Non-uniform correction method of push-broom hyperspectral image suitable for water-land junction scene

Abstract

The application discloses a push-broom hyperspectral image non-uniform correction method suitable for a water-land junction scene, and solves the technical problem that, in the prior art process of processing the hyperspectral image under the water-land junction scene, due to unreasonable mean value calculation, stripe residues appear in the water area, the spectrum of the land area is distorted, and false images and boundary distortion are generated at the water-land junction edge, the column mean value vector is smoothed by using a sliding average, and then a correction coefficient is calculated, so that the correction coefficient is not affected by the dramatic change of the scene content, the water body part cannot be forcibly brightened, the land part cannot be forcibly suppressed, and unnatural bright edges or dark edges cannot appear in the water-land transition zone, so that the adaptive non-uniform correction of the water-land junction scene is realized.

Description

Technical Field

[0001] This invention relates to a pushbroom hyperspectral image correction method, specifically a pushbroom hyperspectral image non-uniformity correction method applicable to land-water interface scenarios. Background Technology

[0002] Pushbroom hyperspectral imaging systems use area array detectors to perform pushbroom imaging of ground objects, enabling the simultaneous acquisition of spatial, radiometric, and fine spectral information of targets. They have been widely used in water-land interface scenarios such as coastal zone monitoring, inland river and lake water quality inversion, and wetland ecological surveys.

[0003] Due to differences in detector manufacturing processes, uneven illumination of optical systems, and inconsistent photoelectric responses of individual detector elements, pushbroom hyperspectral images generally exhibit fringe noise along the scanning direction, which directly affects the accuracy of radiometric calibration, the accuracy of ground feature classification, and the reliability of quantitative inversion.

[0004] Chinese invention patent CN119417738A discloses a scene-based pushbroom hyperspectral image adaptive non-uniformity correction method. This method automatically sorts each scan line of a band in the hyperspectral image according to the pixel values ​​of different spatial dimensions from smallest to largest, and adaptively selects at least one dark uniform region and one bright uniform region in the spatial dimension to obtain the correction coefficients for each spatial dimension pixel, thus completing the non-uniformity correction for each spatial dimension. Based on the spectral dimension characteristics, it performs spectral dimension non-uniformity correction based on the spatial dimension non-uniformity correction results. The spatial dimension non-uniformity correction is then repeated based on the spectral dimension non-uniformity correction results, thus completing the hyperspectral image non-uniformity correction. This method is fully automated, overcomes scene limitations, and utilizes joint spatial and spectral information for non-uniformity correction, preserving the spatial structure and spectral information of the hyperspectral image to the greatest extent, thereby improving the quality of the hyperspectral image.

[0005] However, the above-mentioned technical solutions have significant drawbacks in hyperspectral image processing in water-land interface scenarios:

[0006] 1) The radiation value of land is much higher than that of water. When calculating the correction coefficient of each spatial pixel, the mean value of each scan row will be significantly increased by the radiation value of land.

[0007] 2) Due to the low mean value in the water area, insufficient correction occurred, resulting in residual streaks;

[0008] 3) Overcorrection occurred in land areas due to higher mean values, resulting in spectral distortion;

[0009] 4) Artifacts and boundary distortions are easily generated at the boundary between land and water, which cannot meet the requirements of high-precision remote sensing applications. Summary of the Invention

[0010] The purpose of this invention is to solve the technical problems in the process of processing hyperspectral images in water-land boundary scenes, where unreasonable mean calculation leads to stripe residue in water areas, spectral distortion in land areas, artifacts and boundary distortion at the water-land boundary edges. The invention provides a pushbroom hyperspectral image non-uniformity correction method suitable for water-land boundary scenes.

[0011] To achieve the above objectives, the technical solution provided by this invention is as follows:

[0012] A pushbroom-based hyperspectral image non-uniformity correction method suitable for land-water interface scenarios is characterized by the following steps:

[0013] Step 1: Acquire pushbroom hyperspectral image (HSI) of the land-water interface scene, and determine the number of rows H, columns W, and bands C of the pushbroom hyperspectral image (HSI).

[0014] Step 2: Sort the pixel values ​​of each scan column in the first band of the pushbroom hyperspectral image (HSI) from smallest to largest, and then select the centered and continuous pixels. The region where the row is located is a uniform region of interest (ROI) of the pushbroom hyperspectral image HSI;

[0015] Step 3: Calculate the column mean of each scan column in the first band within the ROI range, and use it as the column mean of that scan column to obtain the column mean vector of the first band. ;

[0016] Step 4: Calculate the column mean vector Smoothing is performed using a moving average with a sliding window size of k and a step size of 1. A smoothing reference value is then calculated. The calculation formula is as follows:

[0017]

[0018] Where m is the m-th scan column within the current sliding window range in column W, and w is the w-th scan column. , This is the column mean of the m-th scan column within the current sliding window range in column W of the first band;

[0019] Step 5: Calculate the correction coefficients for each scan column of the first band. The calculation formula is as follows:

[0020] ;

[0021] Step 6: Based on the correction coefficient Calculate the corrected pixel value of each pixel in the H-scan row and W-scan column of the first band. Where I represents the I-th scan line of the first band. ;

[0022] Step 7: Using the methods in steps 2-6, calculate the corrected pixel values ​​of each pixel in the H-scan row and W-scan column of all bands to obtain the corrected pushbroom hyperspectral image HSI, thus completing the non-uniform correction of the pushbroom hyperspectral image suitable for water-land interface scenarios.

[0023] Furthermore, in step 3, the column mean vector The calculation formula is as follows:

[0024] ;

[0025] in, This represents the pixel value of the pixel corresponding to the i-th scan row and w-th scan column within the ROI of the first band. .

[0026] Furthermore, in step 2, the centered and continuous The row range is from row 25%H to row 75%H.

[0027] Furthermore, in step 6, the formula for calculating the corrected pixel value of each pixel in the H-scan row and W-scan column of the first band is as follows:

[0028] ;

[0029] in, These are the original pixel values ​​of each pixel in the w-th scan column of the first band of the pushbroom hyperspectral image (HSI).

[0030] Furthermore, in step 4, And round down to the nearest odd number.

[0031] Compared with the prior art, the present invention has the following beneficial technical effects:

[0032] 1. This invention is applicable to a pushbroom-type hyperspectral image non-uniformity correction method for water-land interface scenarios, by adjusting the column mean vector. Smoothing is performed using a moving average, and then the correction factor is calculated. This ensures that the correction coefficients are not affected by drastic changes in scene content, and will not forcibly brighten the water parts or forcibly suppress the land parts. The transition zone between water and land will not have unnatural bright or dark edges, thus achieving adaptive non-uniform correction for water-land boundary scenes.

[0033] 2. This invention is applicable to pushbroom hyperspectral image non-uniformity correction methods for water-land interface scenarios, selecting centered and continuous [images / images]. The area containing the scan lines from line 25%H to line 75%H is a uniform region of origin (ROI) in the pushbroom hyperspectral image HSI, which can effectively remove extreme values ​​of high brightness on land and extreme values ​​of extreme darkness in water, ensuring regional radiation stability.

[0034] 3. This invention is applicable to pushbroom-type hyperspectral image non-uniformity correction methods for water-land interface scenarios, with the sliding window value range set as follows: And round down to the nearest odd number, with an upper limit. This ensures that the correction coefficients do not become ineffective due to excessive smoothing, preserving the effective difference information of each detector element; lower limit The system ensures that the window has sufficient smoothness to effectively filter out abrupt changes in radiation at the water-land interface; the odd-value window guarantees the unbiasedness and positional alignment of the correction coefficients. The synergistic effect of these three factors enables the present invention to achieve an optimal balance between computational efficiency, correction accuracy, and scene adaptability. In particular, for semi-water and semi-land scenes, it achieves beneficial technical effects such as artifact-free water-land transition zones, adaptive adaptation to different water-land ratios, and high-fidelity radiation information. Attached Figure Description

[0035] Figure 1 This is a column mean vector in an embodiment of a pushbroom-type hyperspectral image non-uniformity correction method applicable to land-water interface scenarios according to the present invention. Smoothing was achieved using a moving average. (Structural diagram with W=10 and k=5 as an example);

[0036] Figure 2 The image is a comparison of the original image and the image after non-uniform correction of a pushbroom hyperspectral image of a land-water interface scene using the non-uniform correction method of the present invention, which is applicable to land-water interface scenes. The image is (a) the original image, (b) the image after correction using the comparative patent technology, and (c) the image after correction using the method of this application. Detailed Implementation

[0037] To make the objectives, advantages, and features of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Those skilled in the art should understand that these embodiments are merely used to explain the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0038] This embodiment provides a pushbroom-based hyperspectral image non-uniformity correction method suitable for land-water interface scenarios, characterized by the following steps:

[0039] Step 1: Acquire pushbroom hyperspectral image (HSI) of the land-water interface scene, and determine the number of rows H, columns W, and bands C of the pushbroom hyperspectral image (HSI).

[0040] Step 2: Sort the pixel values ​​of each scan column in the first band of the pushbroom hyperspectral image (HSI) from smallest to largest, and then select the centered and continuous pixels. The row is located in a uniform region of interest (ROI) of a pushbroom hyperspectral image (HSI); in this embodiment, it is centered and continuous. The row range is from row 25%H to row 75%H.

[0041] Step 3: Calculate the column mean of each scan column in the first band within the ROI range, and use it as the column mean of that scan column to obtain the column mean vector of the first band. The calculation formula is as follows:

[0042] ;

[0043] in, This represents the pixel value of the pixel corresponding to the i-th scan row and w-th scan column within the ROI of the first band. .

[0044] Step 4, as follows Figure 1 As shown, for the column mean vector Smoothing is performed using a moving average with a sliding window size of k. And round down to an odd number, with a sliding step of 1, to obtain the smoothing reference value. The calculation formula is as follows:

[0045]

[0046] Where m is the m-th scan column within the current sliding window range in column W, m is to All values ​​within the range, where w is the w-th scanned column. , This is the column mean of the m-th scan column within the current sliding window range in column W of the first band;

[0047] The specific steps are as follows:

[0048] Step 4.1: Set a sliding window of size k with a step size of 1, and apply the column mean vector... The column mean of the first column is smoothed.

[0049] Step 4.2, when When, the selected window size k is [missing information]. of , , ..., Take the average of the above values ​​as The smoothed value;

[0050] when When, the selected window size k is [missing information]. of , , ..., Take the average of the above values ​​as The smoothed value;

[0051] when When, the selected window size k is [missing information]. of , , ..., Take the average of the above values ​​as The smoothed value.

[0052] For example: W=10, k=5, meaning a total of 10 numbers, and a window size of 5. When w=1 or 2, there is no window on the left, only on the right. In this case, the window is manually set to the average of the numbers in columns 1-5. When w=3, there are 2 numbers available on both the left and right sides, so the average of columns 1-5 can be obtained. When w=4, the average of columns 2-6 can be obtained, and so on. When w=7, the average of columns 5-9 can be obtained. When w=8, the average of columns 6-10 can be obtained. When w=9 and 10, there is no window on the right. In this case, the window is manually set to the last 5 numbers, i.e., the average of columns 6-10 can be obtained.

[0053] Step 5: Calculate the correction coefficients for each scan column of the first band. The calculation formula is as follows:

[0054] ;

[0055] Step 6: Based on the correction coefficient Calculate the corrected pixel value of each pixel in the H-scan row and W-scan column of the first band. The calculation formula is as follows:

[0056] ;

[0057] in, Here, I represents the original pixel values ​​of each pixel in the w-th scan column of the first band of the pushbroom hyperspectral image (HSI), and I represents the i-th scan row of the first band. ;

[0058] For example, when w=1, the correction coefficient of the first scan column is used. Multiply by the original pixel value of each pixel in all rows of the first scan column to obtain the corrected pixel values ​​of all pixels in scan row H of the first scan column. This process continues until all cells in the W scan column have been calculated.

[0059] Step 7: Using the methods in steps 2-6, calculate the corrected pixel values ​​of each pixel in the H-scan row and W-scan column of all bands to obtain the corrected pushbroom hyperspectral image HSI, thus completing the non-uniform correction of the pushbroom hyperspectral image suitable for water-land interface scenarios.

[0060] like Figure 2 As shown, by comparing the original pushbroom hyperspectral image, the image corrected using the patented technology, and the image corrected using the technology of this application, it was found that the stripe noise along the column direction in the original pushbroom hyperspectral image was completely eliminated, the radiation restoration of the water and land areas was accurate, and the boundary between water and land was clear and natural. The correction effect was significantly better than the traditional global averaging method.

[0061] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the present invention.

Claims

1. A pushbroom-type hyperspectral image non-uniformity correction method suitable for land-water interface scenarios, characterized in that, Includes the following steps: Step 1: Acquire pushbroom hyperspectral image (HSI) of the land-water interface scene, and determine the number of rows H, columns W, and bands C of the pushbroom hyperspectral image (HSI). Step 2: Sort the pixel values ​​of each scan column in the first band of the pushbroom hyperspectral image (HSI) from smallest to largest, and then select the centered and continuous pixels. The region where the row is located is a uniform region of interest (ROI) of the pushbroom hyperspectral image HSI; Step 3: Calculate the column mean of each scan column in the first band within the ROI range, and use it as the column mean of that scan column to obtain the column mean vector of the first band. ; Step 4: Calculate the column mean vector Smoothing is performed using a moving average with a sliding window size of k and a step size of 1. A smoothing reference value is then calculated. The calculation formula is as follows: ; Where m is the m-th scan column within the current sliding window range in column W, and w is the w-th scan column. , This is the column mean of the m-th scan column within the current sliding window range in column W of the first band; Step 5: Calculate the correction coefficients for each scan column of the first band. The calculation formula is as follows: ； Step 6: Based on the correction coefficient Calculate the corrected pixel values ​​of each pixel in the H-scan row and W-scan column of the first band. Where I represents the I-th scan line of the first band. ; Step 7: Using the methods in steps 2-6, calculate the corrected pixel values ​​of each pixel in the H-scan row and W-scan column of all bands to obtain the corrected pushbroom hyperspectral image HSI, thus completing the non-uniform correction of the pushbroom hyperspectral image suitable for land-water interface scenarios.

2. The pushbroom hyperspectral image non-uniformity correction method for land-water interface scenarios according to claim 1, characterized in that, In step 3, the column mean vector The calculation formula is as follows: ； in, This represents the pixel value of the pixel corresponding to the i-th scan row and w-th scan column within the ROI of the first band. .

3. The pushbroom hyperspectral image non-uniformity correction method applicable to land-water interface scenarios according to claim 1, characterized in that: In step 2, the centered and continuous The row range is from row 25%H to row 75%H.

4. The pushbroom hyperspectral image non-uniformity correction method for land-water interface scenarios according to claim 1, characterized in that, In step 6, the formula for calculating the corrected pixel value of each pixel in the H-scan row and W-scan column of the first band is as follows: ； in, These are the original pixel values ​​of each pixel in the w-th scan column of the first band of the pushbroom hyperspectral image (HSI).

5. The pushbroom hyperspectral image non-uniformity correction method for land-water interface scenarios according to claim 1, characterized in that: In step 4, And round down to the nearest odd number.

Images