing image registration method combined with geographic information, comprising: step S1, inputting a reference image Ref and an image to be registered Tar, detecting metadata information of the input image; step S2, screening a significant area and completing a rough positioning; step S3, extracting image features, performing local feature matching, and obtaining an initial matching point pair; step S4, outputting the registered image through coordinate correction and transformation. The present invention utilizes the geographic information carried by satellite images, combined with SIFT feature extraction and matching technology, to complete high-efficiency and high-precision registration of high-resolution satellite remote sensing images.

Description

Technical Field

[0001] The present invention relates to the technical field of remote sensing image processing, and in particular to a high-resolution satellite remote sensing image registration method combined with geographic information. Background Art

[0002] In the field of remote sensing image processing technology, remote sensing image registration is a fundamental problem in remote sensing image analysis. It is mainly used to spatially align remote sensing images acquired at different phases, different sensors or different observation angles for subsequent image fusion, change detection, semantic intelligence generation and other processing. However, with the improvement of remote sensing image resolution, traditional remote sensing image registration methods face the following two main problems:

[0003] (1) Large amount of computation: High-resolution remote sensing images usually contain a large amount of pixel data. Traditional remote sensing image registration methods (such as pixel-based correlation calculation) often require a large amount of computing resources, have slow processing speeds, and are not suitable for real-time registration of large-scale images.

[0004] (2) Low registration accuracy: Existing methods usually rely on the grayscale or color information of the image, while high-resolution satellite remote sensing images often contain complex ground features. Direct registration may be affected by factors such as illumination changes, scale changes, and image noise, resulting in insufficient registration accuracy.

[0005] In order to solve the above problems, more and more research has begun to focus on how to use the feature information of images (such as corners, edges, etc.) for registration. The feature point matching method can reduce the amount of calculation and improve the registration accuracy by accurately matching the feature points. However, when facing high-resolution images and complex geographic information, the existing registration methods often ignore the auxiliary information such as geographic coordinates of satellite remote sensing images. Summary of the invention

[0006] In order to solve the technical problems existing in the above-mentioned prior art, the purpose of the present invention is to provide a high-resolution satellite remote sensing image registration method combined with geographic information. By combining geographic information and SIFT feature matching method, the problems of large computational complexity and low registration accuracy in the existing registration methods are solved, and the efficiency and accuracy of remote sensing image registration can be significantly improved.

[0007] To achieve the above-mentioned object of the invention, the present invention provides a high-resolution satellite remote sensing image registration method combined with geographic information, comprising the following steps:

[0008] Step S1, input a reference image Ref and an image to be registered Tar, and detect metadata information of the input image;

[0009] Step S2, screening the salient regions and completing rough positioning;

[0010] Step S3, extracting image features and performing local feature matching to obtain initial matching point pairs;

[0011] Step S4: Output the registered image through coordinate correction and transformation.

[0012] According to a technical solution of the present invention, in step S1, the metadata information of the input image detected includes image resolution, image coordinate system, geographic coordinates and affine transformation GeoTransform for mutual conversion of image coordinates. The affine transformation GeoTransform of each image consists of 6 parameters in the following form:

[0013] GT=(GT(0),GT(1),GT(2),GT(3),GT(4),GT(5))

[0014] GT(0) represents the x-coordinate of the pixel in the upper left corner of the image, GT(1) represents the pixel resolution in the east-west direction, GT(2) represents the row rotation, GT(3) represents the y-coordinate of the pixel in the upper left corner of the image, GT(4) represents the column rotation, and GT(5) represents the pixel resolution in the north-south direction.

[0015] According to a technical solution of the present invention, based on the affine transformation GeoTransform, the conversion between geographic coordinates and image coordinates is realized, and the conversion relationship is as follows:

[0016] X _geo =GT(0)+X _col *GT(1)+Y _row *GT(2)

[0017] Y _geo =GT(3)+X _col *GT(4)+Y _row *GT(5)

[0018] (X _col , Y _row ) is the pixel coordinate in the image coordinate system, (X _geo , Y _geo ) is the geographic coordinate corresponding to the pixel coordinate.

[0019] According to a technical solution of the present invention, the step S2 specifically includes:

[0020] Step S21, based on the feature information screening strategy of the reference image, obtaining a significant image block Q1 of the reference image, that is, a first candidate region Q1;

[0021] Step S22: By converting the geographic information of the reference image and the image to be registered, the candidate region Q1 is roughly positioned on the image to be registered, and the second candidate region Q2 is obtained.

[0022] According to a technical solution of the present invention, in step S21, the feature information screening strategy is mainly summarized as a saliency strategy and a distinguishability strategy, including:

[0023] The selected candidate region should be clearly visible in the reference image and have strong salient features;

[0024] The distinguishable strategy is to select features that have unique characteristics in the entire image scene.

[0025] According to a technical solution of the present invention, the step S22 specifically includes:

[0026] Step S221, completing the conversion of pixel coordinates to geographic coordinates according to the affine transformation GeoTransform of the reference image Ref based on the upper left corner coordinates and the lower right corner coordinates of the area where the first candidate area Q1 is located;

[0027] Step S222: Based on the affine transformation GeoTransform of the image to be registered Tar, the geographic coordinates are transformed into the pixel coordinate system of the image to be registered, and the second candidate area Q2 is obtained.

[0028] According to a technical solution of the present invention, in step S4, the image features of the first candidate region Q1 and the second candidate region Q2 are extracted using the SIFT feature extraction algorithm, and local feature matching is performed to obtain an initial matching point pair M. init .

[0029] According to a technical solution of the present invention, the step S4 specifically includes:

[0030] Step S41, correcting the coordinates of the initial matching point pairs and mapping them to the coordinates in the respective original image coordinate systems;

[0031] Step S42, using a random consistency algorithm to remove incorrect matching point pairs from the initial matching point pairs after coordinate correction, and calculating the affine transformation matrix M;

[0032] Step S43: Apply the affine transformation matrix M to perform affine transformation on the image to be registered, and output the registered image.

[0033] According to a technical solution of the present invention, step S4 further includes:

[0034] The root mean square error RMSE is used to measure the registration accuracy, and the calculation formula is as follows:

[0035]

[0036] Among them, n pixels SP are uniformly extracted from the image to be registered Tari , i = 1, 2, ..., n, select the corresponding n pixels DP in the reference image Ref i ,i=1,2,…,n, M represents the affine transformation matrix.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] The present invention proposes a high-resolution satellite remote sensing image registration method combined with geographic information, which fully combines auxiliary data such as geographic information of remote sensing images and adopts an efficient feature extraction algorithm. It can greatly improve the accuracy of high-resolution satellite remote sensing image registration, and solves the problems of large calculation amount and low registration accuracy in existing registration methods. The method is suitable for the registration of high-resolution optical remote sensing images, especially for high-precision registration in specific scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0040] Figure 1 A schematic diagram schematically shows a flow chart of a high-resolution satellite remote sensing image registration method combined with geographic information according to an embodiment of the present invention;

[0041] Figure 2 A schematic diagram of a first candidate region Q1 of a reference image according to an embodiment of the present invention is schematically shown;

[0042] Figure 3 The figure schematically shows the conversion from the first candidate region Q1 to the second candidate region Q2 according to one embodiment of the present invention. DETAILED DESCRIPTION

[0043] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0044] like Figures 1 to 3 As shown, a high-resolution satellite remote sensing image registration method combining geographic information of the present invention comprises the following steps:

[0045] Step S1, input a reference image Ref and an image to be registered Tar, and detect metadata information of the input image;

[0046] The reference image Ref and the image to be registered Tar are read using the GDAL geographic information library, and their metadata information is extracted, including image resolution, image coordinate system, geographic coordinates, and affine transformation coefficients GeoTransform for mutual conversion of image coordinates, which are denoted as GT ref and GT tar is the affine transformation coefficient GeoTransform of the reference image Ref and the image to be registered Tar.

[0047] In step S1, the metadata information of the input image is detected, including image resolution, image coordinate system, geographic coordinates and affine transformation GeoTransform for mutual conversion of image coordinates. The affine transformation GeoTransform of each image consists of 6 parameters in the following form:

[0048] GT=(GT(0),GT(1),GT(2),GT(3),GT(4),GT(5))

[0049] GT(0) represents the x-coordinate of the pixel in the upper left corner of the image, GT(1) represents the pixel resolution in the east-west direction, GT(2) represents the row rotation, GT(3) represents the y-coordinate of the pixel in the upper left corner of the image, GT(4) represents the column rotation, and GT(5) represents the pixel resolution in the north-south direction.

[0050] Based on the affine transformation GeoTransform, the conversion between geographic coordinates and image coordinates is realized. The conversion relationship is as follows:

[0051] X _geo =GT(0)+X _col *GT(1)+Y _row *GT(2)

[0052] Y _geo =GT(3)+X _col *GT(4)+Y _row *GT(5)

[0053] (X _col , Y _row ) is the pixel coordinate in the image coordinate system, (X _geo , Y _geo ) is the geographic coordinate corresponding to the pixel coordinate.

[0054] Step S2: Screen the salient regions and complete rough positioning, including:

[0055] Step S21, based on the feature information screening strategy of the reference image, obtaining a significant image block Q1 of the reference image, that is, a first candidate region Q1;

[0056] Feature information screening strategies can be summarized as saliency strategy and distinguishability strategy. The selected candidate regions should be clearly visible in the reference image and have strong saliency characteristics, such as corner points, edge contours and other regions with rich feature information; the distinguishability strategy is that the selected features should have unique features in the whole scene and can be effectively distinguished from other regions in the image. When searching for matching points in multiple images, the registration algorithm can quickly and accurately find the corresponding features.

[0057] like Figure 2 , schematic diagram of the position of the candidate area Q1 in the reference image obtained according to the feature information screening strategy, where the coordinates of the upper left corner of the candidate area Q1 are (x1, y1), the length and width of the candidate area are (w1, h1), so the coordinates of the lower right corner of Q1 are (x1+w1, y1+h1), and the values ​​of w1 and h1 here are both set to 1024.

[0058] Step S22, by converting the geographic information of the reference image and the image to be registered, roughly positioning the candidate area Q1 on the image to be registered, and obtaining the second candidate area Q2, specifically includes:

[0059] Step S221, completing the conversion of pixel coordinates to geographic coordinates according to the affine transformation GeoTransform of the reference image Ref based on the upper left corner coordinates and the lower right corner coordinates of the area where the first candidate area Q1 is located;

[0060] Step S222: Based on the affine transformation GeoTransform of the image to be registered Tar, the geographic coordinates are transformed into the pixel coordinate system of the image to be registered, and the second candidate area Q2 is obtained.

[0061] like Figure 3 , the candidate area Q1 is converted into the candidate area Q2 of the image to be registered after the geographic information is converted. The positioning of the candidate area Q2 is specifically achieved by converting the upper left corner coordinates (x1, y1) and the lower right corner coordinates (x1+w1, y1+h1) of the candidate area Q1 into the upper left corner coordinates (x2, y2) and the lower right corner coordinates (x2+w2, y2+h2) of the candidate area Q2.

[0062] Step S3, extracting image features and performing local feature matching to obtain initial matching point pairs;

[0063] Since SIFT features have good feature representation capabilities and scale invariance, the SIFT feature extraction algorithm is used to extract image features from candidate regions Q1 and Q2, and local feature matching is performed to obtain the initial matching point pair M init ;

[0064] Step S4: outputting the registered image through coordinate correction and transformation, specifically including:

[0065] Step S41, correcting the coordinates of the initial matching point pairs and mapping them to the coordinates in the respective original image coordinate systems;

[0066] Since SIFT is extracted based on the candidate area, and the coordinates of the extracted SIFT feature points are based on the upper left corner of the candidate area as the origin of the pixel coordinate system, it is necessary to correct the coordinates of the initial matching point pairs and map them to the coordinates of the respective original image coordinate systems. In theory, only the translation calculation of the coordinate points needs to be performed, where the feature matching points of the reference image need to be processed by the translation coefficient (x1, y1), and the coordinates of the feature matching points of the image to be registered need to be processed by the translation coefficient (x2, y2).

[0067] Step S42, using a random consistency algorithm to remove incorrect matching point pairs from the initial matching point pairs after coordinate correction, and calculating the affine transformation matrix M;

[0068] Step S43: Apply the affine transformation matrix M to perform affine transformation on the image to be registered, and output the registered image.

[0069] Step S4 also includes:

[0070] The root mean square error RMSE is used to measure the registration accuracy, and the calculation formula is as follows:

[0071]

[0072] Among them, n pixels SP are uniformly extracted from the image to be registered Tar i , i = 1, 2, ..., n, select the corresponding n pixels DP in the reference image Ref i ,i=1,2,…,n, M represents the affine transformation matrix.

[0073] In the present invention, the terms "include", "comprises" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or terminal device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or terminal device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or terminal device including the element.

[0074] Finally, it should be noted that the above is a preferred embodiment of the present invention. It should be pointed out that although the preferred embodiment of the present invention has been described, for those skilled in the art, once the basic creative concept of the present invention is known, several improvements and modifications can be made without departing from the principles of the present invention, and these improvements and modifications should also be regarded as the protection scope of the present invention. Therefore, the attached claims are intended to be interpreted as including the preferred embodiment and all changes and modifications that fall within the scope of the embodiments of the present invention.

Claims

1. A high-resolution satellite remote sensing image registration method combined with geographic information, characterized in that: The following steps are involved: Step S1, input a reference image Ref and an image to be registered Tar, and detect metadata information of the input image; Step S2, screening the salient regions and completing rough positioning; Step S3, extracting image features and performing local feature matching to obtain initial matching point pairs; Step S4: Output the registered image through coordinate correction and transformation.

2. The high-resolution satellite remote sensing image registration method combined with geographic information according to claim 1, characterized in that: In step S1, the metadata information of the input image is detected, including image resolution, image coordinate system, geographic coordinates and affine transformation GeoTransform for mutual conversion of image coordinates. The affine transformation GeoTransform of each image consists of 6 parameters in the following form: GT=(GT(0),GT(1),GT(2),GT(3),GT(4),GT(5)) GT(0) represents the x-coordinate of the pixel in the upper left corner of the image, GT(1) represents the pixel resolution in the east-west direction, GT(2) represents the row rotation, GT(3) represents the y-coordinate of the pixel in the upper left corner of the image, GT(4) represents the column rotation, and GT(5) represents the pixel resolution in the north-south direction.

3. The high-resolution satellite remote sensing image registration method combined with geographic information according to claim 2, characterized in that: Based on the affine transformation GeoTransform, the conversion between geographic coordinates and image coordinates is realized. The conversion relationship is as follows: X _geo =GT(0)+X _col *GT(1)+Y _row *GT(2) Y _geo =GT(3)+X _col *GT(4)+Y _row *GT(5) (X _col , Y _row ) is the pixel coordinate in the image coordinate system, (X _geo , Y _geo ) is the geographic coordinate corresponding to the pixel coordinate.

4. The high-resolution satellite remote sensing image registration method combined with geographic information according to claim 3, characterized in that: The step S2 specifically includes: Step S21, based on the feature information screening strategy of the reference image, obtaining a significant image block Q1 of the reference image, that is, a first candidate region Q1; Step S22: By converting the geographic information of the reference image and the image to be registered, the candidate region Q1 is roughly positioned on the image to be registered, and the second candidate region Q2 is obtained.

5. The high-resolution satellite remote sensing image registration method combined with geographic information according to claim 4, characterized in that: In step S21, the feature information screening strategy is mainly summarized as a saliency strategy and a distinguishability strategy, including: The selected candidate region should be clearly visible in the reference image and have strong salient features; The distinguishable strategy is to select features that have unique characteristics in the entire image scene.

6. The high-resolution satellite remote sensing image registration method combined with geographic information according to claim 4, characterized in that: The step S22 specifically includes: Step S221, completing the conversion of pixel coordinates to geographic coordinates according to the affine transformation GeoTransform of the reference image Ref based on the upper left corner coordinates and the lower right corner coordinates of the area where the first candidate area Q1 is located; Step S222: Based on the affine transformation GeoTransform of the image to be registered Tar, the geographic coordinates are transformed into the pixel coordinate system of the image to be registered, and the second candidate area Q2 is obtained.

7. The high-resolution satellite remote sensing image registration method combined with geographic information according to claim 1, characterized in that: In step S4, the image features are extracted from the first candidate region Q1 and the second candidate region Q2 using the SIFT feature extraction algorithm, and local feature matching is performed to obtain an initial matching point pair M. init .

8. The high-resolution satellite remote sensing image registration method combined with geographic information according to claim 1, characterized in that: The step S4 specifically includes: Step S41, correcting the coordinates of the initial matching point pairs and mapping them to the coordinates in the respective original image coordinate systems; Step S42, using a random consistency algorithm to remove incorrect matching point pairs from the initial matching point pairs after coordinate correction, and calculating the affine transformation matrix M; Step S43: Apply the affine transformation matrix M to perform affine transformation on the image to be registered, and output the registered image.

9. The high-resolution satellite remote sensing image registration method combined with geographic information according to claim 8, characterized in that: The step S4 also includes: The root mean square error RMSE is used to measure the registration accuracy, and the calculation formula is as follows: Among them, n pixels SP are uniformly extracted from the image to be registered Tar i , i = 1, 2, ..., n, select the corresponding n pixels DP in the reference image Ref i ,i=1,2,…,n, M represents the affine transformation matrix.

Images