A method for radiometric uniformity correction of aerial hyperspectral images

Abstract

The present invention relates to a method for correcting the radiometric uniformity of aerial hyperspectral images, comprising the following steps: S1, acquiring aerial hyperspectral images and dividing the images to be corrected into two categories: water bodies and land features; S2, using least squares regression to solve the regression coefficients and perform band-by-band solar flare correction; S3, calculating the observation angle and azimuth corresponding to each scan line, establishing a binary quartic empirical correction model, and performing one-by-one normalization correction; S4, calculating the corrected land radiometric brightness value; S5, calculating the corrected water radiometric brightness value; S6, repeating S1-S5 to complete the image uniformity correction for multiple flight strips and obtain seamless mosaic data for the multiple flight strips. Compared with the prior art, the present invention eliminates the radiometric brightness inhomogeneity caused by the surface birefringence effect and achieves uniformity correction and seamless mosaicking of multi-strip hyperspectral images under different imaging conditions.

Description

Technical Field

[0001] The present invention relates to the field of hyperspectral image correction, and in particular to a method for correcting radiation uniformity of aerial hyperspectral images. Background Art

[0002] During the imaging process of aerial hyperspectral images, due to the influence of the bidirectional reflectance distribution function (BRDF) effect, there are changes in the radiation intensity in the acquired images, resulting in overall image inhomogeneity. At the same time, the radiation intensity of the same feature is inconsistent between adjacent strips, resulting in obvious brightness gradients when multiple flight strips are stitched together, affecting the accuracy of subsequent qualitative and quantitative applications and the reliability of the data. Therefore, it is necessary to correct the radiation uniformity of the images to provide accurate radiation brightness data for the construction of models such as fine feature classification, soil heavy metal inversion, and vegetation physical and chemical parameter inversion.

[0003] The anisotropic reflectance characteristics caused by the BRDF are related to the surface structure of the observed object. Common examples include specular reflection from objects, sunlight reflection from water surfaces, volume scattering from leaves, and reflection from vegetation canopies. Due to the unique anisotropy of different observed object surfaces, establishing correction models for scenes containing multiple ground feature categories is difficult. This is especially true for imaging scenes with high spatial complexity and large differences in radiant brightness, such as aquatic and terrestrial environments. Uniform correction and seamless mosaicking of data across multiple flight strips under different imaging conditions are difficult. Summary of the Invention

[0004] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and provide a method for correcting the radiation uniformity of aerial hyperspectral images, eliminate the uneven radiation brightness caused by the bidirectional reflection effect of the surface, and realize the uniformity correction and seamless mosaicking of multi-band hyperspectral images under different imaging conditions based on the reflection characteristics of the surfaces of different observed objects.

[0005] The purpose of the present invention can be achieved by the following technical solutions:

[0006] A method for correcting the radiometric uniformity of an aerial hyperspectral image comprises the following steps:

[0007] S1. Acquire an aerial hyperspectral image, use the image as the image to be corrected, perform water body separation on the image to be corrected based on the normalized water body index, and divide the image to be corrected into two categories: water body and land surface features;

[0008] S2. Filter pixels in the image to be corrected after water separation obtained in S1, use least squares regression to solve the regression coefficient, and perform solar flare correction on all pixels in the image to be corrected after water separation, band by band;

[0009] S3. For the image to be corrected after solar flare correction obtained in S2, the observation angle and azimuth corresponding to each scanning line of the image are calculated line by line, the observation angle and azimuth are used as independent variables, and the corresponding ground object radiation brightness is used as the dependent variable, a binary quartic empirical correction model is established, and the observation angle and azimuth are normalized and corrected one by one based on the correction model;

[0010] S4. Masking the image to be corrected after the observation angle and azimuth correction obtained in S3 to obtain a first image retaining land features, performing land feature uniformity correction on the first image, and calculating a corrected land radiance value;

[0011] S5. Masking the image to be corrected after the solar flare correction in S2 to obtain a second image containing only the water body, performing water body uniformity correction on the second image according to the correction method for the observation angle and azimuth angle in S3, and calculating the corrected water body radiance value;

[0012] S6. Repeat S1-S5 to complete the image uniformity correction of multiple flight strips, and superimpose the first image after the land surface uniformity correction and the second image after the water body uniformity correction respectively. Based on the corrected land radiation brightness value and the corrected water body radiation brightness value, obtain complete corrected radiation brightness image data, perform geometric correction processing and mosaic splicing on the data, and obtain seamless mosaic data of multiple flight strips.

[0013] Furthermore, pixels are screened in S2, and the least squares regression is used to solve the regression coefficient for solar flare correction. Specifically, the water body radiation brightness value of the near-infrared band part of the image to be corrected after water body separation is counted, and the pixel with the lowest radiation brightness value in the near-infrared band is screened and recorded. The linear relationship between the near-infrared band and the visible light band is established through least squares regression, and the regression coefficient is solved. The solar flare correction is performed based on the radiation brightness value and regression coefficient of the pixel with the lowest radiation brightness value.

[0014] Furthermore, the calculation expression for solar flare correction is:

[0015] L′ i =L i -b i (L NIR -L NIR_min )

[0016] Among them, L i is the original radiation brightness of the i-th band, L′ i is the radiance after solar flare correction in band i, b i is the regression coefficient, L NIR is the original radiation brightness in the near-infrared band, L NIR_minIt is the radiance value of the pixel with the lowest radiance value.

[0017] Furthermore, the expressions of observation angle and azimuth angle are:

[0018]

[0019]

[0020] Where θ is the observation angle, is the azimuth, and x, y, and z are the three-dimensional coordinates corresponding to the pixel point.

[0021] Furthermore, the expression of the corrected land radiance value is:

[0022]

[0023] in, is the corrected land radiation brightness value, α is the solar zenith angle, θ is the observation angle, is the azimuth, is the normalized solar zenith angle, θ v is the normalized observation angle, is the normalized azimuth, is the uncorrected land radiance value, and f is the correction function of the observation angle and azimuth.

[0024] Furthermore, the expression of the corrected water body radiation brightness value is:

[0025]

[0026] in, is the corrected water body radiation brightness value, α is the solar zenith angle, θ is the observation angle, is the azimuth, is the normalized solar zenith angle, θ v is the normalized observation angle, is the normalized azimuth, is the uncorrected water body radiation brightness value, and f is the correction function of observation angle and azimuth angle.

[0027] Furthermore, the normalized observation angle is 10°, the normalized azimuth angle is 180°, and the normalized observation angle and the normalized azimuth angle are obtained by calculating the average observation angle and azimuth angle values ​​of each flight strip.

[0028] Furthermore, the expression of the binary quartic empirical correction model is:

[0029]

[0030] Where, f is the correction function of observation angle and azimuth angle, θ is the observation angle, is the azimuth, and a0, a1, a2, a3, a4, b1, b2, b3, and b4 are all model coefficients.

[0031] Furthermore, the model coefficients are obtained by fitting the sensor viewing angle and the corresponding image average radiance value based on the assumption of uniformity of ground object radiance.

[0032] Furthermore, water bodies are separated from the image to be corrected based on the normalized water body index, and the image to be corrected is divided into two categories: water bodies and land surface features. Specifically, the normalized water body index is calculated, and a threshold is set to perform binarization processing, and the image to be corrected is divided into two categories: water bodies and land surface features. The threshold is 0.2.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] (1) The present invention considers the difference in radiation brightness between land and water, separates the water body of the image to be corrected based on the normalized water body index, and divides the image to be corrected into two categories: water body and land surface features. The water body and other surfaces with large differences in radiation brightness are corrected separately, which is conducive to establishing a more accurate correction model.

[0035] (2) The present invention takes into account the push-scan characteristics of hyperspectral imaging and calculates the observation angle and azimuth angle corresponding to each scanning line line by line, thereby eliminating the differences between scanning lines.

[0036] (3) The present invention performs normalization correction on the radiation difference of water bodies, taking into account the influence of multiple factors such as radiation transmission, solar flares, and sensor viewing angle, to achieve uniformity correction and seamless mosaicking of data in multiple flight strips under different imaging conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 is a flow chart of the present invention;

[0038] Figure 2 Schematic diagram of the water body solar flare correction method of the present invention;

[0039] Figure 3 This is a comparison chart of the spectral radiation brightness of water before and after correction;

[0040] Figure 4 is the land surface feature image before correction;

[0041] Figure 5 is the corrected land surface feature image;

[0042] Figure 6 This is the water body image before correction;

[0043] Figure 7is the corrected water body image;

[0044] Figure 8 This is the full mosaic result of the image before correction;

[0045] Figure 9 This is the full mosaic result of the corrected image. DETAILED DESCRIPTION

[0046] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0047] The present invention provides a method for correcting the radiation uniformity of aerial hyperspectral images applied to water and land scenes. The method eliminates the unevenness of radiation brightness caused by the bidirectional reflection effect of the surface, and realizes the uniformity correction and seamless mosaicking of multi-band hyperspectral images under different imaging conditions based on the reflection characteristics of the surfaces of different observed objects.

[0048] The flow chart of the radiometric uniformity correction method for aerial hyperspectral images is as follows: Figure 1 As shown, the method includes the following steps:

[0049] S1. Acquire an aerial hyperspectral image, use the image as the image to be corrected, perform water body separation on the image to be corrected based on the normalized water body index, and divide the image to be corrected into two categories: water body and land surface features;

[0050] S2. Filter pixels in the image to be corrected after water separation obtained in S1, use least squares regression to solve the regression coefficient, and perform solar flare correction on all pixels in the image to be corrected after water separation, band by band;

[0051] S3. For the image to be corrected after solar flare correction obtained in S2, the observation angle and azimuth corresponding to each scanning line of the image are calculated line by line, the observation angle and azimuth are used as independent variables, and the corresponding ground object radiation brightness is used as the dependent variable, a binary quartic empirical correction model is established, and the observation angle and azimuth are normalized and corrected one by one based on the correction model;

[0052] S4. Masking the image to be corrected after the observation angle and azimuth correction obtained in S3 to obtain a first image retaining land features, performing land feature uniformity correction on the first image, and calculating a corrected land radiance value;

[0053] S5. Masking the image to be corrected after the solar flare correction in S2 to obtain a second image containing only the water body, performing water body uniformity correction on the second image according to the correction method for the observation angle and azimuth angle in S3, and calculating the corrected water body radiance value;

[0054] S6. Repeat S1-S5 to complete the image uniformity correction of multiple flight strips, and superimpose the first image after the land surface uniformity correction and the second image after the water body uniformity correction respectively. Based on the corrected land radiation brightness value and the corrected water body radiation brightness value, obtain complete corrected radiation brightness image data, perform geometric correction processing and mosaic splicing on the data, and obtain seamless mosaic data of multiple flight strips.

[0055] In S1, the NDWI normalized water index is used to extract water bodies. By setting a threshold and performing binarization, the image is divided into two categories: water bodies and land features. The threshold can be set to 0.2. Pixels with NDWI values ​​greater than 0.2 are set to 1, and the remaining pixel values ​​are set to 0, dividing the image into two categories: water bodies (1) and land features (0). In S1, the radiance of the ground objects includes the radiance of both water bodies and land features.

[0056] In S2, the pixels are screened, and the least squares regression is used to solve the regression coefficient to perform solar flare correction. Specifically, the water body radiation brightness value of the near-infrared band part of the image to be corrected after water separation is counted, and the pixel with the lowest radiation brightness value in the near-infrared band is screened and recorded. The linear relationship between the near-infrared band and the visible light band i is established by least squares regression, and the regression coefficient b is solved. i , based on the radiance value and regression coefficient of the pixel with the lowest radiance value, the solar flare correction is performed. The schematic diagram of the water body solar flare correction method is shown in the figure. Figure 2 The comparison of the spectral radiation brightness of water before and after correction is shown in the figure below. Figure 3 shown.

[0057] The calculation expression for solar flare correction is:

[0058] L′ i =L i -b i (L NIR -L NIR_min )

[0059] Among them, L i is the original radiation brightness of the i-th band, L′ i is the radiance after solar flare correction in band i, b i is the regression coefficient, L NIR is the original radiation brightness in the near-infrared band, L NIR_min It is the radiance value of the pixel with the lowest radiance value.

[0060] In S3, the observation angle and azimuth corresponding to each scan line are calculated line by line. According to the collinear equation of photogrammetry, the observation angle θ and azimuth corresponding to the pixel point (x0, y0) are The calculation formula is as follows:

[0061]

[0062] Where θ is the observation angle, is the azimuth, and x, y, and z are the three-dimensional coordinates corresponding to the pixel point (x0, y0).

[0063] Taking the observation angle and azimuth as independent variables and the corresponding ground object radiation brightness as the dependent variable, a binary quartic empirical correction model is established:

[0064]

[0065] Where, f is the correction function of observation angle and azimuth angle, θ is the observation angle, is the azimuth angle, and a0, a1, a2, a3, a4, b1, b2, b3, and b4 are all model coefficients. The model coefficients are obtained by fitting the sensor viewing angle and the corresponding image average radiance value based on the assumption of uniformity of the ground object radiance. Based on this correction model, the BRDF correction function model coefficients are solved, and the observation angle and azimuth angle are normalized one by one.

[0066] In S4, after obtaining the first image that retains the land surface features, the Lommel-Seeliger factor is introduced to normalize the solar incidence angle and the sensor observation angle. The expression of the corrected land radiation brightness value is:

[0067]

[0068] in, is the corrected land radiation brightness value, α is the solar zenith angle, θ is the observation angle, is the azimuth, is the normalized solar zenith angle, θ v is the normalized observation angle, is the normalized azimuth, is the uncorrected land radiance value, f is the correction function for the observation angle and azimuth. The Lommel-Seeliger factor is a factor of the ellipsoidal photometric model.

[0069] The land surface image before and after correction is as follows Figure 4 and Figure 5 shown.

[0070] In S5, after obtaining the second image containing only the water body, the Lommel-Seeliger factor and BRDF correction function are also introduced and correction is performed according to the method in S3. The expression of the corrected water body radiance value is:

[0071]

[0072] in, is the corrected water body radiation brightness value, α is the solar zenith angle, θ is the observation angle, is the azimuth, is the normalized solar zenith angle, θ v is the normalized observation angle, is the normalized azimuth, is the uncorrected water body radiation brightness value, and f is the correction function of observation angle and azimuth angle.

[0073] The water body images before and after correction are as follows Figure 6 and Figure 7 shown.

[0074] In some embodiments, the normalized observation angle in S4 and S5 is 10°, the normalized azimuth angle is 180°, and the normalized solar zenith angle is 40°.

[0075] In S6, the result of the corrected image superposition is as follows Figure 9 As shown, the results of image superposition before and after correction are compared Figure 8 , Figure 8 There is a significant radiometric inconsistency, with noticeable brightness gradients and edge banding. Therefore, the correction method of the present invention effectively eliminates radiometric inconsistencies, corrects the image's radiometric brightness gradients and the brightness differences between the obvious bands, and produces seamless mosaicking data of multiple flight strips with high radiometric brightness consistency.

[0076] The above describes in detail the preferred embodiments of the present invention. It should be understood that those skilled in the art can make numerous modifications and variations based on the concepts of the present invention without inventive effort. Therefore, any technical solutions that can be derived by those skilled in the art through logical analysis, reasoning, or limited experimentation based on the concepts of the present invention and the prior art should be within the scope of protection defined by the claims.

Claims

1. A method for correcting the radiometric uniformity of aerial hyperspectral images, characterized in that: The following steps are involved: S1. Acquire an aerial hyperspectral image, use the image as the image to be corrected, perform water body separation on the image to be corrected based on the normalized water body index, and divide the image to be corrected into two categories: water body and land surface features; S2. Filter pixels in the image to be corrected after water separation obtained in S1, use least squares regression to solve the regression coefficient, and perform solar flare correction on all pixels in the image to be corrected after water separation, band by band; S3. For the image to be corrected after solar flare correction obtained in S2, the observation angle and azimuth corresponding to each scanning line of the image are calculated line by line, the observation angle and azimuth are used as independent variables, and the corresponding ground object radiation brightness is used as the dependent variable, a binary quartic empirical correction model is established, and the observation angle and azimuth are normalized and corrected one by one based on the correction model; S4. Masking the image to be corrected after the observation angle and azimuth correction obtained in S3 to obtain a first image retaining land features, performing land feature uniformity correction on the first image, and calculating a corrected land radiance value; S5. Masking the image to be corrected after the solar flare correction in S2 to obtain a second image containing only the water body, performing water body uniformity correction on the second image according to the correction method for the observation angle and azimuth angle in S3, and calculating the corrected water body radiance value; S6. Repeat S1-S5 to complete the image uniformity correction of multiple flight strips, and superimpose the first image after the land surface uniformity correction and the second image after the water body uniformity correction respectively. Based on the corrected land radiation brightness value and the corrected water body radiation brightness value, obtain complete corrected radiation brightness image data, perform geometric correction processing and mosaic splicing on the data, and obtain seamless mosaic data of multiple flight strips.

2. The method for correcting the radiometric uniformity of aerial hyperspectral images according to claim 1, characterized in that: In S2, pixels are screened, and least squares regression is used to solve the regression coefficient for solar flare correction. Specifically, the water body radiation brightness values ​​of the near-infrared band part of the image to be corrected after water body separation are counted, and the pixel with the lowest radiation brightness value in the near-infrared band is screened and recorded. The linear relationship between the near-infrared band and the visible light band is established through least squares regression, and the regression coefficient is solved. Solar flare correction is performed based on the radiation brightness value and regression coefficient of the pixel with the lowest radiation brightness value.

3. The method for correcting the radiometric uniformity of aerial hyperspectral images according to claim 2, characterized in that: The calculation expression for solar flare correction is: The i =L i -b i (L NIR -L NIR_min ) Among them, L i is the original radiation brightness of the i-th band, L′ i is the radiance after solar flare correction in band i, b i is the regression coefficient, L NIR is the original radiation brightness in the near-infrared band, L NIR_min It is the radiance value of the pixel with the lowest radiance value.

4. The method for correcting radiometric uniformity of aerial hyperspectral images according to claim 1, characterized in that: The expressions of observation angle and azimuth angle are: Where θ is the observation angle, is the azimuth, and x, y, and z are the three-dimensional coordinates corresponding to the pixel point.

5. The method for correcting radiometric uniformity of aerial hyperspectral images according to claim 1, characterized in that: The expression of the corrected land radiation brightness value is: in, is the corrected land radiation brightness value, α is the solar zenith angle, θ is the observation angle, is the azimuth, is the normalized solar zenith angle, θ v is the normalized observation angle, is the normalized azimuth, is the uncorrected land radiance value, and f is the correction function of the observation angle and azimuth.

6. The method for correcting the radiometric uniformity of aerial hyperspectral images according to claim 1, characterized in that: The expression of the corrected water body radiation brightness value is: in, is the corrected water body radiation brightness value, α is the solar zenith angle, θ is the observation angle, is the azimuth, is the normalized solar zenith angle, θ v is the normalized observation angle, is the normalized azimuth, is the uncorrected water body radiation brightness value, and f is the correction function of observation angle and azimuth angle.

7. The method for correcting the radiometric uniformity of aerial hyperspectral images according to claim 5 or 6, characterized in that: The normalized observation angle is 10°, and the normalized azimuth is 180°. The normalized observation angle and normalized azimuth are obtained by calculating the average observation angle and azimuth values ​​of each flight strip.

8. The method for correcting radiometric uniformity of aerial hyperspectral images according to claim 1, characterized in that: The expression of the binary quartic empirical calibration model is: Where, f is the correction function of observation angle and azimuth angle, θ is the observation angle, is the azimuth, and a0, a1, a2, a3, a4, b1, b2, b3, and b4 are all model coefficients.

9. The method for correcting radiometric uniformity of aerial hyperspectral images according to claim 8, characterized in that: The model coefficients are obtained by fitting the sensor viewing angle and the corresponding image average radiance value based on the assumption of uniformity of ground object radiance.

10. The method for correcting radiometric uniformity of aerial hyperspectral images according to claim 1, characterized in that: Based on the normalized water body index, the water body of the image to be corrected is separated, and the image to be corrected is divided into two categories: water body and land surface features. Specifically, the normalized water body index is calculated, and a threshold is set, and a binarization process is performed to divide the image to be corrected into two categories: water body and land surface features. The threshold is 0.2.

Images