A highly robust demosaicing bilinear interpolation method for polarization hyperspectral imaging

Abstract

A robust demosaic bilinear interpolation method for polarization hyperspectral imaging is presented, addressing issues such as optical system platform vibration, frame loss in acquired images, and spectral aliasing in existing technologies. The method includes acquiring a set of polarization hyperspectral images and performing robust demosaic reconstruction on them. Image I is the acquired image of the Kth spectral band, and mosaic image II is an empty matrix used as the interpolation space for image I. Horizontal values ​​are assigned to mosaic image II, and boundary conditions are applied for filling to complete the horizontal interpolation. The intensity images of the four polarization channels in different directions in mosaic image II are extracted, and vertical interpolation is performed on these polarization channel intensity images. This completes the polarization demosaic reconstruction of the Kth spectral band image. This method is applicable to suppressing image artifacts generated during reconstruction.

Description

Technical Field

[0001] The invention relates to the technical field of polarization spectral imaging, and in particular to a highly robust demosaicing bilinear interpolation method for polarization hyperspectral imaging. Background Art

[0002] In recent years, the target detection scheme combining polarization imaging technology with spectral imaging technology has become one of the important research hotspots. Polarization spectral imaging technology combines the two technologies to obtain the two-dimensional image information of the target object, as well as the spectral information of each object point and the polarization information at each wavelength. The two technologies complement each other and can understand the target more comprehensively, accurately and meticulously. Polarization and spectral integration can improve the ability to identify military targets in complex background environments, achieve compact structure, miniaturization, low cost and modular production, and open up new directions for target detection and identification. Although people can camouflage targets by changing the color of objects, most military targets can be identified if polarization detection and spectral detection are used at the same time.

[0003] Focus plane polarization imaging has high timeliness and can obtain multi-dimensional polarization information at the same time. In addition, its compact structure facilitates the miniaturization of the imaging system. Therefore, the use of focus plane polarization detectors has become the mainstream application solution for polarization hyperspectral imaging systems. Common polarization hyperspectral imaging systems use a combination of push-broom hyperspectral imaging methods and focus plane polarization detectors to simultaneously collect hyperspectral information and polarization information, combining the technical advantages of the two imaging solutions. However, due to the characteristics of its own polarization array structure, the focus plane polarization sensor has an inherent spatial resolution loss problem, which greatly affects the measurement accuracy of polarization information. Therefore, the demosaicing super-resolution reconstruction of the polarization image is an indispensable step in the polarization imaging system built with the focus plane polarization detector. Accurate and effective polarization image super-resolution reconstruction can enhance the detection and recognition capabilities of the target.

[0004] At present, the demosaicing reconstruction method of polarization hyperspectral images follows the interpolation method of classical polarization image demosaicing reconstruction, such as bilinear interpolation, bicubic interpolation, Newton polynomial interpolation, etc. These reconstruction algorithms rely on the pixels of adjacent different polarization channels to carry the corresponding different adjacent spatial information for demosaicing reconstruction. In fact, the information carried in the pixel space of the image collected by the push-broom hyperspectral imaging system often does not strictly follow the spatial information distribution of the observation scene due to problems such as platform vibration of the optical system and frame loss of the collected image. For the polarization hyperspectral imaging system with spectral aliasing, the spatial information in the dispersion direction of each acquisition frame is also unreliable due to the influence of spectral aliasing. The direct use of the classical interpolation reconstruction method will lead to errors in the calculation of spatial information and polarization information of the polarization hyperspectral image, and the occurrence of image artifacts, which greatly affects the detection and recognition ability of the polarization hyperspectral imaging system for the target. Therefore, the existing super-resolution reconstruction scheme of polarization hyperspectral images is difficult to meet the use requirements of practical applications. Summary of the invention

[0005] The present invention aims to solve the problems in the prior art that the optical system is in platform vibration, the collected images have frame loss, the optical system has spectral aliasing, etc.

[0006] To solve the above technical problems, the present invention is achieved through the following technical solutions:

[0007] Solution 1: The present invention proposes a highly robust demosaicing bilinear interpolation method for polarization hyperspectral imaging, the method comprising the following steps:

[0008] Step 1: Collect a set of polarization hyperspectral images and perform high robustness on the collected set of polarization hyperspectral images.

[0009] De-mosaicing and reconstruction;

[0010] The highly robust demosaicing reconstruction of a set of acquired polarization hyperspectral images includes the following steps:

[0011] Step 2: Image I is the image of the Kth spectrum segment obtained, and mosaic image II is an empty matrix used as the interpolation space of image I;

[0012] Step 3, assigning values ​​to the mosaic image II in the horizontal direction;

[0013] Step 4: After assigning values ​​to the i-th column images in the mosaic image II in step 3, the boundary conditions are applied to fill the images and complete the horizontal interpolation;

[0014] Step 5, extract the polarization channel intensity images in four different directions from the mosaic image II that has completed the horizontal interpolation, put them into the tensor img, and perform vertical interpolation on the polarization channel intensity images in four different directions;

[0015] Step 6, complete the polarization demosaic reconstruction of the Kth spectral band image, and determine whether the images of all spectral bands have been reconstructed. If the reconstruction is not completed, set K=K+1 and repeat steps 1 to 5; if the reconstruction is completed, the polarization demosaic super-resolution reconstruction of the polarization hyperspectral image is completed.

[0016] Furthermore, a preferred implementation manner is provided, wherein the set of polarization hyperspectral images collected in step 1 is a set of vertical spatial dispersion polarization hyperspectral images corresponding to different horizontal positions.

[0017] Furthermore, a preferred implementation is provided, in step 2, the mosaic image II is an empty matrix, and the determination method for using it as the interpolation space of the image I is:

[0018] Array A=[1, 2, …, ly], where ly is the pixel width of a spectral segment in the spectral dimension of the image during a single acquisition.

[0019] Furthermore, a preferred embodiment is provided, wherein the horizontal assignment of the mosaic image II in step 3 further includes the step of confirming the relative position offset between the column coordinates of the mosaic image II and the image I.

[0020] Further, a preferred embodiment is provided, wherein the method for confirming the relative position offset between the mosaic image II and the image I column coordinates is as follows:

[0021] Let i = 1, 2, ..., 2 n-ly, q=ly floor(i+ly-1) / (ly 2),

[0022] Among them, i represents the assignment of values ​​to the i-th column image in II, q represents the offset of the relative position between the coordinates of columns II and I in the process of assigning values ​​to II, and n is the image length.

[0023] Furthermore, a preferred implementation is provided, in which the method of applying boundary conditions for filling in step 4 is to use a nearest neighbor interpolation method for filling.

[0024] Furthermore, a preferred embodiment is provided, in which the polarization channel intensity images in four different directions extracted in step 5 are obtained by extracting the polarization images using a split-focus plane detector.

[0025] Furthermore, a preferred embodiment is provided, wherein the size of the mosaic image II is larger than that of the image I.

[0026] Solution 2: A computer device includes a processor and a storage medium, wherein the memory stores a computer program. When the processor runs the computer program stored in the memory, the processor executes any one of the methods described in Solution 1.

[0027] Solution 3: A computer storage medium is used to store a computer program. When the computer program is read by a computer, the computer executes any one of the methods described in Solution 1.

[0028] The present invention is beneficial in that:

[0029] The present invention proposes a highly robust demosaicing bilinear interpolation method for polarization hyperspectral imaging, which uses the spatial information between adjacent frames to perform interpolation calculations to achieve demosaicing reconstruction of polarization hyperspectral images, uniformly improves the spatial resolution of images in each polarization direction between frames, and effectively suppresses image artifacts generated by traditional interpolation methods during the reconstruction process. Compared with the demosaicing reconstruction method of polarization hyperspectral images used in the prior art, the method has higher robustness and better reconstruction results, increases the detection and recognition capabilities of targets and the adaptability of the polarization hyperspectral imaging system to the environment.

[0030] The present invention improves on the bilinear interpolation method to achieve demosaicing and reconstruction of polarized hyperspectral images, with short calculation time and small hardware resource requirements, thus meeting the application scenarios of real-time processing.

[0031] The present invention is also applicable to the field of suppressing image artifacts generated in a reconstruction process. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 It is a schematic diagram of the polarization degree image in the 750nm spectrum band after reconstruction by the Newton polynomial polarization demosaicing method described in the eleventh embodiment.

[0033] Figure 2 It is a schematic diagram of the polarization degree image in the 750nm spectrum band after being reconstructed by the polarization demosaicing method proposed by the present invention as described in the eleventh embodiment.

[0034] Figure 3 It is a schematic diagram of a polarization degree image at a location where frame loss occurs in the 750nm spectrum band after reconstruction using the Newton polynomial polarization demosaicing method described in the eleventh embodiment.

[0035] Figure 4 It is a schematic diagram of a polarization degree image at a location where frame loss occurs in the 750nm spectrum band after reconstruction using the polarization demosaicing method proposed in the present invention as described in the eleventh embodiment.

[0036] Figure 5Schematic diagram of the acquisition principle of polarization hyperspectral images in the highly robust demosaicing bilinear interpolation method for polarization hyperspectral imaging described in Implementation Example 1.

[0037] Figure 6 This is a schematic diagram of the corresponding relationship between each polarization channel in the polarization hyperspectral image described in Implementation Example 1.

[0038] Figure 7 This is a flow chart of the highly robust demosaicing bilinear interpolation method for polarization hyperspectral imaging described in implementation mode 1. DETAILED DESCRIPTION

[0039] In order to make the purpose, technical solutions and advantages of the implementation methods of the present application clearer, the technical solutions in the implementation methods of the present application will be clearly and completely described below in conjunction with the drawings in the implementation methods of the present application. Obviously, the described implementation methods are only part of the implementation methods of the present application, not all of the implementation methods.

[0040] Embodiment 1: This embodiment provides a highly robust demosaicing bilinear interpolation method for polarization hyperspectral imaging, the method comprising the following steps:

[0041] Step 1, collecting a set of polarization hyperspectral images, and performing highly robust demosaicing reconstruction on the collected set of polarization hyperspectral images;

[0042] The highly robust demosaicing reconstruction of a set of acquired polarization hyperspectral images includes the following steps:

[0043] Step 2: Image I is the image of the Kth spectrum segment obtained, and mosaic image II is an empty matrix used as the interpolation space of image I;

[0044] Step 3, assigning values ​​to the mosaic image II in the horizontal direction;

[0045] Step 4: After assigning values ​​to the i-th column images in the mosaic image II in step 3, the boundary conditions are applied to fill the images and complete the horizontal interpolation;

[0046] Step 5, extract the polarization channel intensity images in four different directions from the mosaic image II that has completed the horizontal interpolation, put them into the tensor img, and perform vertical interpolation on the polarization channel intensity images in four different directions;

[0047] Step 6, complete the polarization demosaic reconstruction of the Kth spectral band image, and determine whether the images of all spectral bands have been reconstructed. If the reconstruction is not completed, set K=K+1 and repeat steps 1 to 5; if the reconstruction is completed, the polarization demosaic super-resolution reconstruction of the polarization hyperspectral image is completed.

[0048] Implementation method 2. This implementation method is a further limitation of the highly robust de-mosaicing bilinear interpolation method for polarization hyperspectral imaging described in implementation method 1. The set of polarization hyperspectral images collected in step 1 is a set of vertical spatial dispersion polarization hyperspectral images corresponding to different horizontal positions.

[0049] Implementation method 3: This implementation method further limits the highly robust demosaicing bilinear interpolation method for polarization hyperspectral imaging described in implementation method 1. In step 2, the mosaic image II is an empty matrix, and the determination method used as the interpolation space of the image I is:

[0050] Array A=[1, 2, …, ly], where ly is the pixel width of a spectral segment in the spectral dimension of the image during a single acquisition.

[0051] Implementation method 4. This implementation method is a further limitation of the highly robust de-mosaicing bilinear interpolation method for polarization hyperspectral imaging described in implementation method 1. The horizontal assignment of mosaic image II in step 3 also includes the step of confirming the relative position offset between the column coordinates of mosaic image II and image I.

[0052] Embodiment 5: This embodiment further limits the highly robust demosaicing bilinear interpolation method for polarization hyperspectral imaging described in Embodiment 4. The method for determining the relative position offset between the column coordinates of mosaic image II and image I is as follows:

[0053] Let i = 1, 2, ..., 2 n-ly, q=ly floor(i+ly-1) / (ly 2),

[0054] Among them, i represents the assignment of values ​​to the i-th column image in II, q represents the offset of the relative position between the coordinates of columns II and I in the process of assigning values ​​to II, and n is the image length.

[0055] Implementation method six: This implementation method further limits the highly robust demosaicing bilinear interpolation method for polarization hyperspectral imaging described in implementation method one. The method of applying boundary conditions for filling in step 4 is to use the nearest neighbor interpolation method for filling.

[0056] Implementation method seven: This implementation method further limits the highly robust de-mosaicing bilinear interpolation method for polarization hyperspectral imaging described in implementation method one. The polarization channel intensity images of four different directions extracted in step 5 are obtained by extracting a split-focus plane polarization detector.

[0057] Implementation 8: This implementation further limits the highly robust de-mosaicing bilinear interpolation method for polarization hyperspectral imaging described in Implementation 1, and the size of the mosaic image II is larger than that of image I.

[0058] Embodiment 9. This embodiment proposes a computer device, including a processor and a storage medium, wherein a computer program is stored in the memory. When the processor runs the computer program stored in the memory, the processor executes the method described in any one of embodiments 1 to 8.

[0059] Embodiment 10: This embodiment proposes a computer storage medium for storing a computer program. When the computer program is read by a computer, the computer executes the method described in any one of Embodiments 1 to 8.

[0060] Implementation 11. This implementation includes an example, which is used to explain the above implementation. Figures 1 to 7 To illustrate this implementation, in this embodiment, the spectral range of the polarization hyperspectral imaging system is 400nm-900nm wavelength, the spectral resolution is 2nm, and a total of 251 band images are generated. The imaging principle is as follows Figure 2 As shown, Figure 5 The acquisition principle of polarized hyperspectral images is shown in Figure 2. Figure 6 is the corresponding relationship between each polarization channel in the polarization hyperspectral image. The image collected at one instant is the image collected through a vertical space slit after dispersion by the spectroscopic system. The horizontal direction of the image is the spectral dimension, and the vertical direction is the spatial dimension.

[0061] In this embodiment, a highly robust demosaicing bilinear interpolation method for polarization hyperspectral imaging is proposed, and the method includes the following steps:

[0062] 1. Obtain a set of polarization hyperspectral images. While the polarization hyperspectral imaging system performs a uniform push-sweep motion in the horizontal direction, it uniformly collects a set of vertical spatial dispersion images corresponding to different horizontal positions at 1639 different moments to complete data acquisition. A single-wavelength spectral image is formed by taking a certain width of the part carrying the spectral information from this set of dispersion images and splicing it in the horizontal direction. The optical system detector adopts a focal plane polarization detector, in which each four pixels form a group, and are sensitive to polarization vectors in four different directions. In this embodiment, the vertical length of the dispersion image collected at a certain instant is 9834 pixels, which contains the vertical spatial information of the instant, and the 6-pixel width along the horizontal direction is taken as the spectral information of a specific wavelength in the vertical space. After 1639 acquisitions and completion of horizontal spatial domain splicing, the size of the single-wavelength spatial domain image is 2048×9834.

[0063] A highly robust demosaic reconstruction is performed on the set of polarization hyperspectral images obtained above. The specific steps are as follows:

[0064] 2. Let image I be the image of the kth spectrum segment, where k = 1, 2, …, 251, indicating that this polarization demosaicing reconstruction will be performed on the full-frame mosaic polarization image of the kth spectrum segment. Let image II be an empty matrix, which serves as the interpolation space of image I.

[0065] Let array A = [1, 2, ..., ly], as the basis for determining whether to assign pixel values ​​in image I to interpolation space II in the subsequent algorithm. Horizontal interpolation will be performed below. In this embodiment, the pixel width ly = 6 occupied by a certain spectrum segment in the image spectral dimension during a single acquisition, the spatial domain image width of a single wavelength m = 2048, and the image length n = 9834.

[0066] 3. Because the size of II is larger than I, in the process of assigning values ​​to II in the horizontal direction, the relative position offset between the column coordinates of II and I will accumulate and increase with the assignment process.

[0067] Let i = 1, 2, ..., 2 n-ly, q=ly floor(i+ly-1) / (ly 2), where i represents the value assigned to the i-th column image in II, and q represents the relative position offset between the coordinates of columns II and I in the process of assigning values ​​to II.

[0068] If, i / (ly 2) belongs to the element of array A, then II is assigned the original pixel value in I:

[0069] II( : , i)=I( : , iq);

[0070] Otherwise, II is interpolated:

[0071] II(:, i)=0.5 (I( : , iq)+I( : , i-q+6));

[0072] 4. Complete the 1st, 2nd, …, 2nd in II After assigning values ​​to the n-ly columns, n-(ly-1), 2 n-(ly-1)+1, …, 2 The n columns are filled using boundary conditions. In this embodiment, the nearest neighbor interpolation method is used to fill and complete the horizontal interpolation:

[0073] II( : , 2 n-(ly-1) : 2 n)=I( : , n-(ly-1) : n) ;

[0074] 5. Extract the intensity information of the four polarization channels in the mosaic image II that has completed the horizontal interpolation, and put it into the tensor img. Let the polarization vector intensity images in four different directions be img(:,:,1), img(:,:,2), img(:,:,3), and img(:,:,4), respectively. Among them, img(:,:,1), img(:,:,2), img(:,:,3), and img(:,:,4) are extracted from the upper left pixel, upper right pixel, lower right pixel, and lower left pixel of each superpixel, respectively. In this embodiment, the upper left pixel, upper right pixel, lower right pixel, and lower left pixel are respectively sensitive as follows: Figure 6 Linear polarized light with polarization angles of 0°, 135°, 90°, and 45° as shown:

[0075] img(1:2:m, 1:n, 1)=II(1:2:m, 1:2:2 n);

[0076] img(1:2:m, 1:n, 2)=II(1:2:m, 2:2:2 n);

[0077] img(2:2:m, 1:n, 3)=II(2:2:m, 2:2:2 n);

[0078] img(2:2:m, 1:n, 4)=II(2:2:m, 1:2:2 n);

[0079] 6. The following will perform vertical interpolation:

[0080] Let j = 2, 4, 6, …, m-1, and assign the following value to each j-th row of img( : , : , 1) and img( : , : , 2):

[0081] img(j, : , 1)=0.5*(img(j-1, : , 1)+img(j+1, : , 1));

[0082] img(j, : , 2)=0.5*(img(j-1, : , 2)+img(j+1, : , 2));

[0083] 7. After completing the assignment of the 2nd, 4th, 6th, …, m-1th rows in img( : , : , 1) and img( : , : , 2), apply the boundary condition to fill the mth row. In this embodiment, the nearest neighbor interpolation method is used for filling:

[0084] img(m, : , 1)=img(m-1, : , 1);

[0085] img(m, : , 2)=img(m-1, : , 2);

[0086] 8. Let v = 3, 5, 7, …, m, and assign the following values ​​to each v-th row of img( : , : , 3) ​​and img( : , : , 4):

[0087] img(v, : , 3)=0.5 (img(v-1, : , 3)+img(v+1, : , 3));

[0088] img(v, : , 4)=0.5 (img(v-1, : , 4)+img(v+1, : , 4));

[0089] 9. After completing the assignment of the 3rd, 5th, 7th, …, mth rows in img( : , : , 3) ​​and img( : , : , 4), apply the boundary condition to fill the first row. In this embodiment, the nearest neighbor interpolation method is used for filling to complete the highly robust demosaiced bilinear interpolation super-resolution reconstruction of the full-frame mosaic polarization image of the kth spectral segment:

[0090] img(1, : , 3)=img(2, : , 3);

[0091] img(1, : , 4)=img(2, : , 4);

[0092] 10. Let k = k + 1, and repeat steps 2 to 9. Apply this super-resolution method to the full-frame mosaic polarization images of all s spectral bands to complete the super-resolution reconstruction of the set of polarization hyperspectral images.

[0093] Figure 1This is a comparison diagram of the effects of the highly robust demosaicing bilinear interpolation method and the Newton polynomial polarization demosaicing reconstruction method for polarization hyperspectral imaging shown in this embodiment. Figure 1 is the polarization image at 750nm after reconstruction by Newton polynomial polarization demosaicing method, Figure 2 is a polarization degree image at the 750nm spectrum band after reconstruction by the polarization demosaicing method proposed in this embodiment, Figure 3 This is the polarization image at the point where the frame loss occurs in the 750nm spectrum after reconstruction using the Newton polynomial polarization demosaicing method. Figure 4 This is a polarization degree image at the point where frame loss occurs in the 750nm spectrum after reconstruction by the polarization demosaicing method proposed in this embodiment. It can be seen that Figure 5 There are obvious vertical white stripe artifacts, and Figure 2 The resulting polarization image eliminates Figure 1 For polarization hyperspectral images with frame loss, Figure 3 The reconstruction shown shows more severe vertical white streak artifacts, while Figure 4 It effectively inhibits Figure 3 It shows higher robustness to image artifacts generated in the image.

[0094] Figure 5 Schematic diagram of the imaging principle of the polarization hyperspectral imaging system. Figure 5 The acquisition principle of polarized hyperspectral images is shown in Figure 2. Figure 6 It is the corresponding relationship between each polarization channel in the polarization hyperspectral image. , , …, Corresponding to the image information of s different spectral segments, Y1, Y2, …, Y t Corresponding to the image information collected at different horizontal spaces for t times, Y( ),Y( ), …, Y( ) corresponds to the full-space image information of s different spectral bands after data conversion, and ly is the pixel width occupied by a spectral band in the spectral dimension of the image during a single acquisition.

[0095] Those skilled in the art will appreciate that the above are only preferred embodiments of the present invention, and the various embodiments of the present disclosure and / or the features described in the claims may be combined or combined in various ways, even if such combinations or combinations are not explicitly described in the present disclosure. It is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art may still modify the technical solutions described in the aforementioned embodiments, or perform equivalent substitutions on some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention shall be included in the protection scope of the present invention.

[0096] Although preferred embodiments of the present invention have been described, additional changes and modifications may be made to these embodiments by those skilled in the art once the basic inventive concepts are known. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention. Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A highly robust demosaicing bilinear interpolation method for polarization hyperspectral imaging, characterized in that: The method comprises the following steps: A set of polarization hyperspectral images are collected, and highly robust demosaicing and reconstruction is performed on the collected set of polarization hyperspectral images; Step 1: Image I is the image of the Kth spectrum segment obtained, and mosaic image II is an empty matrix used as the interpolation space of image I; Step 2, assigning values ​​to the mosaic image II in the horizontal direction; Step 3: After assigning values ​​to the i-th column image I in the mosaic image II in step 2, the boundary conditions are applied to fill the image and complete the horizontal interpolation; Step 4, extract the polarization channel intensity images in four different directions from the mosaic image II that has completed the horizontal interpolation, put them into the tensor img, and perform vertical interpolation on the polarization channel intensity images in four different directions; Step 5, complete the polarization demosaic reconstruction of the Kth spectral band image, and determine whether the images of all spectral bands have been reconstructed. If the reconstruction has not been completed, set K=K+1 and repeat steps 1 to 4; if the reconstruction has been completed, the polarization demosaic super-resolution reconstruction of the polarization hyperspectral image is completed; The step 2 of assigning values ​​to the mosaic image II in the horizontal direction also includes the step of confirming the relative position offset between the mosaic image II and the column coordinates of the image I; The method for confirming the relative position offset between the column coordinates of mosaic image II and image I is: Let \(i = 1, 2,\cdots, 2\) n - ly, \(q = ly\) \(\lfloor\frac{i + ly - 1}{ly}\) 2), Among them, i represents the assignment of values ​​to the i-th column image in II, q represents the relative position offset between the coordinates of columns II and I in the process of assigning values ​​to II, n is the image length, and ly is the pixel width occupied by a spectral segment in the spectral dimension of the image during a single acquisition.

2. The highly robust demosaicing bilinear interpolation method for polarization hyperspectral imaging according to claim 1, characterized in that: The set of polarization hyperspectral images collected in step 1 is a set of vertical spatial dispersion polarization hyperspectral images corresponding to different horizontal positions.

3. The highly robust demosaicing bilinear interpolation method for polarization hyperspectral imaging according to claim 1, characterized in that: In step 2, the mosaic image II is an empty matrix, and the determination method for using it as the interpolation space of image I is: Array A=[1, 2, …, ly], where ly is the pixel width of a spectral segment in the spectral dimension of the image during a single acquisition.

4. The highly robust demosaicing bilinear interpolation method for polarization hyperspectral imaging according to claim 1, characterized in that: The method for filling by applying boundary conditions in step 3 is to use the nearest neighbor interpolation method for filling.

5. The highly robust demosaicing bilinear interpolation method for polarization hyperspectral imaging according to claim 1, characterized in that: The intensity images of the polarization channels in four different directions extracted in step 4 are obtained by extracting the polarization detector in the focal plane.

6. The highly robust demosaicing bilinear interpolation method for polarization hyperspectral imaging according to claim 1, characterized in that: The mosaic image II is larger than image I in size.

7. A computer device comprising a memory and a processor, characterized in that A computer program is stored in the memory. When the processor runs the computer program stored in the memory, the processor executes the method according to any one of claims 1 to 6.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images