A remote sensing image color uniformization method based on surface reflectivity and surface type

Abstract

This invention discloses a method for color balancing remote sensing images based on surface reflectance and surface type. The method first inputs the remote sensing image and its auxiliary parameters, performs image calibration, converts DN values ​​to apparent radiance, and then further converts them to apparent reflectance. Subsequently, based on aerosol parameters, atmospheric correction is performed using a radiative transfer model to obtain the surface reflectance. Next, orthorectification of the surface reflectance is performed based on RPC information, and combined with surface type data, surface type information within the image range is obtained through geometric location matching. Then, a reference image is selected, and the mean and variance of surface reflectance for each channel of both the reference image and the image to be sized are calculated for different surface types. Finally, the mean-variance method is used to color-balance the image to be sized, obtaining a color-balanced surface reflectance image. This invention can effectively eliminate color deviations between remote sensing images, improving the color consistency and ground feature realism of mosaic images.

Description

Technical Field

[0001] This invention relates to the field of satellite remote sensing image processing technology, and in particular to a method for color balancing remote sensing images based on surface reflectance and surface type. Background Technology

[0002] In large-scale remote sensing applications, mosaicking of remote sensing images from different orbits and time phases is often required. However, due to the combined effects of various factors such as differences in sensor parameters, changes in lighting conditions, atmospheric effects, and seasonal changes in ground features, there are often significant differences in color and radiometry between adjacent images. This results in poor color consistency in the mosaicking results, which seriously restricts the accuracy and reliability of subsequent applications such as visualization interpretation and thematic mapping.

[0003] Therefore, before implementing any specific application, the mosaic images need to undergo color balancing to eliminate color deviations between images. Currently, the mainstream color balancing methods include mathematical modeling, histogram matching, and mean-variance methods. Mathematical modeling typically constructs a statistical model based on "invariant pixel" samples in overlapping areas of multiple images and uses linear or nonlinear transformations to achieve color correction. Histogram matching forcibly adjusts the grayscale distribution of the image to be matched, making it consistent with the overall histogram shape of the reference image, thereby achieving overall color uniformity. The mean-variance method maps the mean and variance of the image to be processed to the corresponding statistics of the target image, making them more consistent in brightness, contrast, and grayscale dynamic range.

[0004] While the aforementioned methods can achieve certain results under specific conditions, they all have limitations in practical applications. First, the mathematical model method is extremely sensitive to the selection of invariant pixels, resulting in significant fluctuations in correction effectiveness. Although histogram matching and mean-variance methods perform well in images with similar land cover types, in large-scale remote sensing scenarios, due to the wide image coverage, complex land cover types, and significant regional differences, these methods are prone to causing overall or local color shifts, making it difficult to maintain the naturalness and consistency of land cover colors. Second, optical remote sensing images are susceptible to atmospheric scattering and absorption, and existing color-balancing methods mostly process the raw digital quantization (DN) values ​​obtained from satellite sensors, failing to effectively remove atmospheric interference, resulting in the mosaicked image still retaining atmospheric blurring effects.

[0005] Therefore, in order to improve the color consistency and ground feature realism of mosaic images, it is urgent to develop a remote sensing image color balancing method that integrates surface reflectance and surface type information to achieve radiation consistency and color that is more in line with visual perception. Summary of the Invention

[0006] The purpose of this invention is to provide a method for color balancing remote sensing images based on surface reflectance and surface type, thereby solving the aforementioned problems existing in the prior art.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0008] A method for color balancing remote sensing images based on surface reflectance and surface type includes the following steps:

[0009] Input remote sensing images, including the DN value, RPC information, spectral response function, calibration coefficients, solar zenith angle, solar azimuth angle, observed zenith angle, and observed azimuth angle;

[0010] The remote sensing images are calibrated to convert DN values ​​into apparent radiance.

[0011] Convert apparent radiance to apparent reflectance;

[0012] Based on aerosol parameters, atmospheric correction of apparent reflectance is performed using a radiative transfer model to obtain surface reflectance.

[0013] Orthorectification of surface reflectance is performed based on RPC information to obtain orthorectified remote sensing images of surface reflectance.

[0014] Based on the land surface type data, the corresponding land surface type within the remote sensing image area is obtained through geometric location matching;

[0015] Select internal or external reference images, and calculate the mean and variance of the surface reflectance values ​​of each channel of the reference image and the remote sensing image to be uniformly colored under different surface types, based on the surface type data.

[0016] The mean-variance method is used to color homogenize the image to be homogenized, and the surface reflectance after color homogenization is obtained.

[0017] Preferably, the formula used in the step of converting DN values ​​to apparent radiance is:

[0018] L λ = DN λ × a λ + b λ ;

[0019] in, L λ Apparent radiance, in W·m -2 ·sr -1 ·μm -1 , a λ and b λ These are the scaling factor gain and offset, respectively. λ Representative band.

[0020] Preferably, the formula used in the step of converting apparent radiance to apparent reflectance is:

[0021] ;

[0022] in, Apparent reflectance, d The Earth-Sun distance, ESUN This represents the solar irradiance at the top of the atmosphere. This is the solar zenith angle.

[0023] Preferably, the formula used in the atmospheric correction step is:

[0024] ;

[0025] in, SR λ For surface reflectance, Atmospheric path radiation, S Atmospheric albedo The downward transmittance of the atmosphere. It represents the upward transmittance of the atmosphere.

[0026] Preferably, the formula used in the step of color-smoothing the image to be color-smoothed is:

[0027] ;

[0028] in, The surface reflectance after color homogenization. and These are the mean and variance of the reference image, respectively. and These are the mean and variance of the image to be corrected, respectively.

[0029] Preferably, the remote sensing image is multispectral remote sensing data from the Gaofen-7 satellite, including observation data from three channels: 460nm, 550nm, and 670nm.

[0030] Preferably, atmospheric correction is based on the 6SV radiative transfer model.

[0031] The beneficial effects of this invention are:

[0032] This invention improves the color balancing method for remote sensing images based on surface reflectance and surface type data. Compared with traditional color balancing methods, it has the following effects: (1) It removes the influence of the atmosphere in remote sensing images, improves the clarity and contrast of remote sensing images, and makes the colors of ground features more realistic; (2) It constructs a mean-variance method for different surface types, which can maintain better color consistency among different ground features; (3) It is suitable for large-scale remote sensing image mosaic processing and does not require additional preparation of reference base maps. Attached Figure Description

[0033] Figure 1 This is a flowchart of a remote sensing image color balancing method based on surface reflectance and surface type proposed in this invention.

[0034] Figure 2 This is a comparison image of remote sensing images before and after color balancing according to the present invention. Detailed Implementation

[0035] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0036] Reference Figure 1 and Figure 2 The method for color balancing remote sensing images based on surface reflectance and surface type, as shown, includes the following steps:

[0037] (1) Input remote sensing image, including the DN value, RPC information, spectral response function, calibration coefficient, solar zenith angle, solar azimuth angle, observation zenith angle, and observation azimuth angle;

[0038] (2) According to formula (1), convert the DN value of the remote sensing image into apparent radiance. The unit is W·m -2 ·sr -1 ·μm -1 , a λ and b λ These are the scaling factor gain and offset, respectively:

[0039] L λ = DN λ × a λ + b λ (1)

[0040] Representative band, a Represents the scaling gain coefficient.b This represents the offset.

[0041] (3) According to formula (2), convert radiance to apparent reflectance:

[0042] (2)

[0043] Represents apparent reflectance. d Represents the Earth-Sun distance. ESUN Represents solar irradiance at the top of the atmosphere. It represents the zenith angle of the sun.

[0044] (4) Based on aerosol parameters, atmospheric correction is performed according to formula (3) to obtain the surface reflectance. :

[0045] (3)

[0046] This indicates that the atmosphere radiates. S Represents atmospheric albedo. Represents atmospheric transmittance downwards.

[0047] Represents atmospheric upward transmittance, based on aerosol parameters. , S， ， The parameters can be obtained from the radiative transfer model.

[0048] (5) Perform orthorectification on the surface reflectance SR of the remote sensing image based on the RPC information to obtain the orthorectified surface reflectance remote sensing image.

[0049] (6) Based on the land surface type data, obtain the corresponding land surface type within the remote sensing image range through geometric location matching;

[0050] (7) Select internal / external reference images, and calculate the mean and variance of the surface reflectance values ​​of each channel of the reference and the remote sensing images to be uniformly colored in different surface types according to formulas (4) and (5).

[0051] (4)

[0052] (5)

[0053] In the formula, This represents the mean. Represents variance. t Represents land surface type. Representative channel,n This represents the total number of pixels. This represents the surface reflectance value of a pixel;

[0054] (8) Using the mean-variance method, the image to be color-matched is color-matched according to formula (6) to obtain the surface reflectance of the image after color-matching. ;

[0055] (6)

[0056] In the formula, The mean of the reference image. The mean of the image to be corrected is... The variance of the reference image, The variance of the image to be corrected is denoted as . Example

[0057] To achieve this invention, a remote sensing image downscaling and stretching method based on land surface type is provided. This method is applied to multispectral remote sensing data from the Gaofen-7 satellite, and the specific implementation scheme is as follows:

[0058] (1) For the multispectral remote sensing data of Gaofen-7 satellite, the data used in this invention includes observation data of three channels: 460, 550, and 670 nm, as well as RPC information, spectral response function, calibration coefficient, solar zenith angle, solar azimuth angle, observation zenith angle, and observation azimuth angle;

[0059] (2) Based on the RPC information, the observation data of the three channels of Gaofen-7 satellite (460, 550, and 670 nm) are orthorectified to obtain orthorectified remote sensing images;

[0060] (3) According to formula (1), convert the DN values ​​of the three bands of the remote sensing image into apparent radiance. , , The unit is W·m -2 ·sr -1 ·μm -1 , a λ and b λ These are the scaling factor gain and offset, respectively:

[0061] (1)

[0062] a Represents the scaling gain coefficient. b This represents the offset.

[0063] (4) According to formula (2), convert the radiance of the three bands into apparent reflectance. , , :

[0064] (2)

[0065] Represents apparent reflectance, subscript Representative band, L Represents radiance, d Represents the Earth-Sun distance. ESUN Represents solar irradiance at the top of the atmosphere. It represents the zenith angle of the sun.

[0066] (5) Based on the spatiotemporal matching MODIS aerosol parameters, atmospheric correction is performed according to formula (3) to obtain the surface reflectance. , , :

[0067] (3)

[0068] This indicates that the atmosphere radiates. S Represents atmospheric albedo. Represents atmospheric transmittance downwards. Represents atmospheric upward transmittance, based on aerosol parameters. , S， ， The parameters can be obtained from the 6SV radiative transfer model.

[0069] (6) Based on RPC information, analyze the surface reflectance of remotely sensed images. , , Perform orthorectification to obtain orthorectified remote sensing images of the land surface reflectance;

[0070] (7) Based on the land surface type data, obtain the corresponding land surface type within the remote sensing image range through geometric location matching;

[0071] (8) Select internal / external reference images, and calculate the average surface reflectance values ​​of each channel of the reference and the remote sensing images to be uniformly colored in different surface type data according to formulas (4) and (5). , , and variance , , ;

[0072] (4)

[0073] (5)

[0074] In the formula, This represents the mean. Represents variance. t Represents land surface type. Representative channel, n This represents the total number of pixels. This represents the surface reflectance value of a pixel;

[0075] (9) Using the mean-variance method, the image to be color-matched is color-matched according to formula (6) to obtain the surface reflectance of the image after color-matching. , , ;

[0076] (6)

[0077] In the formula, The mean of the reference image. The mean of the image to be corrected is... The variance of the reference image, The variance of the image to be corrected is denoted as .

[0078] By adopting the above-disclosed technical solution of this invention, the following beneficial effects are obtained:

[0079] (1) It removes the influence of the atmosphere in remote sensing images, improves the clarity and contrast of remote sensing images, and makes the colors of ground features more realistic; (2) It constructs the mean-variance method according to the land surface type, which can maintain better color consistency among different ground features; (3) It is suitable for large-scale remote sensing image mosaic processing and does not require additional preparation of reference base maps.

[0080] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for color balancing remote sensing images based on surface reflectance and surface type, characterized in that, Includes the following steps: Input remote sensing images, including the DN value, RPC information, spectral response function, calibration coefficients, solar zenith angle, solar azimuth angle, observed zenith angle, and observed azimuth angle; The remote sensing images are calibrated to convert DN values ​​into apparent radiance. Convert apparent radiance to apparent reflectance; Based on aerosol parameters, atmospheric correction of apparent reflectance is performed using a radiative transfer model to obtain surface reflectance. Orthorectification of surface reflectance is performed based on RPC information to obtain orthorectified remote sensing images of surface reflectance. Based on the land surface type data, the corresponding land surface type within the remote sensing image area is obtained through geometric location matching; Select internal or external reference images, and calculate the mean and variance of the surface reflectance values ​​of each channel of the reference image and the remote sensing image to be uniformly colored under different surface types, based on the surface type data. Based on the mean-variance method, the image to be color-matched is color-matched to obtain the surface reflectance after color matching. The formula used in the atmospheric correction step is: ； Among them, SR λ For surface reflectance, Where S is atmospheric path radiation and S is atmospheric albedo. The downward transmittance of the atmosphere. The upward transmittance of the atmosphere; The formula used in the step of color-smoothing the image to be color-smoothed is: ； in, After color mixing Surface reflectivity For the image to be corrected Band mean For the image to be corrected Band variance For reference image Band mean For reference image Band variance; t represents land surface type. Represents the image to be corrected. Representative reference image.

2. The method according to claim 1, characterized in that, The formula used in the step of converting DN values ​​to apparent radiance is: L λ =DN λ ×a λ +b λ ; Among them, L λ The apparent radiance of the band is expressed in W·m⁻²·sr⁻¹·μm⁻¹, a λ and b λ These are the band calibration coefficients, gain, and offset, respectively, where λ represents the band.

3. The method according to claim 2, characterized in that, The formula used in the step of converting apparent radiance to apparent reflectance is: ； in, denoted as , where d is the Earth-Sun distance, and ESUN is the solar irradiance at the top of the atmosphere. This is the solar zenith angle.

4. The method according to claim 3, characterized in that, The remote sensing image is multispectral remote sensing data from the Gaofen-7 satellite, including observation data from three channels: 460nm, 550nm, and 670nm.

5. The method according to claim 4, characterized in that, The atmospheric correction is based on the 6SV radiative transfer model.

Images