CN117058008A

CN117058008A - Remote sensing image geometry and radiation integrated correction method, device, equipment and medium

Info

Publication number: CN117058008A
Application number: CN202310655699.4A
Authority: CN
Inventors: 谭福宏; 刘洋; 王明省; 秦亮军; 余锐; 张郁; 丁晶; 王楠; 李爽; 吴辉
Original assignee: Guangzhou Urban Planning Survey and Design Institute
Current assignee: Guangzhou Urban Planning Survey and Design Institute
Priority date: 2023-06-02
Filing date: 2023-06-02
Publication date: 2023-11-14

Abstract

The invention discloses a remote sensing image geometry and radiation integrated correction method, a device, equipment and a medium, wherein the method comprises the following steps: clipping the history orthophoto image based on the image control points in the area range to obtain a plurality of point location index images taking the image control points as the center and constructing an image control point database; according to the geographic range of the image to be corrected, acquiring a plurality of target image control points positioned in the geographic range and determining corresponding point position index images in the image to be corrected; and matching the point index image of each target image control point with the corresponding point index image by using a SIFT feature matching method to obtain an image control point homonymous image point pair set and a non-image control point homonymous image point pair set, calculating geometric correction parameters by using a rational function model, calculating radiation correction parameters by using a constant target method and a least square linear regression method, and respectively carrying out geometric and radiation correction on the image to be corrected by using a geometric correction model and a radiation correction model. The invention can improve the efficiency and the precision of the geometric and radiation correction of the remote sensing image.

Description

Remote sensing image geometry and radiation integrated correction method, device, equipment and medium

Technical Field

The invention relates to the technical field of remote sensing image processing, in particular to a remote sensing image geometry and radiation integrated correction method, a device, terminal equipment and a computer readable storage medium.

Background

For the original remote sensing image obtained by the sensor, geometric correction and radiation correction are required for the original remote sensing image due to geometric distortion and radiation distortion. Geometric correction is the task of correcting and eliminating geometric errors on remote sensing images. Geometric errors refer to deformation of geometric features such as the size, shape, position and the like of each feature, which are inconsistent with theoretical values under a reference system. The geometrical deformation is caused by various reasons, and is generally caused by deformation of an image material, distortion of an objective lens, refraction of the atmosphere, curvature of the earth surface, rotation of the earth, fluctuation of the topography, and the like at the time of imaging. The geometric correction can be subdivided into absolute geometric correction and relative geometric correction, wherein the absolute geometric correction is to correct an image to be corrected by taking a ground control point as a reference; the relative geometric correction is to correct the remote sensing image geometrically corrected in the same area as the reference. Radiation correction is the task of correcting and eliminating radiation errors on remote sensing images. The radiation correction can be subdivided into absolute radiation correction, which is corrected with reference to the radiation value measured in the field in a certain area on the image to be corrected, and relative radiation correction, which is corrected with reference to the remote sensing image of the same area, which has been corrected by radiation.

In the prior art, a step-by-step correction method is generally adopted for geometric correction and radiation correction, the image control point selection of high-precision geometric correction depends on a manual stabbing mode, the point selection precision depends on manual experience, the correction efficiency is low, the correction precision cannot be ensured, a dodging technology is generally adopted for radiation correction, only visual color consistency is met, real radiation information of the ground surface is difficult to reflect, and the correction precision is low.

Disclosure of Invention

The invention provides a geometric and radiation integrated correction method, a device, equipment and a medium for remote sensing images, which are used for constructing an image control point database by utilizing historical orthographic images which are subjected to geometric correction and radiation correction, matching point index images corresponding to all target image control points with corresponding point index images by combining a SIFT feature matching method to obtain an image control point homonymous image point pair set and a non-image control point homonymous image point pair set, thereby realizing the geometric and radiation integrated automatic correction of the remote sensing images, needing no relying on artificial stabs, improving the correction efficiency and correction precision, and in addition, obtaining radiation correction parameters for radiation correction by utilizing a invariant target method and a least square linear regression method based on the non-image control point homonymous image point pair set, reflecting the ground surface real radiation information and obviously improving the correction precision.

In order to solve the above technical problems, a first aspect of the embodiments of the present invention provides a method for integrally correcting geometry and radiation of a remote sensing image, including the following steps:

based on a plurality of image control points in a preset area range and a history orthophoto image subjected to geometric correction and radiation correction, cutting the history orthophoto image to obtain a plurality of point location index images taking the image control points as the center, and constructing an image control point database according to the plurality of image control points and the point location index images;

according to the geographic range corresponding to the remote sensing image to be corrected, acquiring a plurality of target image control points positioned in the geographic range from the image control point database, and acquiring corresponding point position index images of each target image control point in the remote sensing image to be corrected;

matching point index images corresponding to each target image control point with each corresponding point index image by using a SIFT feature matching method to obtain an image control point homonymous image point pair set and a non-image control point homonymous image point pair set;

based on the image control point homonymy image point pair set, calculating by using a preset rational function model to obtain a geometric correction parameter;

based on the non-image control point homonymous image point pair set, calculating by using a constant target method and a least square linear regression method to obtain a radiation correction parameter;

And according to the geometric correction parameters and the radiation correction parameters, respectively utilizing a preset geometric correction model and a radiation correction model to sequentially perform geometric correction and radiation correction on the remote sensing image to be corrected, so as to obtain a target correction remote sensing image.

As a preferred solution, the obtaining the index image of the corresponding point position of each target image control point in the remote sensing image to be corrected specifically includes the following steps:

acquiring position information of each target image control point in the remote sensing image to be corrected, and determining a plurality of candidate image control points which are not positioned in a background area of the remote sensing image to be corrected according to the position information corresponding to each target image control point;

acquiring corresponding point position index images of each candidate image control point in the remote sensing image to be corrected according to the width, the height and the resolution of the point position index image corresponding to each candidate image control point, the resolution of the remote sensing image to be corrected, a preset image width compensation value and an image height compensation value; the geographic range corresponding to the point location index image of any one candidate image control point is larger than or equal to the geographic range corresponding to the point location index image of the any one candidate image control point;

Then, matching the point index image corresponding to each target image control point with each corresponding point index image by using a SIFT feature matching method to obtain an image control point homonymous image point pair set and a non-image control point homonymous image point pair set, which specifically are:

and matching the point index image corresponding to each candidate image control point with each corresponding point index image by using a SIFT feature matching method to obtain an image control point homonymous image point pair set and a non-image control point homonymous image point pair set.

As a preferred scheme, the method further comprises the following steps after obtaining the index image of the corresponding point position of each candidate image control point in the remote sensing image to be corrected:

judging whether the difference between the resolution of the point index image corresponding to each candidate image control point and the resolution of the image to be corrected is larger than a preset resolution difference;

when the difference between the resolution of the point location index image corresponding to any one candidate image control point and the resolution of the image to be corrected is larger than the preset resolution difference, resampling the point location index image corresponding to any one candidate image control point to enable the resolution of the point location index image corresponding to any one candidate image control point to be the same as the resolution of the image to be corrected;

When the difference between the resolution of the point index image corresponding to each candidate image control point and the resolution of the image to be corrected is smaller than or equal to the preset resolution difference, histogram prescribing is carried out on the corresponding point index image of each candidate image control point so as to enable the gray scale distribution of the point index image corresponding to any one candidate image control point to be consistent with the gray scale distribution of the point index image of any one candidate image control point.

As a preferred solution, the matching method of SIFT features is used to match the point index image corresponding to each candidate control point with each corresponding point index image to obtain a pair of image control point homonymous image points and a pair of non-image control point homonymous image points, and specifically includes the following steps:

matching the point index image corresponding to each candidate image control point with each corresponding point index image by using a SIFT feature matching method to obtain a plurality of matching point pairs between the point index image corresponding to each candidate image control point and the corresponding point index image;

calculating homography matrixes between the point index images corresponding to each candidate image control point and the corresponding point index images by using a RANSAC algorithm according to a plurality of matching point pairs between the point index images corresponding to each candidate image control point and the corresponding point index images, eliminating error matching point pairs in the plurality of matching point pairs according to the homography matrixes, obtaining a plurality of candidate matching point pairs and calculating a target homography matrix based on the plurality of candidate matching points;

Grouping a plurality of candidate matching point pairs between the point location index image corresponding to each candidate image control point and the corresponding point location index image to obtain an image control point matching point pair group and a non-image control point matching point pair group;

judging whether the number of pairs of the image control point matching points of the image control point matching point pair group is smaller than a preset number or not;

when the number of the matching point pairs of the image control points is not smaller than the preset number, calculating the image coordinates of the same-name image points of each candidate image control point in the remote sensing image to be corrected based on the target homography matrix;

when the number of the matching point pairs of the image control points is smaller than the preset number, re-matching the point position index image corresponding to each candidate image control point with the corresponding point position index image based on a geographic coordinate constraint matching strategy to obtain a plurality of correct matching point pairs between the point position index image corresponding to each candidate image control point and the corresponding point position index image, calculating a new homography matrix between the point position index image corresponding to each candidate image control point and the corresponding point position index image by re-using a RANSAC algorithm, removing the error matching point pairs in the plurality of correct matching point pairs according to the new homography matrix to obtain a plurality of new candidate matching points, calculating a new target homography matrix based on the plurality of new candidate matching points, and calculating the homonymy point image coordinates of each candidate image control point in the remote sensing image to be corrected based on the new target homography matrix;

And obtaining the image control point homonymy point pair set according to homonymy point image coordinates of each candidate image control point in the remote sensing image to be corrected, and obtaining the non-image control point homonymy point pair set according to a plurality of non-image control point matching point pairs in the non-image control point matching point pair group.

As a preferred solution, the matching strategy based on geographic coordinate constraint re-matches the point location index image corresponding to each candidate image control point with the corresponding point location index image to obtain a plurality of correct matching point pairs between the point location index image corresponding to each candidate image control point and the corresponding point location index image, which specifically includes the following steps:

according to the similarity between the SIFT feature vector of the point index image corresponding to each candidate image point and the SIFT feature vector of the corresponding point index image, determining a first similar image point and a second similar image point of each SIFT feature point in the point index image in the corresponding point index image;

acquiring longitude and latitude values of each SIFT feature point in the point index image corresponding to each candidate image point, carrying out coordinate conversion on the longitude and latitude values of each SIFT feature point by using 7 parameters, and determining corresponding image points of each SIFT feature point in the corresponding point index image;

Judging whether the distance between the first similar image point, the second similar image point and the corresponding image point corresponding to each SIFT feature point is smaller than a preset distance threshold value or not;

and when the distance between the second similar image point corresponding to any SIFT feature point and the corresponding image point is smaller than the preset distance threshold, judging that the first similar image point and the any SIFT feature point are the correct matching point pairs, and when the distance between the second similar image point corresponding to any SIFT feature point and the corresponding image point is smaller than the preset distance threshold, judging that the second similar image point and the any SIFT feature point are the correct matching point pairs until the correct matching points corresponding to all SIFT feature points are determined, and obtaining a plurality of correct matching point pairs between the point index image corresponding to each candidate image control point and the corresponding point index image according to the correct matching points corresponding to all SIFT feature points.

Preferably, the method further includes the following steps after obtaining the image coordinates of the same name image point of each candidate image control point in the remote sensing image to be corrected:

optimizing the image coordinates of each non-image control point matching point in the non-image control point matching point pair group and the image coordinates of the same-name image point of each candidate image control point in the remote sensing image to be corrected by utilizing a single-point least square algorithm to obtain the optimized image coordinates of each non-image control point matching point in the non-image control point matching point pair group and the optimized image coordinates of each candidate image control point in the remote sensing image to be corrected;

And obtaining the image control point homonymous image point pair set according to the homonymous image point coordinates of each candidate image control point in the remote sensing image to be corrected, and obtaining the non-image control point homonymous image point pair set according to a plurality of non-image control point matching point pairs in the non-image control point matching point pair group, wherein the method specifically comprises the following steps:

and obtaining the image control point homonymous image point pair set according to the optimized homonymous image point coordinates of each candidate image control point in the remote sensing image to be corrected, and obtaining the non-image control point homonymous image point pair set according to a plurality of non-image control point matching points in the non-image control point matching point pair set and the optimized image coordinates of each non-image control point matching point.

As a preferred solution, according to the geometric correction parameter and the radiation correction parameter, a preset geometric correction model and a radiation correction model are respectively used to sequentially perform geometric correction and radiation correction on the remote sensing image to be corrected, so as to obtain a target corrected remote sensing image, which specifically includes the following steps:

according to the geometric correction parameters, performing geographic coordinate transformation on any one pixel in the remote sensing image to be corrected by using the geometric correction model to obtain a remote sensing image after geometric correction;

Resampling the geometrically corrected remote sensing image by using a bilinear interpolation method to obtain a resampled remote sensing image;

and according to the radiation correction parameters, carrying out radiation value transformation on each channel radiation vector of any pixel in the resampled remote sensing image by using the radiation correction model to obtain the target corrected remote sensing image.

A second aspect of the embodiments of the present invention provides a remote sensing image geometry and radiation integrated correction device, including:

the image control point database construction module is used for cutting the history orthographic image based on a plurality of image control points in a preset area range and the history orthographic image subjected to geometric correction and radiation correction to obtain a plurality of point location index images taking the image control points as the center, and constructing an image control point database according to the plurality of image control points and the point location index images;

the corresponding point position index image acquisition module is used for acquiring a plurality of target image control points positioned in the geographical range in the image control point database according to the geographical range corresponding to the remote sensing image to be corrected, and acquiring corresponding point position index images of the target image control points in the remote sensing image to be corrected;

The image matching module is used for matching the point index image corresponding to each target image control point with each corresponding point index image by using a SIFT feature matching method to obtain an image control point homonymous image point pair set and a non-image control point homonymous image point pair set;

the geometric correction parameter acquisition module is used for obtaining geometric correction parameters by means of calculation of a preset rational function model based on the image control point homonymy image point pair set;

the radiation correction parameter acquisition module is used for acquiring radiation correction parameters by utilizing a constant target method and a least square linear regression method based on the non-image control point homonymous image point pair set;

and the image correction module is used for sequentially carrying out geometric correction and radiation correction on the remote sensing image to be corrected by using a preset geometric correction model and a radiation correction model respectively according to the geometric correction parameters and the radiation correction parameters to obtain a target correction remote sensing image.

A third aspect of the embodiment of the present invention provides a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, where the processor implements the remote sensing image geometry and radiation integrated correction method according to any one of the first aspects when the processor executes the computer program.

A fourth aspect of the embodiments of the present invention provides a computer readable storage medium, where the computer readable storage medium includes a stored computer program, where the computer program when executed controls a device in which the computer readable storage medium is located to execute the remote sensing image geometry and radiation integrated correction method according to any one of the first aspect.

Compared with the prior art, the embodiment of the invention has the beneficial effects that the imaging control point database is constructed by utilizing the history orthographic images subjected to geometric correction and radiation correction, the point index images corresponding to the target imaging control points are matched with the corresponding point index images by combining the SIFT feature matching method, and the imaging control point homonymous image point pair set and the non-imaging control point homonymous image point pair set are obtained, so that the automatic correction of the remote sensing image geometry and radiation integration is realized, the dependence on artificial stabs is avoided, the correction efficiency and correction precision are improved, and in addition, the radiation correction parameters are obtained for radiation correction by utilizing the invariant target method and the least square linear regression method based on the non-imaging control point homonymous image point pair set, the real radiation information of the earth surface can be reflected, and the correction precision is remarkably improved.

Drawings

FIG. 1 is a flow chart of a remote sensing image geometry and radiation integrated correction method according to an embodiment of the invention;

FIG. 2 is a schematic diagram of local extremum point detection of adjacent scale spaces in an embodiment of the present invention;

FIG. 3 is a schematic diagram of generation of SIFT feature vectors in an embodiment of the present invention;

FIG. 4 is a schematic diagram of the principle of geographic coordinate constraint matching in an embodiment of the invention;

FIG. 5 is a schematic diagram of a single-point least squares image matching scheme according to an embodiment of the present invention;

fig. 6 is a schematic structural diagram of a remote sensing image geometry and radiation integrated correction device in an embodiment of the invention.

Detailed Description

The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

Referring to fig. 1, a first aspect of the present invention provides a method for integrating geometric and radiation correction of a remote sensing image, which includes steps S1 to S6 as follows:

Step S1, based on a plurality of image control points in a preset area range and a history orthophoto image subjected to geometric correction and radiation correction, clipping the history orthophoto image to obtain a plurality of point location index images taking the image control points as the center, and constructing an image control point database according to the plurality of image control points and the point location index images;

step S2, according to the geographic range corresponding to the remote sensing image to be corrected, acquiring a plurality of target image control points positioned in the geographic range from the image control point database, and acquiring corresponding point position index images of each target image control point in the remote sensing image to be corrected;

step S3, matching point index images corresponding to each target image control point with each corresponding point index image by using a SIFT feature matching method to obtain an image control point homonymous image point pair set and a non-image control point homonymous image point pair set;

step S4, based on the image control point homonymy image point pair set, calculating by using a preset rational function model to obtain geometric correction parameters;

s5, based on the non-image control point homonymous image point pair set, obtaining radiation correction parameters by utilizing a constant target method and a least square linear regression method;

and S6, according to the geometric correction parameters and the radiation correction parameters, respectively utilizing a preset geometric correction model and a radiation correction model to sequentially perform geometric correction and radiation correction on the remote sensing image to be corrected, and obtaining a target correction remote sensing image.

In step S1, a plurality of image control points are laid in advance in the range of the measurement area, then the image control points are selected from the history orthographic images subjected to geometric correction and radiation correction and cut, as an optional embodiment, the cut size is 1023 x 1023 pixels, but the embodiment does not limit the cut size specifically, the cut size can be adjusted according to actual needs, a plurality of point index images taking the image control points as the center are obtained after cutting, attribute editing is performed on each point index image and the image control points thereof, and batch storage is performed to construct an image control point database.

Step S2 is a preprocessing step, wherein the preprocessing aims to acquire corresponding point position index images of each target image control point in the remote sensing image to be corrected, and ensure that the data types of the two images meet the requirement of a matching program. Specifically, according to the four-corner coordinates of the remote sensing image to be corrected, the corresponding geographic range can be determined, so that the search range in the image control point database is reduced, then according to the conditions of resolution, correction precision requirements and the like of the image, the applicability of a plurality of image control points located in the geographic range is analyzed, a plurality of target image control points meeting the conditions are finally determined, and further, the corresponding point position index image of each target image control point in the remote sensing image to be corrected is obtained.

In step S3, since the SIFT feature operator has the characteristics of unchanged scale, unchanged rotation, unchanged affine and unchanged brightness, the SIFT feature operator can well adapt to the distortion and brightness difference of the aerial image, and extracts the linearly related homonymous image point pairs, so that the embodiment matches the point index image corresponding to each target image control point with each corresponding point index image by using the SIFT feature matching method to obtain an image control point homonymous image point pair set and a non-image control point homonymous image point pair set.

It is worth to say that the SIFT feature matching process includes point location index images corresponding to each target image control point and SIFT feature points and feature descriptor sub-vectors of each corresponding point location index image. The generation step of the SIFT feature points comprises four steps of scale space extreme point detection, extreme point accurate positioning, feature point main direction determination and feature point descriptor generation.

(1) Detection of extreme points in scale space

First, a scale space is generated, the theoretical purpose of which is to simulate multi-scale features of image data. The gaussian convolution kernel is the only linear kernel that implements the scaling, and thus the scale space of a two-dimensional image is defined as:

L(x,y,σ)＝G(x,y,σ)*I(x,y) (1)

wherein G (x, y, σ) represents a variable-scale gaussian function:

(x, y) represents spatial coordinates, and σ represents scale coordinates.

In order to effectively detect stable key points in the scale space, a Gaussian differential scale-space (DOG scale-space) is established, and the Gaussian differential kernels with different scales are used for convolution generation with the image.

D(x,y,σ)＝(G(x,y,kσ)-G(x,y,σ))*I(x,y)＝L(x,y,kσ)-L(x,y,σ) (3)

To find the extreme points of the scale space, each sample point is compared with all its neighbors to see if it is larger or smaller than its neighbors in the image and scale domains. As shown in fig. 2, the middle detection point is compared with its 8 adjacent points of the same scale and the 9×2 points corresponding to the up-down adjacent scales for 26 points in total to ensure that extreme points are detected in both the scale space and the two-dimensional image space.

(2) Accurate positioning of extreme points

The positions and scales of the key points are accurately determined (sub-pixel precision is achieved) through fitting three-dimensional quadratic functions, and meanwhile, the key points with low contrast and unstable edge response points (because the DoG operator can generate stronger edge response) are removed, so that the matching stability is improved, and the noise immunity is improved. Extrema of gaussian difference operators, which are generally not good, have a larger principal curvature across the edges and a smaller principal curvature in the direction perpendicular to the edges. Unstable edge response points can be removed by the ratio of principal curvature to vertical curvature.

(3) Principal direction determination of feature points

And designating a direction parameter for each key point by utilizing the gradient direction distribution characteristic of the key point neighborhood pixels, so that the operator has rotation invariance.

The above formula is a model value and a direction formula of the gradient at (x, y), wherein the scale used by L is the scale of each key point. In actual calculation, sampling is carried out in a neighborhood window with a key point as a center, and the gradient direction of the neighborhood pixels is counted by using a histogram. The gradient histogram ranges from 0 to 360 degrees with one bin per 10 degrees for a total of 36 bins. The peak of the histogram then represents the main direction of the neighborhood gradient at the keypoint, i.e. the direction that is the keypoint.

In the gradient direction histogram, when there is another peak corresponding to 80% of the energy of the main peak, then this direction is considered as the auxiliary direction of the key point. A key point may be designated to have multiple directions (one primary direction, more than one secondary direction), which may enhance the robustness of the matching. So far, the key points of the image are detected, and each key point has three pieces of information: position, scale, direction, a SIFT feature region can be determined.

(4) Feature point descriptor generation

The coordinate axis is first rotated in the direction of the key point to ensure that the rotation is not changed. Next, an 8×8 window is taken centered around the keypoint. As shown in fig. 3, the central black dot in the left part is the position of the current keypoint, each cell represents a pixel in the scale space where the neighborhood of the keypoint is located, the arrow direction represents the gradient direction of the pixel, the arrow length represents the gradient modulus value, and the circle in the figure represents the gaussian weighted range (the closer to the keypoint, the greater the contribution of the pixel gradient direction information). Then calculating gradient direction histograms of 8 directions on each 4×4 small block, and drawing accumulated values of each gradient direction to form a seed point. One key point in this figure consists of a total of 4 seed points of 2×2, each seed point having 8 direction vector information. This idea of neighborhood directional information union enhances the noise immunity of the algorithm, while also providing better fault tolerance for feature matching that contains positioning errors. In the actual calculation process, in order to enhance the robustness of matching, a total of 16 seed points are used for describing each key point, so that 128 data can be generated for one key point, namely, 128-dimensional SIFT feature vectors are finally formed. At this time, the SIFT feature vector has removed the influence of geometric deformation factors such as scale change and rotation, and then the length of the feature vector is normalized, so that the influence of illumination change can be further removed.

Further, after generating the SIFT feature vector, the euclidean distance of the two SIFT feature vectors may be utilized as a similarity measure between the points to be matched. And judging whether the two points are the same-name image points or not by calculating the ratio of the nearest Euclidean distance and the next nearest Euclidean distance between the target point and all candidate point feature descriptors, wherein the ratio calculation formula is as follows. If the ratio is less than the threshold, the two points may be considered to be homonymous points, otherwise unacceptable. In order to reduce the number of mismatch points, the threshold value may be appropriately reduced.

Based on the acquired plurality of matching homonymous image point pairs, an image control point homonymous image point pair set and a non-image control point homonymous image point pair set can be obtained in a dividing mode.

In step S4, based on the image control point homonymy image point pair set, the geometric correction parameter is obtained by using a preset rational function model for calculation.

It should be noted that, geometric correction is also called orthographic correction, which is a process of geometric distortion of an image, and is caused by external factors such as topography and sensors. The result of the orthophoto correction is a real remote sensing image. The orthographic correction is to resample the image into an orthographic image by selecting some ground control points on the photo and utilizing the DEM data of the original photo and correcting the inclination and projection difference of the image. And splicing and embedding a plurality of orthographic images, and obtaining the orthographic image by cutting out the images within a certain range after color balance treatment.

The present embodiment uses a rational function model (RFM, rational Function Model) based on rational polynomial coefficients (RPC, rational Polynomial Coefficients) for the resolution of the geometrical correction parameters.

The RFM model is a generalized sensor imaging correction model, and can achieve a correction effect similar to a strict physical model needing to obtain internal core parameters of a satellite and basically consistent in precision, so that the RFM model can replace a complex strict physical correction model in many remote sensing image processing works. The RFM model has more accurate and more common expression form relative to various sensor geometric models, but an error compensation model is needed to compensate RPC parameter errors of the image.

The RPC model is to associate the image coordinates P (r, c) with the ground coordinates P (X, Y, Z) to establish the mathematical relationship of the ratio polynomial, and the expression form is as follows:

where P (X, Y, Z) is a polynomial function whose power of X, Y, Z is at most 3 for each term, and the sum of the power values of X, Y, Z for each term is not higher than 3, typically taking the values of three of 1, 2, and 3. The specific expression is as follows:

P(X，Y，Z)＝a ₀ +a ₁ X+a ₂ Y+a ₃ Z+a ₄ XY+a ₅ XZ+a ₆ YZ+a ₇ X ² +a ₈ Y ² +a ₉ Z ² +a ₁₀ XYZ+a ₁₁ X ² Y+a ₁₂ X ² Z+a ₁₃ Y ² X+a ₁₄ Y ² X+a ₁₅ XZ ² +a ₁₆ YZ ² +a ₁₇ X ³ +a ₁₈ Y ³ +a ₁₉ Z ³

wherein a is ₀ ，a ₁ ，a ₂ ，……，a ₁₉ Is a coefficient of a rational function. Similarly, P ₂ ，P ₃ ，P ₄ Usable b _i ，c _i ，d _i Polynomial representation of b) ₀ And d ₀ Typically "1".

In the calculation, rounding errors are often generated due to the gap between the data size levels, so in order to maintain the stability of the calculation process and reduce the errors, the image coordinates (r, c) and the ground coordinates (X, Y, Z) need to be regularized by scaling and translation to obtain standardized coordinates with a value range between (-1, 1), and the transformation form is as follows:

wherein (X) ₀ ，Y ₀ ，Z ₀ ，r ₀ ，c ₀ ) Is a standardized translation parameter; (X) _s ，，Y _s ，Z _s ，r _s ，c _s ) Is a standardized scale parameter. (X) _n ，Y _n ，Z _n ，r _c ，c _n ) Is the normalized coordinates.

In step S5, the radiation correction parameters are obtained by using a constant target method and a least square linear regression method based on the non-image control point homonymous image point pair set.

It is worth noting that the invariant targeting method was first proposed by Schott in 1988. The core of the method is a pseudo-invariant target (PIF), which refers to a pixel with relatively stable radiation characteristics and definite geographic significance on a remote sensing image, and the reflection radiation characteristics of the PIF have small change within a certain time, so that the PIF can be used as a radiation reference for radiation correction of a multi-temporal remote sensing image. The earliest proposed objective of PIF technology was to achieve scene-to-scene normalization by implementing some kind of radiation conversion to maintain the radiation consistency of the images in both scenes. This technique is based on basic radiation theory and can ultimately be defined in terms of the earth's surface reflectivity properties. It is assumed that the radiation reaching a given spectral channel of an on-board or satellite sensor can be expressed as a linear function of reflectivity, based on which the relation between the gray values of the multi-temporal image can be deduced:

DN _1i ＝m _i ×DN _2j +n _i (6)

Wherein DN is _1i For the image gray value, DN, of the target image band i _2j For the image gray value, m of the reference image band j _i 、n _i For regression equation coefficients, i=j when the target image and the reference image are homologous, and i and j when the target image and the reference image are heterologous are the wave bands corresponding to the same or similar wavelength ranges.

Based on the above, PIF point gray values in the reference image and the remote sensing image to be corrected are substituted into a regression equation to calculate a correction coefficient, so that the relative radiation correction between the multi-temporal images can be performed, and if the reference image reflects the ground surface real radiation information, the effect of absolute radiation correction can be approximated.

Linear regression refers to building a linear model over an existing dataset to fit the relationship between components in the feature vector of the dataset. The expected outcome of the "new data" can be predicted by the already fitted linear model. The two-dimensional linear model is a straight line, and the three-dimensional linear model is a plane. At present, regarding the algorithm implementation of linear regression, the least square method is one of the most widely used algorithms. For a dataset, the data relationship is fitted by using a linear model, and the method for constructing the linear model can be realized relative to one-dimensional or multi-dimensional data.

According to the principle of the invariant target method, the linear relation between the remote sensing image to be corrected and the radiation reference image is established through least square linear regression by utilizing the extracted non-image control point homonymous image point pair sets, and the radiation correction parameters are calculated.

In step S6, according to the geometric correction parameter and the radiation correction parameter, the geometric correction and the radiation correction are sequentially performed on the remote sensing image to be corrected by using the preset geometric correction model and the radiation correction model, so as to obtain the target corrected remote sensing image.

Specifically, the position information of each target image control point in the remote sensing image to be corrected is obtained, the position information is a rough position, for the point location index image corresponding to each target image control point, the rough position of the target image control point on the remote sensing image to be corrected can be calculated according to the geodetic coordinates of the target image control point and the RPC model of the remote sensing image to be corrected, the deviation of the target image control point from the real position is few, namely tens of pixels, more than hundreds of pixels, and a plurality of candidate image control points which are not in the background area of the remote sensing image to be corrected are determined.

Further, since the remote sensing image to be corrected often has a large data size, some of the remote sensing images are as large as 2G, and cannot be read into the memory at one time, in order to facilitate program processing, in this embodiment, a remote sensing image with a certain size is taken with each candidate image control point as a center, and the remote sensing image is used as a corresponding point position index image of each candidate image control point in the remote sensing image to be corrected, and the calculation formula of the corresponding point position index image is as follows:

Wherein w is ₁ 、h ₁ The width and the height of the point position index image corresponding to each candidate image control point are calculated; res _x1 、res _y1 The pixel number of the point index image in the horizontal direction and the pixel number of the point index image in the vertical direction are obtained; w (w) ₂ 、h ₂ Indexing the width and the height of the image for the corresponding point position; res _x2 、res _y2 The pixel number of the remote sensing image to be corrected in the horizontal direction and the pixel number of the remote sensing image to be corrected in the vertical direction; Δw and Δh are preset image width compensation values and image height compensation values.

It should be noted that, due to the variation of the illumination condition, the shooting angle or the ground feature, the gray values of the point index image corresponding to each candidate image control point and the corresponding point index image may have a large difference, so that the same-name image point is a SIFT feature point on one image and is a non-SIFT feature point on the other image. This situation may deteriorate the image matching effect or even fail the matching. In order to avoid the influence of the difference of the gray values of the images on the image matching result, the distribution of the radiation gray values of the point index image and the corresponding point index image is required to be consistent. Because the point index image is an orthographic image, the gray value of the image can reflect the real ground object radiation more after geometric correction and radiation correction. Therefore, in this embodiment, the histogram specification is performed on the corresponding point location index image of each candidate image control point, so that the gray level distribution of the corresponding point location index image of any one candidate image control point is consistent with the gray level distribution of the point location index image of the any one candidate image control point.

It should be noted that, in order to improve the reliability of the matching point pairs, the present embodiment eliminates the wrong matching point pairs by using a RANSAC (random sample consistency check) algorithm.

The RANSAC algorithm is a random parameter estimation algorithm. The RANSAC algorithm randomly extracts a sample subset from samples, calculates model parameters for the subset by using a minimum variance estimation algorithm, calculates the deviation between all samples and the model, compares the deviation with a preset threshold value, and when the deviation is smaller than the threshold value, the sample point belongs to a sample point in the model, otherwise, the sample point is a sample point outside the model, records the number of the current inner points, and repeats the process. The best model parameters are recorded at each repetition, namely the best model parameters are the most of the inner points, and the model parameters are the final model parameter estimated values.

In image matching, the RANSAC algorithm calculates a homography matrix H of two images, where the homography matrix H represents a conversion relationship between homonymous pixels of the two images. The principle of the RANSAC algorithm for rejecting the mismatching point pairs is as follows:

(1) Randomly selecting 8 matching point pairs, and estimating a homography matrix H;

(2) Determining the logarithm of the matching point pair supporting the calculated homography matrix H, and taking the logarithm as a consistent set;

(3) If the number of the matching point pairs is larger than the set threshold value, re-estimating the homography matrix H by using all the matching points in the consistent set, eliminating the matching point pairs which do not support the homography matrix H, taking the rest point pairs as the point pairs which are correctly matched, otherwise, returning to the step (1);

(4) And if the sampling times reach the threshold value, discarding the estimation of the homography matrix H, and considering that the two images are not correctly matched with the point pairs.

And finally obtaining a plurality of candidate matching point pairs.

As shown in fig. 4, as a preferred solution, the matching policy based on geographic coordinate constraint re-matches the point location index image corresponding to each candidate image control point with the corresponding point location index image to obtain a plurality of correct matching point pairs between the point location index image corresponding to each candidate image control point and the corresponding point location index image, which specifically includes the following steps:

It should be noted that, the image coordinates of the same-name image points calculated according to the homography matrix may have errors of several pixels, so that the matching result is optimized by using the single-point least square algorithm in this embodiment, and finally, the matching result of the sub-pixel level is obtained.

The basic principle of single-point least square image matching is as follows:

the geometric distortion between two-dimensional images is not only relative displacement, but also graphical changes. As shown in fig. 5, the left image is a rectangular image window, and the right image is a random quadrangle. Only if the geometric deformation of the image is sufficiently considered, the best image matching can be obtained. However, since the size of the image matching window is small, only one distortion is generally considered:

x ₂ ＝a ₀ +a ₁ x+a ₂ y (9)

y ₂ ＝b ₀ +b ₁ x+b ₂ y (10)

sometimes only affine deformation or one orthomorphism transformation is considered. If the linear gray scale distortion of the right image relative to the left image is considered at the same time, the following can be obtained:

g ₁ (x,y)+n ₁ (x,y)＝h ₀ +h ₁ g ₂ (a ₀ +a ₁ x+a ₂ y,b ₀ +b ₁ x+b ₂ y)+n ₂ (x,y)

after linearization, the error equation of least square image matching can be obtained:

v＝c ₁ dh ₀ +c ₂ dh ₁ +c ₃ da ₀ +c ₄ da ₁ +c ₅ da ₂ +c ₆ db ₀ +c ₇ db ₁ +c ₈ db ₂ -Δg

where the unknown dh ₀ ，dh ₁ ，…，db ₂ Is the correction value of undetermined parameters, and their initial values are respectively:

h ₀ ＝0，h ₁ ＝1

a ₀ ＝0，a ₁ ＝1，a ₂ ＝0

b ₀ ＝0，b ₁ ＝0，b ₂ ＝1

the observed value Δg is the gray level difference of the corresponding pixel, and the coefficient of the error equation is:

c ₁ ＝1

c ₂ ＝g ₂

In digital image matching, gray scales are all discrete arrays arranged according to a regular grid, and sampling intervals are constant delta, which can be regarded as unit length, so that partial derivatives in the above formula are replaced by differences:

an error equation is established for each pixel (within the target area) in the form of a matrix:

V＝CX-L

X＝[dh ₀ dh ₁ da ₀ da ₁ da ₂ db ₀ db ₁ db ₂ ] ^T

in establishing the error equation, a local coordinate system with the center of the target area as the origin of coordinates may be used. Error equation building method equation:

(C ^T C)X＝(C ^T L)

the specific steps of single-point least square image matching are as follows:

(1) And correcting geometric deformation. Transforming the photo coordinates of the left image window to the right image array according to the geometric deformation correction parameters:

x ₂ ＝a ₀ +a ₁ x+a ₂ y

y ₂ ＝b ₀ +b ₁ x+b ₂ y

(2) Resampling. Because the scaled coordinates are generally not possible to be the integer row and column numbers in the right image array, resampling is necessary, from resampling g ₂ (x ₂ ，y ₂ ) Obtained. In general, resampling may employ bilinear interpolation.

(3) And correcting radiation distortion. And (3) utilizing the radiation distortion correction parameters obtained by least square image matching to carry out radiation correction on the resampling result.

(4) And calculating a correlation coefficient between the left image window and the gray scale array of the right image window subjected to geometric and radiation correction, and judging whether iteration needs to be continued or not. Generally, an iteration is considered to be ended if the correlation coefficient is smaller than the correlation coefficient obtained after the previous iteration. The iteration is further determined to be over, or it may be determined whether the geometric deformation parameter (in particular the shift correction value) is smaller than a predetermined threshold.

(5) And solving the correction value of the deformation parameter by adopting least square image matching.

(6) And calculating deformation parameters. Since the correction value of the deformation parameter is obtained according to the right image gray scale array after geometric and radiation correction, the deformation parameter should be obtained according to the following algorithm:

for radiation distortion parameters:

(7) And calculating the point position of the best match. The purpose of image matching is to obtain homonymous image points. A target image window is usually established with a target point to be determined, namely, the center point of the window is the target point. However, in high-precision image correlation, it is necessary to consider whether the center point of the target window is the best matching point. The accuracy theory of least square matching can be known: the matching accuracy depends on the gradient affecting the gray scale. Thus, the coordinates can be weighted averaged over the left image window, weighted by the square of the gradient. Taking the coordinate as the target point coordinate, the coordinate of the same name image point can be obtained by the geometric transformation parameter obtained by least square image matching.

Specifically, according to a geometric correction transformation function T (r, c) in the geometric correction parameters, any one pixel (r, c) in the remote sensing image to be corrected is transformed by using a geometric correction model, so that unique real geographic coordinates (X, Y, Z) are obtained. In the geometric correction, because of the geometric distortion of the remote sensing image to be corrected, the image needs to be resampled after geometric correction, and the resampled remote sensing image is obtained by resampling the geometric corrected remote sensing image by using a bilinear interpolation method in the embodiment. Then according to the radiation correction transformation function R (L (R, c)) in the radiation correction parameters, transforming the radiation vector L (R, c) of each channel of any pixel (R, c) in the resampled remote sensing image by using the radiation correction model, thereby obtaining a corrected radiation value L ₁ (r，c)。

According to the geometric and radiation integrated correction method for the remote sensing image, an image control point database is built by utilizing the historical orthographic images subjected to geometric correction and radiation correction, point index images corresponding to all target image control points are matched with corresponding point index images by combining a SIFT feature matching method, an image control point homonymous image point pair set and a non-image control point homonymous image point pair set are obtained, so that the geometric and radiation integrated automatic correction of the remote sensing image is achieved, manual stabbing is not needed, correction efficiency and correction precision are improved, and in addition, radiation correction parameters are obtained for radiation correction by utilizing a invariant target method and a least square linear regression method based on the non-image control point homonymous image point pair set, so that ground surface real radiation information can be reflected, and correction precision is remarkably improved.

Referring to fig. 6, a second aspect of the embodiment of the present invention provides a remote sensing image geometry and radiation integrated correction device, including:

the image control point database construction module 601 is configured to cut out a historical orthographic image that has undergone geometric correction and radiation correction based on a plurality of image control points within a preset area range, obtain a plurality of point location index images centered on the image control points, and construct an image control point database according to the plurality of image control points and the point location index images;

The corresponding point location index image obtaining module 602 is configured to obtain a plurality of target image control points located in a geographic range in the image control point database according to the geographic range corresponding to the remote sensing image to be corrected, and obtain corresponding point location index images of each target image control point in the remote sensing image to be corrected;

the image matching module 603 is configured to match the point index image corresponding to each target image control point with each corresponding point index image by using a SIFT feature matching method, so as to obtain an image control point homonymous image point pair set and a non-image control point homonymous image point pair set;

the geometric correction parameter obtaining module 604 is configured to obtain geometric correction parameters by using a preset rational function model to calculate based on the image control point homonymy point pair set;

a radiation correction parameter obtaining module 605, configured to obtain radiation correction parameters based on the non-image control point homonymous image point pair set by using a invariant objective method and a least square linear regression method;

the image correction module 606 is configured to sequentially perform geometric correction and radiation correction on the remote sensing image to be corrected by using a preset geometric correction model and a radiation correction model according to the geometric correction parameter and the radiation correction parameter, so as to obtain a target corrected remote sensing image.

It should be noted that, the remote sensing image geometry and radiation integrated correction device provided by the embodiment of the present invention can implement all the processes of the remote sensing image geometry and radiation integrated correction method described in any one of the embodiments, and the functions and the implemented technical effects of each module in the device are respectively the same as those of the remote sensing image geometry and radiation integrated correction method described in the embodiment, and are not repeated herein.

A third aspect of the embodiments of the present invention provides a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, where the processor implements the remote sensing image geometry and radiation integrated correction method according to any of the embodiments of the first aspect when the processor executes the computer program.

The terminal equipment can be computing equipment such as a desktop computer, a notebook computer, a palm computer, a cloud server and the like. The terminal device may include, but is not limited to, a processor, a memory. The terminal device may also include input and output devices, network access devices, buses, and the like.

The processor may be a central processing unit (Central Processing Unit, CPU), other general purpose processors, digital signal processors (Digital Signal Processor, DSP), application specific integrated circuits (Application Specific Integrated Circuit, ASIC), field programmable gate arrays (Field-Prog rammable Gate Array, FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or the like. The general purpose processor may be a microprocessor or the processor may be any conventional processor or the like, which is a control center of the terminal device, and which connects various parts of the entire terminal device using various interfaces and lines.

The memory may be used to store the computer program and/or module, and the processor may implement various functions of the terminal device by running or executing the computer program and/or module stored in the memory and invoking data stored in the memory. The memory may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program (such as a sound playing function, an image playing function, etc.) required for at least one function, and the like; the storage data area may store data (such as audio data, phonebook, etc.) created according to the use of the handset, etc. In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as a hard disk, memory, plug-in hard disk, smart Media Card (SMC), secure Digital (SD) Card, flash Card (Flash Card), at least one disk storage device, flash memory device, or other volatile solid-state storage device.

A fourth aspect of the embodiments of the present invention provides a computer readable storage medium, where the computer readable storage medium includes a stored computer program, where the computer program when executed controls a device in which the computer readable storage medium is located to execute the remote sensing image geometry and radiation integrated correction method according to any one of the embodiments of the first aspect.

From the above description of the embodiments, it will be clear to those skilled in the art that the present invention may be implemented by means of software plus necessary hardware platforms, but may of course also be implemented entirely in hardware. With such understanding, all or part of the technical solution of the present invention contributing to the background art may be embodied in the form of a software product, which may be stored in a storage medium, such as ROM/RAM, a magnetic disk, an optical disk, etc., including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to perform the method described in the embodiments or some parts of the embodiments of the present invention.

While the foregoing is directed to the preferred embodiments of the present invention, it will be appreciated by those skilled in the art that changes and modifications may be made without departing from the principles of the invention, such changes and modifications are also intended to be within the scope of the invention.

Claims

1. The geometric and radiation integrated correction method for the remote sensing image is characterized by comprising the following steps of:

2. The method for integrated geometric and radiation correction of a remote sensing image according to claim 1, wherein the step of obtaining the index image of the corresponding point of each target image control point in the remote sensing image to be corrected comprises the following steps:

3. The method for geometric and radiation integrated correction of a remote sensing image according to claim 2, further comprising the steps of, after obtaining the index image of the corresponding point position of each candidate image control point in the remote sensing image to be corrected:

4. The method for geometric and radiation integrated correction of remote sensing images according to claim 2, wherein the matching of the point index image corresponding to each candidate image control point with each corresponding point index image by SIFT feature matching method, to obtain an image control point homonymous image point pair set and a non-image control point homonymous image point pair set, comprises the following steps:

5. The method for geometric and radiation integrated correction of remote sensing images according to claim 4, wherein the matching strategy based on geographic coordinate constraint re-matches the point index image corresponding to each candidate image control point with the corresponding point index image to obtain a plurality of correct matching point pairs between the point index image corresponding to each candidate image control point and the corresponding point index image, specifically comprising the following steps:

6. The method for integrated geometric and radiation correction of a remote sensing image according to claim 4, further comprising the steps of, after obtaining the image coordinates of the same-name image point of each candidate image control point in the remote sensing image to be corrected:

7. The method for geometric and radiation integrated correction of a remote sensing image according to claim 1, wherein the geometric correction and radiation correction are sequentially performed on the remote sensing image to be corrected by using a preset geometric correction model and a radiation correction model according to the geometric correction parameter and the radiation correction parameter, respectively, so as to obtain a target corrected remote sensing image, and the method specifically comprises the following steps:

8. A remote sensing image geometry and radiation integrated correction device, comprising:

9. A terminal device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, the processor implementing the remote sensing image geometry and radiation integration correction method according to any one of claims 1 to 7 when the computer program is executed.

10. A computer readable storage medium, characterized in that the computer readable storage medium comprises a stored computer program, wherein the computer program when run controls a device in which the computer readable storage medium is located to perform the remote sensing image geometry and radiation integration correction method according to any one of claims 1 to 7.