Image motion detection and correction method for multi-slit hyperspectral image data

Abstract

The invention discloses an image motion detection and correction method for multi-slit hyperspectral image data, and solves the problem that the embodiment of the advantages of a multi-slit hyperspectral imaging system and the high-quality application of data of the multi-slit hyperspectral imaging system are restricted due to the fact that a method for detecting the image motion difference of each pixel among the multi-slit hyperspectral data does not exist in the prior art. According to the method, the image motion amount of each pixel among the multi-slit hyperspectral data is regarded as the image motion amount of each image, continuity exists among the image motion amounts of the pixels, and detection and correction of the image motion amount of each pixel are realized by using different image motion continuity constraint HS dense optical flow methods in the longitudinal and transverse directions. And then, on the basis of distortion image motion deviation of the multi-slit hyperspectral imaging system and imaging image motion deviation caused by platform motion, correcting an optical flow numerical value obtained by calculation through an HS dense optical flow method, removing an abnormal optical flow numerical value, and supplementing the optical flow numerical value at the position of a removed pixel through a linear difference calculation result, so as to obtain a high-resolution image. And obtaining the final image motion amount of each pixel of the multi-slit hyperspectral data.

Description

Technical Field

[0001] The invention relates to multi-slit hyperspectral image data, and in particular to an image motion detection and correction method for multi-slit hyperspectral image data. Background Art

[0002] The multi-slit hyperspectral imaging system acquires the overall image data of the entire object surface element by push-scanning along the track. Since the multi-slit hyperspectral imaging system adopts multi-slit spectrometry, it can acquire the hyperspectral image data of different columns of elements at a single moment, and then acquire multiple sets of hyperspectral image data of the same element after different slit spectrometry at different moments by push-scanning along the track, so there is correlation between the multi-slit hyperspectral image data; at the same time, due to the influence of optical distortion, imaging conditions and other factors, the multi-slit hyperspectral imaging system has slight image shift differences between the hyperspectral image data of each slit, so there is also difference between the multi-slit hyperspectral image data. As an emerging hyperspectral imaging technology system, the detection method for the image shift of its independent pixels is still relatively lacking, which restricts the embodiment of the advantages of the multi-slit hyperspectral imaging system and the high-quality application of its data.

[0003] Image motion detection of multi-slit hyperspectral image data is the basis for supporting high-quality applications of multi-slit hyperspectral data. It is of great significance for high-precision reconstruction of multi-slit data, super-resolution reconstruction based on image motion, and the embodiment of the advantages of multi-slit hyperspectral imaging systems. However, there is currently no detection method for the image motion difference of each pixel between multi-slit hyperspectral data. Summary of the invention

[0004] In order to solve the technical problem that there is no detection method for the image motion difference between pixels in multi-slit hyperspectral data in the prior art, which restricts the realization of the advantages of the multi-slit hyperspectral imaging system and the high-quality application of its data, the present invention provides an image motion detection and correction method for multi-slit hyperspectral image data.

[0005] The inventive concept of the present invention:

[0006] Since the image shift of each pixel between multi-slit hyperspectral image data can be regarded as the movement of each image pixel, the HS dense optical flow method (i.e., Horn-Schunck dense optical flow method) can be used to detect and correct the image shift of independent pixels. The HS dense optical flow method is a classic dense optical flow calculation method that can calculate the optical flow value of each pixel. In the calculation, the smoothness of the calculated optical flow field is characterized and adjusted by the smoothness coefficient factor. The multi-slit hyperspectral imaging system is an on-track push-scan imaging system. During the push-scan process, the smoothness of the optical flow field in the along-track direction and the vertical track direction is greatly different, but the optical flow smoothness of each pixel in the vertical track direction is better than the optical flow smoothness in the along-track direction. Therefore, different smoothness coefficient factors can be used in the two directions for optical flow calculation.

[0007] The image shift amount of each pixel can also be obtained by superimposing the known image shift deviation of the multi-slit hyperspectral imaging system and the imaging image shift deviation caused by the platform motion to obtain the image shift measurement amount. Since the image shift measurement amount is relatively rough, it can reflect the distribution law of image shift. It can be used to correct the optical flow calculation results, eliminate abnormal optical flow values, and obtain more accurate image shift amounts.

[0008] Therefore, the present invention regards the image shift of each pixel between multi-slit hyperspectral data as the image shift of each image, and there is continuity between the image shifts of each pixel. The HS dense optical flow method is used to detect and correct the image shift of each pixel using different image shift continuity constraints in the vertical and horizontal directions. Then, based on the distortion image shift deviation of the multi-slit hyperspectral imaging system and the imaging image shift deviation caused by platform motion, the optical flow value calculated by the HS dense optical flow method is corrected, the abnormal optical flow value is eliminated, and the optical flow value of the eliminated pixel position is supplemented by the linear difference calculation result to obtain the final image shift of each pixel in the multi-slit hyperspectral data.

[0009] In order to achieve the above purpose and complete the above invention concept, the present invention adopts the following technical solutions:

[0010] A method for detecting and correcting image motion of multi-slit hyperspectral image data, which is special in that it includes the following steps:

[0011] Step 1, obtaining multiple hyperspectral image data of the same target pushed by a multi-slit hyperspectral imaging system, and preprocessing them, setting the hyperspectral image data obtained by one of the slits as a benchmark, recorded as the first hyperspectral image data, and the rest as the second hyperspectral image data;

[0012] Step 2, taking the average of the grayscale values ​​of each spectrum segment in the first hyperspectral image data, determining the spectrum segment T1 with the largest average value, and finding the spectrum segment T2 corresponding to the spectrum segment T1 in each second hyperspectral image data;

[0013] Step 3, using the HS dense optical flow method, different smoothing coefficient factors are set for the along-track direction and the perpendicular-track direction, and the optical flow value of each pixel in the spectrum segment T2 relative to each pixel in the spectrum segment T1 is calculated, and all the optical flow values ​​are recorded as the first image shift array;

[0014] Step 4, superimposing the distortion image shift deviation and the imaging image shift deviation respectively to obtain a second image shift array; the distortion image shift deviation is the system deviation of the multi-slit hyperspectral imaging system; the imaging image shift deviation is the imaging deviation caused by each pixel in the spectrum segment T1 and each corresponding pixel in the spectrum segment T2 due to imaging at different push-scanning moments;

[0015] Step 5, correcting the first image shift array by using the second image shift array;

[0016] 5.1) Calculating the pixel deviations of the first image shift array and the second image shift array accordingly;

[0017] 5.2) Compare the deviation of each pixel with the preset threshold:

[0018] If the pixel deviation is less than or equal to the preset threshold, the corresponding optical flow value is retained;

[0019] If the pixel deviation is greater than a preset threshold, the corresponding optical flow value is removed, and the remaining and adjacent optical flow values ​​in the front and rear positions in the vertical direction are used for linear interpolation calculation, and the calculation result is used as the optical flow value of the removed pixel position;

[0020] The correction of the first image shift array is completed, and the corrected first image shift array is the image shift detection correction result of the multi-slit hyperspectral image data.

[0021] Furthermore, in step 3, the vertical-to-track direction smoothness term coefficient factor α is set to twice the along-track direction smoothness term coefficient factor β.

[0022] Furthermore, in step 3, the smoothing term coefficient factor β is set to (0, 0.5].

[0023] Furthermore, in step 5.2, the preset threshold is 1-2 times of the average deviation of the corresponding pixels in the first image shift array and the second image shift array.

[0024] Furthermore, in step 5.2, the preset threshold is 1.5 times the average deviation of the corresponding pixels in the first image shift array and the second image shift array.

[0025] Furthermore, in step 1, the hyperspectral image data obtained by a slit with the best imaging quality is set as a reference, which is recorded as the first hyperspectral image data.

[0026] Furthermore, in step 1, the preprocessing includes dark current correction, relative radiation correction and bad pixel correction.

[0027] Furthermore, in step 3, the HS dense optical flow method is solved using a Gauss-Seidel iterative method.

[0028] Beneficial effects of the present invention:

[0029] In view of the characteristics of multi-slit hyperspectral image data, the present invention uses the HS dense optical flow method and the system's distortion image motion deviation and imaging image motion deviation to provide an image motion detection and correction method for multi-slit hyperspectral image data. The method can obtain high-precision image motion of each pixel between multi-slit hyperspectral image data, provides technical support for high-precision reconstruction of multi-slit data, super-resolution reconstruction based on image motion, and high-quality application of multi-slit hyperspectral data, and is of great significance for embodying the advantages of multi-slit hyperspectral imaging systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 This is a schematic diagram of push-scan imaging of a three-slit hyperspectral imaging system platform;

[0031] Figure 2 The present invention discloses a flow chart of an image motion detection and correction method for multi-slit hyperspectral image data. DETAILED DESCRIPTION

[0032] The technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings and embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0033] A method for image motion detection and correction of multi-slit hyperspectral image data, combined with Figure 1 and Figure 2 As shown, this embodiment uses a double-slit hyperspectral imaging system, the spatial dimension along the track direction × vertical track direction is 20×10, and the spectral dimension is 5 spectral bands. In other embodiments, a three-slit hyperspectral imaging system can also be used.

[0034] The method comprises the following steps:

[0035] Step 1, obtain multiple hyperspectral image data of the same target pushed by the multi-slit hyperspectral imaging system, set the hyperspectral image data obtained by one of the slits as the benchmark, recorded as the first hyperspectral image data, and the rest as the second hyperspectral image data; independently and sequentially perform the following data preprocessing on the first hyperspectral image data and the second hyperspectral image data: dark current correction, relative radiation correction, and bad pixel correction.

[0036] The first hyperspectral image data is selected as the hyperspectral image data obtained by a slit with the best imaging quality.

[0037] Step 2: average the grayscale values ​​of each spectral segment (5 spectral segments) in the first hyperspectral image data, determine the spectral segment T1 with the largest average value, and find the spectral segment T2 corresponding to the spectral segment T1 in each second hyperspectral image data.

[0038] In this embodiment, the spectrum segment 3 has the largest mean value, and the spectrum segment 3 corresponding to the second hyperspectral image data of the spectrum segment 3 is found.

[0039] Step 3, using the HS dense optical flow method, and setting different smoothing coefficient factors for the along-track direction and the perpendicular-track direction, calculate the optical flow value of each pixel in the spectrum segment T2 relative to each pixel in the spectrum segment T1, and record all the optical flow values ​​as the first image shift array; then each value in the first image shift array corresponds to a corresponding pixel in the spectrum segment T1 and the spectrum segment T2.

[0040] The HS dense optical flow method is solved using the Gauss-Seidel iterative method. The optical flow smoothing coefficient factor of the traditional HS dense optical flow method takes the same value in both the along-track direction and the vertical-track direction. In this embodiment, due to the particularity of multi-slit data acquisition, the vertical-track direction smoothing coefficient factor α is taken as twice the along-track direction smoothing coefficient factor β, specifically α=1, β=0.5.

[0041] Step 4: Superimpose the distortion image shift deviation and the imaging image shift deviation respectively to obtain a second image shift array.

[0042] in:

[0043] The distortion image shift deviation is the system deviation of the multi-slit hyperspectral imaging system, which is a known quantity.

[0044] The imaging shift deviation is the imaging deviation caused by each pixel in the spectrum segment T1 and each corresponding pixel in the spectrum segment T2 due to imaging at different push-scanning moments;.

[0045] Therefore, each value in the second image shift array also corresponds to a corresponding pixel in the spectrum segment T1 and the spectrum segment T2.

[0046] Step 5, correcting the first image shift array by using the second image shift array;

[0047] 5.1) Calculating the pixel deviations of the first image shift array and the second image shift array accordingly;

[0048] 5.2) Compare the deviation of each pixel with the preset threshold:

[0049] If the pixel deviation is less than or equal to the preset threshold, the corresponding optical flow value is retained;

[0050] If the pixel deviation is greater than a preset threshold, the corresponding optical flow value is removed, and linear interpolation calculation is performed using the remaining and adjacent optical flow values ​​at the front and rear positions in the vertical direction, and the calculation result is used as the optical flow value of the removed pixel position.

[0051] The preset threshold is 1-2 times of the average deviation of the corresponding pixels in the first image shift array and the second image shift array, and preferably 1.5 times in this embodiment.

[0052] The correction of the first image shift array is completed, and the corrected first image shift array is the image shift detection correction result of the multi-slit hyperspectral image data.

[0053] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions within the technical scope disclosed by the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for detecting and correcting image motion of multi-slit hyperspectral image data, characterized in that: The following steps are involved: Step 1, obtaining multiple hyperspectral image data of the same target pushed by a multi-slit hyperspectral imaging system, and preprocessing them, setting the hyperspectral image data obtained by one of the slits as a benchmark, recorded as the first hyperspectral image data, and the rest as the second hyperspectral image data; Step 2, taking the average of the grayscale values ​​of each spectrum segment in the first hyperspectral image data, determining the spectrum segment T1 with the largest average value, and finding the spectrum segment T2 corresponding to the spectrum segment T1 in each second hyperspectral image data; Step 3, using the HS dense optical flow method, different smoothing coefficient factors are set for the along-track direction and the perpendicular-track direction, and the optical flow value of each pixel in the spectrum segment T2 relative to each pixel in the spectrum segment T1 is calculated, and all the optical flow values ​​are recorded as the first image shift array; Step 4, superimposing the distortion image shift deviation and the imaging image shift deviation respectively to obtain a second image shift array; the distortion image shift deviation is the system deviation of the multi-slit hyperspectral imaging system; the imaging image shift deviation is the imaging deviation caused by each pixel in the spectrum segment T1 and each corresponding pixel in the spectrum segment T2 due to imaging at different push-scanning moments; Step 5, correcting the first image shift array by using the second image shift array; 5.1) Calculating the pixel deviations of the first image shift array and the second image shift array accordingly; 5.2) Compare the deviation of each pixel with the preset threshold: If the pixel deviation is less than or equal to the preset threshold, the corresponding optical flow value is retained; If the pixel deviation is greater than a preset threshold, the corresponding optical flow value is removed, and the remaining and adjacent optical flow values ​​in the front and rear positions in the vertical direction are used for linear interpolation calculation, and the calculation result is used as the optical flow value of the removed pixel position; The correction of the first image shift array is completed, and the corrected first image shift array is the image shift detection correction result of the multi-slit hyperspectral image data.

2. The image motion detection and correction method for multi-slit hyperspectral image data according to claim 1, characterized in that: In step 3, the smoothness factor α in the vertical-to-track direction is set to twice the smoothness factor β in the along-track direction.

3. The image motion detection and correction method for multi-slit hyperspectral image data according to claim 2, characterized in that: In step 3, set the smoothing factor β to (0, 0.5].

4. The image motion detection and correction method for multi-slit hyperspectral image data according to claim 2 or 3, characterized in that: In step 5.2, the preset threshold is 1-2 times the average deviation of the corresponding pixels in the first image shift array and the second image shift array.

5. The image motion detection and correction method for multi-slit hyperspectral image data according to claim 4, characterized in that: In step 5.2, the preset threshold is 1.5 times the average deviation of the corresponding pixels in the first image shift array and the second image shift array.

6. The image motion detection and correction method for multi-slit hyperspectral image data according to claim 5, characterized in that: In step 1, the hyperspectral image data obtained from a slit with the best imaging quality is set as a reference, which is recorded as the first hyperspectral image data.

7. The image motion detection and correction method for multi-slit hyperspectral image data according to claim 1, characterized in that: In step 1, the preprocessing includes dark current correction, relative radiation correction and bad pixel correction.

8. The image motion detection and correction method for multi-slit hyperspectral image data according to claim 7, characterized in that: In step 3, the HS dense optical flow method is solved using the Gauss-Seidel iterative method.

Images