An Automatic Registration Method for Multispectral Images Based on the Probe Element Pointing Angle Model

Abstract

A multispectral image automatic registration method based on a detector pointing angle model relates to the field of remote sensing image processing technology and solves the problem of low local area registration accuracy of existing multispectral images. The method includes: splitting the multispectral data of the remote sensing image to be processed into images of several different bands, obtaining control point pairs of each image to be registered and a reference image, determining a deviation vector and a registration method based on the control point pairs, fitting the control point pairs using a least squares fitting method based on the deviation vector, performing distortion correction on the camera payload parameters, solving the detector pointing angle model of the payload detector of the spectral segment to be registered, and establishing a strict imaging geometry model of the spectral segment. The present invention has good universality, improves the actual image production efficiency, and has a good effect on improving the inter-spectral registration accuracy of single-view multispectral 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 multispectral image automatic registration method based on a detector pointing angle model. Background Art

[0002] For multispectral images from satellite remote sensing, band registration is a key technology in remote sensing image processing. The accuracy of band registration directly affects the accuracy of subsequent remote sensing image fusion, mosaic content recognition, and other processing algorithms. Currently, for optical remote sensing satellites, the optical camera on board is geometrically calibrated before launch. However, environmental factors such as external vibrations during launch and stress release during on-orbit operation can cause changes in the camera's geometric parameters, resulting in errors in the same-name light rays along and perpendicular to the track for push-scan images of different bands. Traditional band registration methods often use affine transformation models to perform overall image translation, rotation, scaling, and other changes. In fact, most multispectral images have poor registration accuracy in local areas. Starting from registration based on the entire image, there will still be problems with errors in matching local areas. Therefore, there is a need to provide a registration method that improves the registration accuracy of local areas of multispectral images. Summary of the Invention

[0003] In order to solve the above problems, the present invention provides a multispectral image automatic registration method based on a detector pointing angle model.

[0004] The technical solutions adopted by the present invention to solve the technical problems are as follows:

[0005] A multispectral image automatic registration method based on a detector pointing angle model includes the following steps:

[0006] Step 1: Split the multispectral remote sensing image data to be processed into several images of different bands, and select the image of one band as the reference image, and the images of the remaining bands as the images to be registered;

[0007] Step 2: Use the Surf algorithm to extract feature points of the reference image and the image to be registered;

[0008] Step 3: Match each feature point of the image to be registered with the feature point of the reference image to form a control point pair between each image to be registered and the reference image. The control point pair includes an image control point on the image to be registered and an image control point on the reference image corresponding to the image control point.

[0009] Step 4: Divide the reference image and the image to be registered into multiple regions, and calculate the regional deviation vector of each region of the image to be registered relative to the corresponding region in the reference image;

[0010] Step 5: Determine whether the image to be registered requires global registration, local registration, or no registration based on all regional deviation vectors of the image to be registered; if the image to be registered requires registration, proceed to step 6;

[0011] Step 6: Decompose the deviation vector of each image to be registered into the component in the along-track direction and the component in the vertical-track direction, establish the relative relationship between the detector element of the spectral segment of the image to be registered and the detector element of the spectral segment of the reference image, use the least squares fitting to obtain the theoretical deviation corresponding to the detector element position of the spectral segment of the image to be registered, remove the image control points that exceed the tolerance and re-perform the least squares fitting, iterate until the residuals of all image control points of the image to be registered do not exceed the tolerance, and obtain the vertical-track deviation curve f of the image to be registered x (x i2 ) and the deviation curve f of the image to be registered along the track y (x i2 );

[0012] Step 7: According to f x (x i2 ) and f y (x i2 ) Correct the distortion of the camera payload parameters and correct the actual imaging position of each detector in the camera coordinate system of the image spectrum to be registered;

[0013] Step 8: Using the original camera parameters and the corrected actual imaging position, the detector pointing angle model of the detector loaded in the image spectrum segment to be registered is solved, and a strict imaging geometric model of the spectrum segment is established to regenerate the multispectral data of the remote sensing image multispectral data to be processed described in step 1.

[0014] The beneficial effects of the present invention are:

[0015] The present invention proposes an automatic multispectral image registration method based on a detector pointing angle model. The method relies on the relative relationship between spectral segments of the multispectral image. A single image does not need to consider the influence of external orientation elements separately. At the same time, it is not constrained by the acquisition of ground control points and only depends on the multispectral image to be registered itself. The method has good universality in practical applications. Least squares fitting and limit error conditions are used to conditionally constrain, automatically screen, and adapt the algorithm for the deviation vector. The automated process improves the actual production efficiency of the image. When using the detector pointing angle model for inter-spectral registration, the relative distortion difference can be corrected. The present invention has a good effect of improving the inter-spectral registration accuracy for single-scene multispectral remote sensing images. A strict imaging geometry model is used to resample the original scene to avoid the information loss that may be caused by resampling the corrected remote sensing image. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1This is a flow chart of a multispectral image automatic registration method based on a detector pointing angle model of the present invention;

[0017] Figure 2 Schematic diagram of image grid division according to the present invention. DETAILED DESCRIPTION

[0018] In order to more clearly understand the above-mentioned objects, features and advantages of the present invention, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0019] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the scope of protection of the present invention is not limited to the specific embodiments disclosed below.

[0020] A multispectral image automatic registration method based on the detector pointing angle model, such as Figure 1 , including the following steps:

[0021] Step 1: Split the multispectral remote sensing image data to obtain several different band images. For the split images, select one band as the reference image and the remaining bands as the images to be registered. In other words, determine the individual reference bands and the bands to be registered.

[0022] As an example, the Jilin-1 remote sensing satellite is capable of detecting radiation reflected from objects in two or more bands. The multispectral imagery it produces contains grayscale images of each band simultaneously imaged. In actual use, the imagery may suffer from low registration accuracy in localized areas. The multispectral imagery of the Jilin-1 GF03D satellite L1-class remote sensing image to be processed is split into four segments based on the visible and near-infrared spectra, namely, Band 1, Band 2, Band 3, and Band 4. Band 1 is selected as the reference band image, and registration correction is performed on Band 2, Band 3, and Band 4 images.

[0023] Step 2: For the reference image selected in step 1 and the image to be registered, use the Surf algorithm to automatically extract their feature points.

[0024] The selected band 1 image, band 2 image, band 3 image, and band 4 image are degraded and stretched, and their SURF feature points are automatically extracted.

[0025] Step 3: Match each feature point of the image to be registered with the feature point of the reference image to form a control point pair of each registered image and the reference image;

[0026] Match the feature points of each image to be registered with the corresponding feature points of the reference image, constrain the matching results, and eliminate out-of-limit matching points (eliminate point pairs that exceed the preset threshold of the constraint) to form control point pairs (also called image control point pairs) of each image to be registered and the reference image.

[0027] Using a brute force matching algorithm, the feature points of the Band 2 image are matched with those of the Band 1 image to obtain matching results. Other information, such as the distance of the feature vectors, is obtained. Using the distance ratio and the fundamental matrix, qualified image control point pairs are selected from the matching results to serve as the image control point pairs between Band 1 and Band 2. The same operation is performed on Band 3 and Band 4 images to obtain image control point pairs between Band 1 and Band 3, and between Band 1 and Band 4.

[0028] Step 4: Divide the reference image and the image to be registered into multiple regions, and calculate the regional deviation vector of each region of the image to be registered relative to the corresponding region in the reference image.

[0029] The reference image and the image to be registered are divided into N*N regions, where N is a positive integer. In this embodiment, N is the number of linear array sensors used in multispectral imaging. For each image to be registered, the regional deviation vectors of the control point pairs in each region of the image to be registered and the corresponding region in the reference image are calculated. Specifically, the deviation vectors of each control point pair are obtained using the coordinate differences between the image to be registered and the reference image. The regional deviation vectors of each region of each image to be registered are statistically analyzed to obtain the regional deviation vector of each region.

[0030] Taking the deviation vector between Band 1 and Band 2 as an example, since the payload used to capture the multispectral data of the remote sensing image has 3 CCDs in each band, the Band 1 image and the Band 2 image are divided into 9 regions in the same way, that is, the image control point pairs are divided into 9 grid regions of 3x3. The row range of each region satisfies the number of rows of the entire band image, and the column range is determined by the CCD width and the overlap area between CCDs. The region division is as follows: Figure 2 As shown, it is divided into regions ① to ⑨. The two image control points of the control point pair are the same-name image points. One image control point is the image control point on the image to be registered, and the other image control point is the image control point on the reference image. The deviation vector of the following same-name image points is the deviation vector of the image control point on the image to be registered relative to the image control point on the reference image. Suppose the coordinates of the n same-name image points in region ① on the reference image (band 1 image) are x i1 ,y i1 i=1,2,3,...,n, i represents the i-th image control point, n represents the number of image control point pairs in region ①, and the coordinate on the image to be registered (band 2 image) is x i2 ,y i2i=1,2,3,...,n, then the deviation vector of each image point with the same name is represented by x in the image coordinate system. i1 ,y i1 Point to x i2 ,y i2 Calculate the average deviation vector of the image control points in area ① on the band 2 image relative to the image control points in area ① on the band 1 image. The direction of the average deviation vector is the mean of the deviation vector directions of the same-name image points and . The magnitude of the deviation vector in area ① is

[0031]

[0032] That is, the average deviation vector of all image control points in the Band 2 image area is also the mean of the deviation vectors of all image control points with the same name. The average deviation vector of all image control points with the same name in the area is the area deviation vector.

[0033] Based on the Band 1 image, perform the same operation on the Band 3 image and the Band 4 image.

[0034] Step 5: Based on the regional deviation vectors of all regions of the image to be registered (comparing the regional deviation vectors of all regions of the image to be registered with the first preset threshold), determine whether the image to be registered requires global registration, requires local registration, or does not require registration; if the image to be registered requires global registration or requires local registration, proceed to step 6;

[0035] Specifically: first determine whether the image to be registered needs to be registered based on the regional deviation vectors of all regions of the image to be registered; if registration is required, then obtain the consistency of the deviation of the image to be registered based on the consistency of the regional deviation vectors between the regions of the image to be registered; if there is consistency, global registration is required, if not, local registration is required.

[0036] Based on all regional deviation vectors of the image to be registered, the consistency of the deviation of the image to be registered is obtained. As an embodiment, the size of the regional deviation vector and the direction of the regional deviation vector between each region are compared to judge the consistency of each deviation vector. If the deviation of each grid region is less than the threshold value 1, then the scene does not need to be corrected for the inter-spectral registration accuracy. If the deviation between each grid region is greater than or equal to 1 and less than the threshold value 2, then if there is consistency between each grid region, it proves that the image requires global registration; if the deviation between each grid region is not consistent, then the image requires local registration.

[0037] If the deviation vector magnitudes for each region in Band 2 exceed the tolerance, interspectral registration accuracy correction is required. If the deviation vector direction between each grid region in Band 2 exceeds a standard threshold (e.g., 1°) and the difference between the maximum and minimum vector lengths exceeds a length difference threshold (e.g., 1 pixel), the deviation vectors between each grid region in Band 2 are considered inconsistent, and the image requires a local registration model. If they are consistent, exceed the standard threshold, and / or exceed the length difference threshold, the image requires a global registration model.

[0038] Step 6: Decompose the deviation vector of each image to be registered into components in the satellite along-track direction and the satellite vertical track direction, establish the relative relationship between the spectral detector of the image to be registered and the spectral detector of the reference image, use least squares fitting to obtain the theoretical deviation corresponding to the detector position of the spectral detector of the image to be registered, eliminate the image control points that exceed the tolerance limit and re-perform the least squares fitting, and iterate until the residuals of all image control points of the image to be registered do not exceed the tolerance limit.

[0039] The deviation vector of the image to be registered is the deviation vector of all image control points in the area to be registered.

[0040] If global registration is required, the above fitting of the deviation curve of the deviation vector of the image to be registered in the vertical track direction is selected using third-order fitting, and linear fitting is selected in the along-track direction; if local registration is required, the above fitting of the vector deviation in the vertical track direction is selected using sixth-order fitting, and the method with the smaller final average residual between linear fitting and third-order fitting is selected in the along-track direction; if there are fewer image control points, the vector deviation in the vertical track direction is selected using third-order fitting, and linear fitting is selected in the along-track direction.

[0041] Step 6 includes:

[0042] Step 6-1: Subtract the coordinates of the image to be registered (e.g., band 2 image) and the reference image (band 1 image) to obtain the deviation (Δx i , Δy i )(i=1, 2, 3, ..., m), where m is the number of image control point pairs required for registration between the image to be registered and the reference image.

[0043] Step 6-2, Deviation (Δx i , Δy i ) Deviation component Δx in the vertical track direction i The coordinates of the image control point x in the band 2 image i2 (i=1, 2, 3, ..., k) performs least square fitting to obtain the vertical track direction fitting curve f x (x i2 ), deviation (Δx i , Δy i ) Deviation component Δy along the track iand the image coordinate x to be registered i2 (i=1, 2, 3, ..., k) performs least square fitting to obtain the along-track fitting curve f y (x i2 ), find the minimum sum of squares of deviations between the fitting curve and each deviation component, that is,

[0044]

[0045]

[0046] Where, f x (x i2 ) is Δx i The corresponding polynomial fitting curve, f y (x i2 ) is Δy i The corresponding polynomial fitting curve, δ ix Represents the fitting curve f x (x i2 ) and Δx i The residual, δ iy Represents δ i The component fitting curve f along the track direction y (x i2 ) and Δy i k is the number of image control points used to fit the curve. When step 6-2 is executed for the first time, k = m.

[0047] Step 6-3: Based on a second preset threshold (ie, a limit error), the image control point data whose residual exceeds the preset threshold is eliminated, and a point set containing k′ image control points that meets the following conditions is retained.

[0048]

[0049] Among them, δ ix Represents δ i The component in the vertical direction, δ iy Represents δ i Component along the track, TH x Indicates the second preset threshold along the track direction, TH y The second preset threshold representing the vertical track direction, k is an integer.

[0050] Step 6-4: Let k = k′, and return to step 6-2 until the residuals of all the image control points involved in the fitting do not exceed the second preset threshold. At this time, f x (x i2 ) as the final vertical track deviation curve f x (x i2 ), at this time f y (x i2) as the along-track deviation curve f y (x i2 ).

[0051] The same operation is performed on the image control point pairs of each image to be registered and the reference image to obtain the vertical track deviation curve and the along track deviation curve of each image to be registered.

[0052] Step 7: Correct the camera payload parameters for distortion according to the vertical track deviation curve and the along-track deviation curve of the image to be registered, and correct the actual imaging position of each detector of the camera payload in the camera coordinate system of the spectral band of the image to be registered.

[0053] The camera payload parameters are corrected for distortion using the deviation between the reference image and the image to be registered. The actual imaging position of each detector in the camera payload of the spectral band to be registered in the camera coordinate system is back-calculated to obtain the corrected actual imaging position.

[0054] For bands 2 to 4 of the multispectral remote sensing image to be registered, for each pixel position of the sensor, the two direction fitting curves f obtained in step 6 are used. x (x i2 ) and f y (x i2 ), use the inter-spectral (between the reference image and the image to be registered) deviation to correct the distortion of the original theoretical pointing in the camera payload parameters, and the vertical direction of the actual imaging position of each detector in the camera payload of the registered spectrum in the camera coordinate system is the detector pointing angle φ x and the detector pointing angle φ along the track v The calculation is performed as follows:

[0055]

[0056] Where S x0 represents the vertical track coordinate, S y0 represents the along-track coordinate, and F represents the camera principal distance.

[0057] Step 8: Using the original camera parameters and the corrected actual imaging position, the detector pointing angle model of the payload detector of the image spectrum segment to be registered is solved, and a strict imaging geometry model of the image spectrum segment to be registered is established to regenerate the multispectral image of the scene (the multispectral remote sensing image data to be processed in step 1).

[0058] Reconstructing a rigorous imaging geometry model

[0059]

[0060] Where (X, Y, Z) is the coordinate of the ground point to be found in the object space coordinate system, is the position vector of the GPS phase center at time t in the WGS84 coordinate system, is the rotation matrix from the J2000 coordinate system to the WGS84 coordinate system at time t (calculated from the nutation, polar motion and precession matrices at time t), is the rotation matrix from the satellite body coordinate system to the J2000 coordinate system at time t, is the rotation matrix from the camera coordinate system to the satellite body coordinate system (given by ground calibration before the satellite is launched), [tanψ x tanψ y 1] T is the current image point coordinate (xy) T The value in the detector pointing angle model is used to regenerate the original multispectral image according to the above known information for the scene image to obtain the final product.

[0061] For three multispectral images (test image 1, test image 2, and test image 3), the multispectral image automatic registration method based on the detector pointing angle model of the present invention was used to compare the data before and after correction. The results are shown in Table 1.

[0062] Table 1

[0063]

[0064]

[0065] The present invention proposes an automatic multispectral image registration method based on a detector pointing angle model. The method relies on the relative relationship between spectral segments of the multispectral image. A single image does not need to consider the influence of external orientation elements separately. At the same time, it is not constrained by the acquisition of ground control points and only depends on the multispectral image to be registered itself. It has good universality in practical applications. Least squares fitting and limit conditions are used to conditionally constrain, automatically screen, and adapt the algorithm for the deviation vector. The automated process improves the actual production efficiency of the image. When using the detector pointing angle model for inter-spectral registration, the relative distortion difference can be corrected. After testing, it has a good effect on improving the inter-spectral registration accuracy for single-scene multispectral remote sensing images. A strict imaging geometry model is used to perform geometric correction on the original image to avoid the information loss that may be caused by resampling the corrected remote sensing image.

Claims

1. A multispectral image automatic registration method based on a detector pointing angle model, characterized in that: The steps include: Step 1: Split the multispectral remote sensing image data to be processed into several images of different bands, and select the image of one band as the reference image, and the images of the remaining bands as the images to be registered; Step 2: Use the Surf algorithm to extract feature points of the reference image and the image to be registered; Step 3: Match each feature point of the image to be registered with the feature point of the reference image to form a control point pair between each image to be registered and the reference image. The control point pair includes an image control point on the image to be registered and an image control point on the reference image corresponding to the image control point. Step 4: Divide the reference image and the image to be registered into multiple regions, and calculate the regional deviation vector of each region of the image to be registered relative to the corresponding region in the reference image; Step 5: Determine whether the image to be registered requires global registration, local registration, or no registration based on all regional deviation vectors of the image to be registered; if the image to be registered requires registration, proceed to step 6; Step 6: Decompose the deviation vector of each image to be registered into the component in the along-track direction and the component in the vertical-track direction, establish the relative relationship between the detector element of the spectral segment of the image to be registered and the detector element of the spectral segment of the reference image, use the least squares fitting to obtain the theoretical deviation corresponding to the detector element position of the spectral segment of the image to be registered, remove the image control points that exceed the tolerance and re-perform the least squares fitting, iterate until the residuals of all image control points of the image to be registered do not exceed the tolerance, and obtain the vertical-track deviation curve f of the image to be registered x (x i2 ) and the deviation curve f of the image to be registered along the track y (x i2 ); Step 7: According to f x (x i2 ) and f y (x i2 ) Correct the distortion of the camera payload parameters and correct the actual imaging position of each detector in the camera coordinate system of the image spectrum to be registered; Step 8: Using the original camera parameters and the corrected actual imaging position, the detector pointing angle model of the payload detector of the image spectrum segment to be registered is solved, and a strict imaging geometric model of the spectrum segment is established to regenerate the remote sensing image multispectral data to be processed as described in step 1.

2. The multispectral image automatic registration method based on the detector pointing angle model according to claim 1, characterized in that: The step 3 specifically includes: using a brute force matching algorithm to match the feature points of the image to be registered with the feature points of the reference image to obtain a matching result, and screening out qualified point pairs from the matching result as control point pairs.

3. The multispectral image automatic registration method based on the detector pointing angle model according to claim 1, characterized in that: The step 4 is specifically as follows: dividing the reference image and the image to be registered into N*N regions, where N is a positive integer, and calculating the average deviation vector of all control point pairs in each region of the image to be registered, and the average deviation vector is used as the regional deviation vector of the image to be registered.

4. The multispectral image automatic registration method based on the detector pointing angle model according to claim 3, characterized in that: The deviation vector of each control point pair is obtained by using the coordinate difference between each control point pair in the image to be registered and the reference image, and the average deviation vector is calculated based on the deviation vectors of all control point pairs in the region.

5. The multispectral image automatic registration method based on the detector pointing angle model according to claim 4, characterized in that: The control point pair includes two image control points with the same name, and the average deviation vector is the average of the deviation vectors of each image control point in the area of the image to be registered and .

6. The multispectral image automatic registration method based on the detector pointing angle model according to claim 1, characterized in that: The step 5 is specifically as follows: first, determine whether the image to be registered needs to be registered based on the regional deviation vectors of all regions of the image to be registered; if registration is required, obtain the consistency of the deviation of the image to be registered based on all regional deviation vectors of the image to be registered; if there is consistency, global registration is required, and if there is no consistency, local registration is required.

7. The multispectral image automatic registration method based on the detector pointing angle model according to claim 1, characterized in that: If the image to be registered requires global registration, the third-order fitting is used when fitting the deviation curve of the deviation vector of the image to be registered in the vertical track direction, and the linear fitting is used in the along-track direction. If the image to be registered requires local registration, the sixth-order fitting is used when fitting the vector deviation in the vertical track direction, and the linear fitting or third-order fitting is selected in the along-track direction.

8. The multispectral image automatic registration method based on the detector pointing angle model according to claim 1, characterized in that: The step 6 is specifically as follows: Step 6-1: Subtract the coordinates of the image to be registered and the reference image to obtain the deviation of each image control point (Δx i ,Δy i )(i=1,2,3,...,k), k is the number of image control point pairs required for registration between the image to be registered and the reference image; Step 6-2, Deviation (Δx i ,Δy i ) Deviation component Δx in the vertical track direction i The coordinates of the image control point x in the band 2 image i2 (i=1,2,3,...,k) perform least square fitting to obtain the vertical track direction fitting curve f x (x i2 ), deviation (Δx i ,Δy i ) Deviation component Δy along the track i and the image coordinate x to be registered i2 (i=1,2,3,...,k) performs least square fitting to obtain the along-track fitting curve f y (x i2 ), find the minimum sum of squares of deviations between the fitting curve and each deviation component, that is, Where, f x (x i2 ) is Δx i The corresponding polynomial fitting curve, f y (x i2 ) is Δy i The corresponding polynomial fitting curve, δ ix Represents the fitting curve f x (x i2 ) and Δx i The residual, δ iy The component fitting curve f represents the component along the track direction of δi y (x i2 ) and Δy i The residual error is: k is the number of image control points used to fit the curve. When step 6-2 is executed for the first time, k = m. Step 6-3: Based on the error limit, remove the image control point data whose residual exceeds the preset threshold, and retain the point set containing k' image control points that meets the following conditions; Among them, δ ix represents the component of δi in the vertical direction, δ iy represents the component of δi along the track direction, TH x Indicates the tolerance along the track direction, TH y Indicates the tolerance in the vertical track direction, k is an integer; Step 6-4: Let k = k', and return to step 6-2 until the residuals of all the image control points involved in the fitting do not exceed the second preset threshold. At this time, f x (x i2 ) as the final vertical track deviation curve f of the image to be registered x (x i2 ), at this time f y (x i2 ) as the deviation curve f along the track direction of the image to be registered y (x i2 ).

9. The multispectral image automatic registration method based on the detector pointing angle model according to claim 1, characterized in that: The calculation of the actual imaging position in step 7 includes: calculating the vertical direction of the detector pointing angle φ of the actual imaging position of each detector of the spectral band camera payload to be registered in the camera coordinate system x and the detector pointing angle φ along the track y Perform calculations, Where S x0 The vertical coordinate of the actual imaging position of the camera payload laboratory, S y0 represents the along-track coordinate of the actual imaging position of the camera payload laboratory, and F represents the camera principal distance.

10. The multispectral image automatic registration method based on the detector pointing angle model according to claim 1, characterized in that: The strict imaging geometry model is: Where (X, Y, Z) is the coordinate of the ground point to be found in the object space coordinate system, where [X s Y s Z s ] t T is the position vector of the GPS phase center in the WGS84 coordinate system at time t, is the rotation matrix from the J2000 coordinate system to the WGS84 coordinate system at time t, is the rotation matrix from the satellite body coordinate system to the J2000 coordinate system at time t, is the rotation matrix from the camera coordinate system to the satellite body coordinate system, [tanψ x tanψ y 1] T is the current image point coordinate (x,y) T The value in the detector pointing angle model.

Images