A Method and System for Reconstructing BRDF Features of Forests from Multi-Angle Optical Satellite Images

Abstract

The present invention discloses a method and system for reconstructing forest BRDF characteristics of multi - angle optical satellite images. The method includes: performing geometric correction, radiometric calibration, and atmospheric correction on satellite images at multiple observation angles of the same area to obtain satellite reflectance images at multiple observation angles; calculating the solar - observation geometric data of slope pixels of satellite reflectance images at each observation angle; constructing a scattering kernel function library; constructing a vegetation morphological structure parameter library; using a semi - empirical kernel - driven bidirectional reflectance model to screen out the optimal combination of scattering kernel functions and the optimal vegetation morphological structure parameters from the scattering kernel function library and the vegetation morphological structure parameter library; and reconstructing the forest BRDF characteristics of satellite images at multiple observation angles based on the optimal combination of scattering kernel functions and the optimal vegetation morphological structure parameters. The present invention can solve the problem that the BRDF model of existing medium - and low - resolution satellite images cannot accurately reconstruct the forest BRDF characteristics of high - spatial - resolution multi - angle optical satellite images.

Description

Technical Field

[0001] The present invention relates to the technical field of remote sensing image processing, and particularly to a method and system for reconstructing forest BRDF characteristics from multi-angle optical satellite images. Background Art

[0002] As one of the most important ecosystems in the terrestrial ecosystem, how to monitor the forest ecosystem dynamically on a large scale is one of the important contents of the United Nations Sustainable Development Goals (SDGs) and the Global Earth Observation System of Systems (GEOSS). The forest canopy has the characteristic of anisotropic scattering, and its reflection depends not only on the observation direction but also on the incident direction. Therefore, the spectral reflectance of the forest canopy obtained by remote sensing observation is not fixed, but changes with the positions of the sun and the sensor, showing the Bidirectional Reflectance Distribution Function (BRDF) effect. The BRDF characteristics of forests can be used as additional information and are of great value for remote sensing applications such as tree species classification, physical and chemical parameters, and structural parameters. With the continuous update and development of global remote sensing technology, multi-angle spaceborne optical satellites have successfully obtained a large amount of multi-angle observation data and formed a series of model methods for characterizing BRDF. Among them, the semi-empirical linear kernel-driven model based on MODIS data uses a linear expression of the sum of the weights of three groups of kernel functions, namely "isotropic reflectance", "volume scattering", and "geometric optical scattering", to reflect the different scattering characteristics of ground objects, and is one of the most widely used methods for BRDF characteristic reconstruction. The isotropic scattering assumes that the leaves are randomly distributed and only considers the single-scattering approximation, reflecting the reflection characteristics of the isotropic Lambertian surface, which can be regarded as the leaf aggregation characteristic (set as a constant term); the geometric optical scattering describes the scattering effect caused by the shape of the tree crown; the volume scattering reflects the diffusion effect caused by the vertical and horizontal distribution of vegetation elements within the tree crown.

[0003] BRDF models based on medium- and low-resolution satellite images such as MODIS (spatial resolution of 500 m) generally adopt a fixed form of kernel function combination: the Ross-Thick kernel reflecting the scattering of dense vegetation bodies and the Li-Sparse-Reciprocal kernel reflecting the geometric optical scattering of sparsely distributed vegetation, namely RTLSR; and it is assumed that fixed canopy shape parameters (h / b = 2, b / r = 1, where h is the distance from the sphere to the ground, b is the vertical radius of the ellipsoid, and r is the horizontal radius of the ellipsoid) are adopted. Although the reconstruction of forest BRDF feature information from medium- and low-spatial-resolution satellite images has formed an operational processing flow and a corresponding BRDF product system, with the successful launch of high-spatial-resolution multi-angle optical remote sensing satellites, the RTLSR model fails to fully apply to the reconstruction of forest BRDF features from high-spatial-resolution multi-angle optical satellite images under complex terrain in forest areas. The reasons are as follows: (1) The pixel observation scenes of high-spatial-resolution multi-angle optical satellite images are more refined, and the corresponding differences in forest structure types are more obvious (the canopy shape is in sparse or dense distribution; the interior of the canopy is sparse or dense), so the combination form of kernel functions for geometric optical scattering (Ross-Thick kernel, Ross-Thin kernel, Ross-Thick-Maignan kernel) and volume scattering (Li-Sparse-Reciprocal, Li-Dense-Reciprocal, Li-Transit-Reciprocal kernel), as well as the canopy morphological structure parameters (h / b, b / r), need to be more finely adjusted dynamically according to the type of ground objects and different structural states of the same type of ground objects (such as forests); (2) Different from the terrain factors ignored in the BRDF feature reconstruction of 500 m spatial resolution satellite images, the terrain (slope, aspect) corresponding to the pixels of high-spatial-resolution multi-angle optical satellite images cannot be ignored in the impact on the accuracy of forest BRDF feature reconstruction. To sum up, how to solve the scattering mechanism differences caused by different observation scenes (scale, structure) in the reconstruction of forest BRDF features from high-spatial-resolution multi-angle optical satellite images, as well as the BRDF changes of forest canopies caused by undulating terrain, is an unsolved problem and difficulty at present. Summary of the Invention

[0004] The purpose of the present invention is to provide a method and system for reconstructing forest BRDF features from multi-angle optical satellite images, so as to solve the problem that the BRDF model of existing medium- and low-resolution satellite images cannot accurately reconstruct the forest BRDF features of high-spatial-resolution multi-angle optical satellite images.

[0005] To achieve the above purpose, the present invention provides the following solutions:

[0006] A method for reconstructing forest BRDF features from multi-angle optical satellite images includes:

[0007] Perform geometric correction, radiometric calibration, and atmospheric correction on satellite images at multiple observation angles of the same area to obtain satellite reflectance images at multiple observation angles; the observation angles include: observation zenith angle, observation azimuth angle, solar zenith angle, and solar azimuth angle;

[0008] Calculate the solar-observation geometric data of slope pixels in the satellite reflectance images at each observation angle; the solar-observation geometric data of slope pixels includes: slope pixel observation zenith angle, slope pixel observation azimuth angle, slope pixel solar zenith angle, and slope pixel solar azimuth angle;

[0009] Construct a scattering kernel function library; the scattering kernel function library includes: volume scattering kernel function library, volume scattering kernel function coefficient library, geometric optical scattering kernel function coefficient library, and geometric optical scattering kernel function coefficient library;

[0010] Construct a vegetation morphological structure parameter library;

[0011] Use the semi-empirical kernel-driven bidirectional reflectance model to select the optimal combination of scattering kernel functions and the optimal vegetation morphological structure parameters from the scattering kernel function library and the vegetation morphological structure parameter library;

[0012] Based on the optimal combination of scattering kernel functions and the optimal vegetation morphological structure parameters, perform forest BRDF feature reconstruction on satellite images at multiple observation angles.

[0013] Optionally, after performing geometric correction, radiometric calibration, and atmospheric correction on satellite images at multiple observation angles of the same area to obtain satellite reflectance images at multiple observation angles, it further includes:

[0014] Perform spatial resampling on the satellite reflectance images at other observation angles according to the spatial resolution of the satellite reflectance image with the maximum observation zenith angle.

[0015] Optionally, calculating the solar-observation geometric data of slope pixels in the satellite reflectance images at each observation angle specifically includes:

[0016] Based on flat terrain, calculate the observation azimuth angle, observation zenith angle, solar azimuth angle, and solar zenith angle of pixels in the satellite reflectance images at each observation angle;

[0017] Based on digital elevation model data, calculate the slope and aspect of pixels in the satellite reflectance images at each observation angle;

[0018] Based on the slope, aspect, observation azimuth angle, and observation zenith angle, calculate the slope pixel observation zenith angle and slope pixel observation azimuth angle;

[0019] Calculate the solar zenith angle and solar azimuth angle of the slope pixel based on the slope, aspect, solar azimuth angle, and solar zenith angle.

[0020] Optionally, the expression of the volume scattering kernel function library is as follows:

[0021] DK vol = [K Ross-Thick K Ross-Thin K Ross-Thick-Maignan T

[0022] The expression of the volume scattering kernel function coefficient library is as follows:

[0023] Df vol = [f Ross-Thick f Ross-Thin f Ross-Thick-Maignan

[0024] The expression of the geometric optical scattering kernel function coefficient library is as follows:

[0025] DK geo = [K Li-Sparse-Reciprocal K Li-Dense-Reciprocal K Li-Tranist-Reciprocal T

[0026] The expression of the geometric optical scattering kernel function coefficient library is as follows:

[0027] Df geo = [f Li-Sparse-Reciprocal f Li-Dense-Reciprocal f Li-Tranist-Reciprocal

[0028] Where DK vol is the volume scattering kernel function library, Df vol is the volume scattering kernel function coefficient library, DK geo is the geometric optical scattering kernel function coefficient library, Df geo is the geometric optical scattering kernel function coefficient library, K Ross-Thick is the Ross-Thick kernel function, K Ross-rhin is the Ross-Thin kernel function, K Ross-Thick-Maignan is the Ross-Thick-Maignan kernel function, f Ross-Thick is the Ross-Thick kernel function coefficient, f Ross-Thin is the Ross-Thin kernel function coefficient, f Ross-Thick-Maignan is the Ross-Thick-Maignan kernel function coefficient, K Li-Sparse-Reciprocal is the Li-Sparse-Reciprocal kernel function, K Li-Dense-Reciprocal ​​​​is the Li-Dense-Reciprocal kernel function, K Li-Tranist-Reciprocal is the Li-Transit-Reciprocal kernel function, f Li-Sparse-Reciprocal is the coefficient of the Li-Sparse-Reciprocal kernel function, f Li-Dense-Reciprocal is the coefficient of the Li-Dense-Reciprocal kernel function, f Li-Tranist-Reciprocal is the coefficient of the Li-Transit-Reciprocal kernel function.

[0029] Optionally, the expression of the vegetation morphological structure parameter library is as follows:

[0030] Dh / b = [1, 2, 3, 4.....n - 2, n - 1, n]

[0031] Db / r = [1, 2, 3, 4......n - 2, n - 1, n]

[0032] where h is the distance from the tree crown to the ground, b is the vertical radius of the tree crown, r is the horizontal radius of the tree crown, Dh / b and Db / r are the corresponding tree crown morphological structure parameter libraries, and n is the maximum tree crown morphological structure parameter.

[0033] Optionally, the expression of the semi-empirical kernel-driven bidirectional reflectance model is as follows:

[0034]

[0035] where ρ is the reflectance of the pixel in the satellite reflectance image at multiple observation angles, θ′ v is the observation zenith angle of the slope pixel, θ′ s is the solar zenith angle of the slope pixel, is the relative azimuth angle between the sun and the line pushbroom sensor, λ is the wavelength, h / b and b / r are the corresponding tree crown morphological structure parameters, f iso is the isotropic scattering kernel coefficient, f vol is the volume scattering kernel coefficient, K vol is the volume scattering kernel, f geo is the geometric optical scattering kernel coefficient, K geo is the geometric optical scattering kernel.

[0036] The present invention also provides a multi-angle optical satellite image forest BRDF feature reconstruction system, including:

[0037] A processing module for performing geometric correction, radiometric calibration, and atmospheric correction on satellite images at multiple observation angles of the same area to obtain satellite reflectance images at multiple observation angles; the observation angles include: observation zenith angle, observation azimuth angle, solar zenith angle, and solar azimuth angle;

[0038] The slope pixel sun-observation geometry data calculation module is used to calculate the slope pixel sun-observation geometry data of the satellite reflectance image at each observation angle; the slope pixel sun-observation geometry data includes: the slope pixel observation zenith angle, the slope pixel observation azimuth angle, the slope pixel sun zenith angle, and the slope pixel sun azimuth angle;

[0039] The scattering kernel function library construction module is used to construct a scattering kernel function library; the scattering kernel function library includes: a volume scattering kernel function library, a volume scattering kernel function coefficient library, a geometric optics scattering kernel function coefficient library, and a geometric optics scattering kernel function coefficient library;

[0040] The vegetation morphological structure parameter library construction module is used to construct a vegetation morphological structure parameter library;

[0041] The screening module is used to screen out the optimal scattering kernel function combination and the optimal vegetation morphological structure parameters from the scattering kernel function library and the vegetation morphological structure parameter library by using the semi-empirical kernel-driven bidirectional reflectance model;

[0042] The reconstruction module is used to reconstruct the forest BRDF characteristics of the satellite image at multiple observation angles based on the optimal scattering kernel function combination and the optimal vegetation morphological structure parameters.

[0043] Optionally, it further includes:

[0044] The spatial resampling module is used to perform spatial resampling on the satellite reflectance images at other observation angles according to the spatial resolution of the satellite reflectance image with the maximum observation zenith angle.

[0045] Optionally, the slope pixel sun-observation geometry data calculation module specifically includes:

[0046] The planar pixel sun-observation geometry information calculation unit is used to calculate the observation azimuth angle, the observation zenith angle, the sun azimuth angle, and the sun zenith angle of the pixels in the satellite reflectance image at each observation angle based on the flat terrain;

[0047] The slope and aspect calculation unit is used to calculate the slope and aspect of the pixels in the satellite reflectance image at each observation angle based on the digital elevation model data;

[0048] The slope pixel observation geometry information calculation unit is used to calculate the slope pixel observation zenith angle and the slope pixel observation azimuth angle based on the slope, the aspect, the observation azimuth angle, and the observation zenith angle;

[0049] The slope pixel sun geometry information calculation unit calculates the slope pixel sun zenith angle and the slope pixel sun azimuth angle based on the slope, the aspect, the sun azimuth angle, and the sun zenith angle.

[0050] According to the specific embodiments provided by the present invention, the following technical effects are disclosed by the present invention:

[0051] The method and system for reconstructing forest BRDF features from multi-angle optical satellite images provided by the present invention, on the one hand, characterize the scattering features of forests with different structures by optimizing and selecting a scattering kernel function library and a vegetation morphological structure parameter library; on the other hand, combine topographic information to accurately characterize the BRDF features of forests under complex terrain conditions, overcoming the shortcomings of the original BRDF model of coarse-resolution optical images being not fine and accurate enough. Brief Description of the Drawings

[0052] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required to be used in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0053] Figure 1 is a flowchart of the method for reconstructing forest BRDF features from multi-angle optical satellite images provided by the present invention;

[0054] Figure 2 is the 5-angle observation image data of the terrestrial ecosystem carbon monitoring satellite simulated by the Less radiative transfer model;

[0055] Figure 3 is a comparison chart of the image data reconstructed by the forest BRDF model based on the present invention and the observed image data;

[0056] Figure 4 is the R of the reflectance reconstructed by the forest BRDF model based on the present invention and the observed reflectance 2 and RMSE and the comparison chart of the R 2 and RMSE of the reflectance reconstructed by the BRDF model based on RTLSR and the observed reflectance;

[0057] Figure 5 is a schematic diagram of the vegetation morphological structure parameters preferentially selected for each pixel of the high-resolution satellite image;

[0058] Figure 6 is a schematic diagram of the geometric optical scattering kernel function and the volume scattering kernel function preferentially selected for each pixel of the high-resolution satellite image;

[0059] Figure 7 is a schematic diagram of the observation images of the hot spots and cold spots directions reconstructed by the forest BRDF model based on the present invention. Detailed Embodiments

[0060] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0061] The object of the present invention is to provide a method and system for reconstructing the BRDF characteristics of forests in multi-angle optical satellite images, so as to solve the problem that the BRDF models of existing medium and low-resolution satellite images cannot accurately reconstruct the BRDF characteristics of forests in high-spatial-resolution multi-angle optical satellite images. To make the above objects, features, and advantages of the present invention more obvious and understandable, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0062] Embodiment 1

[0063] As Figure 1 shown, the method for reconstructing the BRDF characteristics of forests in multi-angle optical satellite images provided by the present invention includes the following steps:

[0064] Step 101: Geometrically correct, radiometrically calibrate, and atmospherically correct satellite images at multiple observation angles in the same area to obtain satellite reflectance images at multiple observation angles; the observation angles include: the observation zenith angle, observation azimuth angle of the sensor when the satellite passes, and the solar zenith angle and solar azimuth angle at the observation time in this area.

[0065] In practical applications, for satellite images (multi-angle satellite images) of the same area containing at least 3 observation angles, geometrically correct, radiometrically calibrate, and atmospherically correct them respectively to obtain satellite reflectance images corresponding to the observation angles. Figure 2 For the satellite image data of 5 observation angles of the terrestrial ecosystem carbon monitoring satellite simulated by the Less radiative transfer model, the spatial resolution is set to 1.5 m.

[0066] Resample the satellite reflectance images of other observation angles according to the spatial resolution of the satellite reflectance image with the maximum observation zenith angle (i.e., the coarsest spatial resolution) so that the satellite reflectance images of all observation angles have a unified spatial resolution, and extract the area where the number of observations of the same pixel is greater than or equal to 3.

[0067] Step 102: Calculate the solar-observation geometric data of the slope pixels in the satellite reflectance images at each observation angle; the solar-observation geometric data of the slope pixels includes: the observation zenith angle of the slope pixel, the observation azimuth angle of the slope pixel, the solar zenith angle of the slope pixel, and the solar azimuth angle of the slope pixel. Specifically:

[0068] (1) Based on flat terrain, calculate the observation azimuth angle, observation zenith angle, solar azimuth angle, and solar zenith angle of the pixels in the satellite reflectance images at each observation angle.

[0069] When calculating the solar-observation geometry of the pixels in the multi-angle images, the frame coordinate system should be converted to the ground coordinate system by combining the observation angle information of the whiskbroom sensor:

[0070]

[0071]

[0072] Since the multi-angle optical satellite images are sensor whiskbroom images, in Equation (1), (x, 0) and (x0, 0) are the coordinates of the scan line pixel point and the scan line center pixel point in the frame coordinate system respectively; f is the focal length, (x - x0, 0, -f) is the three-dimensional coordinate of the pixel point in the image space coordinate system, and (u, v, w) is the three-dimensional coordinate of the pixel point in the ground space coordinate system. In Equation (2), ω and k are the sensor attitude data, that is, the heading tilt angle (generally set to 0), the cross-track tilt angle (the sensor's tilted observation angle), and the image rotation angle (generally set to 0), α az is the heading azimuth angle, R is the total rotation matrix of φ, ω, and k, R φ is the rotation matrix corresponding to the attitude parameter φ, R ω is the rotation matrix corresponding to the attitude parameter ω, R k is the rotation matrix corresponding to the attitude parameter k.

[0073] Then the plane-based observation zenith angle θ of the pixel in the satellite image v and the azimuth angle φ v The calculation formulas are as shown in the following equations:

[0074]

[0075]

[0076] Then the plane-based solar zenith angle θ of the pixel in the satellite image s and the azimuth angle φ s The calculation formulas are as shown in the following equations:

[0077] θ s = π / 2 - asin(sinψsinχ + cosψcosχcosτ) (5)

[0078] φ s = acos ((sinψ - sinχsinθ s ) / cosχcosθ s ) (6)

[0079] In the formula: ψ is the solar declination angle, χ is the local geographical latitude, and τ is the solar hour angle at that time.

[0080] (2) Calculate the slope (α) and aspect (β) information of each pixel in the image within the observation area based on the digital elevation model data (DEM).

[0081] (3) Combine the slope and aspect information to transform the sun-observation geometry based on planar pixels into the local coordinate system of the corresponding slope surface of the pixel. The transformation formula can be expressed as:

[0082]

[0083] Among them, (x', y', z') is the new coordinate transformation in the local coordinate system of the slope surface.

[0084] Then, in the local coordinate system of the slope surface, the zenith angle θ' of the slope pixel observation v and the azimuth angle are respectively calculated by formulas (8) and (9):

[0085]

[0086]

[0087] Similarly, the zenith angle θ' of the sun of the slope pixel s and the azimuth angle can also be calculated by formulas (7), (8), and (9), that is, by replacing the zenith angle θ v , the observation azimuth angle with the zenith angle θ of the sun s , and the sun azimuth angle θ s is sufficient.

[0088] Step 103: Construct a scattering kernel function library; the scattering kernel function library includes: a volume scattering kernel function library, a volume scattering kernel function coefficient library, a geometric optics scattering kernel function coefficient library, and a geometric optics scattering kernel function coefficient library.

[0089] The main object observed in the present invention is the forest, and its BRDF characteristics can be decomposed into: isotropic scattering, volume scattering, and geometric optics scattering. Isotropic scattering represents the isotropic reflection of the Lambertian surface. Assuming the random distribution of leaves, it can be regarded as the overall effect of the aggregation properties of all leaves. The geometric optics scattering effect is related to the shape of the canopy; the volume scattering effect reflects the vertical and horizontal distribution of the internal elements of the canopy and is related to the wavelength. The long wavelength has a stronger scattering effect.

[0090] Usually, the isotropic scattering kernel K is simplified isois a constant 1, but the volume scattering kernel and the geometric optical scattering kernel have different combination forms. The high-spatial-resolution multi-angle optical satellite imagery has a high spatial resolution, making the coarse-resolution observation scenarios of the fixed volume scattering kernel and geometric optical scattering kernel originally applicable to satellite imagery with coarse spatial resolution such as MODIS no longer fully applicable to high-spatial-resolution imagery. Since different land cover types have different BRDF characteristics, even the same land cover type will cause changes in its scattering characteristics due to differences in its morphology and structure. Therefore, the present invention constructs a scattering kernel function library according to the more refined various land cover types at the pixel scale of the high-spatial-resolution multi-angle optical satellite imagery, and different structural states of the same land cover type, and then uses rule definitions to enable the kernel function to adaptively find the optimal combination form of the geometric optical scattering and volume scattering kernel functions of the BRDF model that conforms to different vegetation morphological structures in this high-spatial-resolution observation scenario.

[0091] Common volume scattering kernels include: Ross-Thick kernel, Ross-Thin kernel, and Ross-Thick-Maignan kernel. The Ross-Thick kernel is used to describe the volume scattering effect of dense vegetation (LAI>>1). The model is based on the assumption that the small scattering surfaces in the background and the medium are all Lambertian scattering, and the orientations of the scattering surfaces in the medium are randomly distributed, only considering single scattering and ignoring the influence of multiple scattering. The Ross-Thin kernel is applicable to the volume scattering effect when the leaf area index of the vegetation is small. The Ross-Thick-Maignan kernel takes into account the hot spot effect of the canopy on the basis of the Ross-Thick kernel. Common geometric optical scattering kernels include: Li-Sparse-Reciprocal kernel, Li-Dense-Reciprocal kernel, and Li-Transit-Reciprocal kernel. The Li-Sparse-Reciprocal kernel is the reciprocal form of the Li-Sparse kernel and is applicable to sparse canopies distributed on a Lambertian scattering ground background. The Li-Dense-Reciprocal kernel is applicable to canopies with a large vegetation density. The Li-Transit-Reciprocal kernel has a similar fitting ability to the Li-Sparse-Reciprocal kernel but is proven to perform better at high solar or observation zenith angles.

[0092] Based on the common volume scattering kernel and geometric optical scattering kernel, the present invention first introduces the solar-observation geometric data of the slope pixels calculated in step 102 and constructs its volume scattering kernel function library DK pixel by pixel vol (Equation 10) and the geometric optical scattering kernel function library DK geo (Equation 12) and its corresponding volume scattering kernel function coefficient library Df vol (Equation 11) and the geometric optical scattering kernel function coefficient library Dfgeo (Equation 13).

[0093] Volume scattering kernel function library DK vol :

[0094] DK vol = [K Ross-Th ick K Ress-Th in K Ress-Th ick-Maignan T (10)

[0095] Volume scattering kernel function coefficient library Df vol :

[0096] Df vol = [f Ross-Th ick f Ross-Th in f R。ss-Th ick-Maignan (11)

[0097] Geometric optical scattering kernel function library DK geo :

[0098] DK geo = [K Li-Sparse-Reciprocal K Li-Dense-Reciprocal K Li-Tranist-Reciprocal T (12)

[0099] Geometric optical scattering kernel function coefficient library Df geo :

[0100] Df geo = [f Li-Sparse-Reciprocal f Li-Dense-Reciprocal f Li-Tranist-Reciprocal (13)

[0101] Step 104: Construct the vegetation morphological structure parameter library.

[0102] The h / b and b / r reflecting the vegetation morphological structure are important constituent parameters of the geometric optical scattering kernel function K geo . Different from the usually assumed tree crown shape of h / b = 2 and b / r = 1 for the pixels observed by the satellite images with coarse spatial resolution such as MODIS. In order to further accurately characterize the BRDF characteristics of the pixels of the high-spatial-resolution multi-angle optical satellite images, based on the optimized selection of the adaptive geometric optical scattering kernel function for the BRDF modeling of the ground objects in Step 103, following the fact of the vegetation morphological differences within the observation scene, the vegetation shape parameters h / b and b / r in the pixels observed by the high-spatial-resolution images are respectively taken from 1 to n in sequence, with an interval of 1 (for example, from 1 to 100, with an interval of 1), to form the morphological parameter libraries Dh / b and Db / r.

[0103] ​​Dh / b = [1, 2, 3, 4......n - 2, n - 1, n] (14)

[0104] Db / r = [1, 2, 3, 4......n - 2, n - 1, n] (15)

[0105] Step 105: Using the semi - empirical kernel - driven bidirectional reflectance model, screen out the optimal scattering kernel function combination and the optimal vegetation morphological structure parameters from the scattering kernel function library and the vegetation morphological structure parameter library.

[0106] In order to accurately construct the forest BRDF feature model for each pixel in the observation scene of high - spatial - resolution multi - angle optical satellite images using the semi - empirical linear kernel - driven bidirectional reflectance model (Equation 16), information such as the pixel reflectance of high - spatial - resolution multi - angle optical satellite images with the number of observation angles ≥ 3, the solar - observation geometric data of slope pixels, the optimal combination form of geometric optical scattering kernels and volume scattering kernels under high - spatial - resolution observation scenes, and the optimal vegetation morphological structure parameters is required.

[0107]

[0108] Among them, corresponding to the solar zenith angle θ′ of the slope pixel calculated in Step 102 s and the viewing zenith angle θ′ of the slope v under the condition of the relative azimuth angle between the sun and the sensor at the slope pixel the reflectance of a pixel at a certain wavelength λ under a certain vegetation morphological condition (h / b, b / r). f iso (λ), f vol (λ) and f geo (λ) represent the coefficients corresponding to the isotropic scattering kernel, volume scattering kernel and geometric optical scattering kernel in the BRDF model respectively.

[0109] The setting rule defines that each time only one geometric optical scattering kernel function (Equation 17) and one volume scattering kernel function are selected from the scattering kernel function library in Step 103 for combination (Equation 18), and each time one h / b (Equation 19) and one b / r (Equation 20) vegetation morphological parameter are selected from the vegetation morphological structure parameter library in Step 104 for combination. Finally, based on this double - iteration, using the semi - empirical kernel - driven BRDF model (Equation 16), screen out the best kernel function combination form and the vegetation morphological structure parameter relationship (h / b, b / r) with the minimum RMSE and the maximum R 2 between the model simulation value and the observed value (Equations 21 and 22). The preferred vegetation morphological structure parameters are as Figure 5 shown, and the schematic diagrams of the preferred geometric optical scattering kernel function and volume scattering kernel function are as Figure 6as shown

[0110] s.t. ||Df geo ||0 = 1 (17)

[0111] s.t. ||Df vol ||0 = 1 (18)

[0112] s.t. ||Dh / b||0 = 1 (19)

[0113] s.t. ||Db / r||0 = 1 (20)

[0114]

[0115]

[0116] wherein, the coefficients f iso 、f vol and f geo corresponding to each kernel function term in the BRDF model of formula (16) can be solved by the least squares method (Equation 23).

[0117]

[0118] Solve X·F = Y, then X T ·X·F = X T ·Y, and F = (X T ·X) -1 X T ·Y can be obtained, that is, the coefficients f iso 、f vol and f geo corresponding to each kernel function term in the BRDF model of the reflectance of a certain ground object type in a certain waveband are solved.

[0119] Step 106: Based on the optimal scattering kernel function combination and the optimal vegetation morphological structure parameters, perform satellite image forest BRDF feature reconstruction at multiple observation angles.

[0120] According to the BRDF model (f iso ,f vol ,f geo ,K geo kernel function form, K vol kernel function form, h / b, b / h parameters have been obtained in step five) constructed for each pixel, only by setting any sun-observation geometry (such as cold and hot spot observations) of the slope pixel, the band reflectance information of the pixel under this condition can be obtained, and the expression is as shown in Equation (24):

[0121]

[0122] is the solar zenith angle with an arbitrarily set wavelength λ view zenith angle and the relative azimuth of the sun-sensor The reflectance reconstructed by the BRDF model under the conditions.

[0123] Figure 3 is a comparison chart of the image data reconstructed by the forest BRDF model based on the present invention and the observed image data. It can be seen that the two have the same realistic effect under this solar observation geometry (view zenith = 41.00, view azimuth = 184.0, solar zenith = 42.16, solar azimuth = 185.00). And as Figure 4 shown, through the R of the pixel-by-pixel BRDF reconstructed reflectance and the observed reflectance in the near-infrared band of 837nm 2 and RMSE, it can be seen that: compared with the BRDF model based on RTLSR for satellite data with coarse spatial resolution, the reflectance reconstructed by the BRDF model of the present invention for high-spatial-resolution satellite images and the observed reflectance have a higher R 2 and a lower RMSE. Therefore, the BRDF feature reconstruction method for high-spatial-resolution multi-angle satellite images provided by the present invention is more accurate.

[0124] Figure 7 are the observed images in the hot spot (view zenith = 41.00, view azimuth = 184.0, solar zenith = 41.00, solar azimuth = 184.00) and cold spot (view zenith = 41.00, view azimuth = 4.0, solar zenith = 41.00, solar azimuth = 184.00) directions reconstructed by the forest BRDF model based on the present invention. From Figure 7 it can be clearly seen that the visual effect of the image in the hot spot observation direction is brighter, while the visual effect of the image in the cold spot observation direction is darker.

[0125] Embodiment 2

[0126] In order to execute the method corresponding to the above Embodiment 1 to achieve the corresponding functions and technical effects, a system for reconstructing forest BRDF features of multi-angle optical satellite images is provided below.

[0127] The system includes:

[0128] A processing module for performing geometric correction, radiometric calibration, and atmospheric correction on satellite images at multiple observation angles in the same area to obtain satellite reflectance images at multiple observation angles; the observation angles include: observation zenith angle, observation azimuth angle, solar zenith angle, and solar azimuth angle;

[0129] A slope pixel solar-observation geometry data calculation module for calculating the slope pixel solar-observation geometry data of the satellite reflectance images at each observation angle; the slope pixel solar-observation geometry data includes: slope pixel observation zenith angle, slope pixel observation azimuth angle, slope pixel solar zenith angle, and slope pixel solar azimuth angle;

[0130] A scattering kernel function library construction module for constructing a scattering kernel function library; the scattering kernel function library includes: a volume scattering kernel function library, a volume scattering kernel function coefficient library, a geometric optics scattering kernel function coefficient library, and a geometric optics scattering kernel function coefficient library;

[0131] A vegetation morphological structure parameter library construction module for constructing a vegetation morphological structure parameter library;

[0132] A screening module for using a semi-empirical kernel-driven bidirectional reflectance model to screen out the optimal combination of scattering kernel functions and the optimal vegetation morphological structure parameters from the scattering kernel function library and the vegetation morphological structure parameter library;

[0133] A reconstruction module for reconstructing the forest BRDF characteristics of satellite images at multiple observation angles based on the optimal combination of scattering kernel functions and the optimal vegetation morphological structure parameters.

[0134] It further includes: a spatial resampling module for spatially resampling the satellite reflectance images at other observation angles according to the spatial resolution of the satellite reflectance image at the maximum observation zenith angle.

[0135] Among them, the slope pixel solar-observation geometry data calculation module specifically includes:

[0136] A planar pixel solar-observation geometry information calculation unit for calculating the observation azimuth angle, observation zenith angle, solar azimuth angle, and solar zenith angle of the pixels in the satellite reflectance images at each observation angle based on a flat terrain;

[0137] A slope and aspect calculation unit for calculating the slope and aspect of the pixels in the satellite reflectance images at each observation angle based on digital elevation model data;

[0138] A slope pixel observation geometry information calculation unit for calculating the slope pixel observation zenith angle and slope pixel observation azimuth angle based on the slope, aspect, observation azimuth angle, and observation zenith angle;

[0139] The slope pixel solar geometric information calculation unit calculates the slope pixel solar zenith angle and the slope pixel solar azimuth angle based on the slope, the aspect, the solar azimuth angle, and the solar zenith angle.

[0140] Each embodiment in this specification is described in a progressive manner. The key point of each embodiment is the difference from other embodiments. For the same or similar parts among the embodiments, reference can be made to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple. For the relevant parts, reference can be made to the description in the method part.

[0141] Specific examples are used in this article to elaborate on the principles and implementation manners of the invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. The described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

Claims

1. A method for reconstructing forest BRDF characteristics from multi-angle optical satellite images, characterized in that Including: Performing geometric correction, radiometric calibration, and atmospheric correction on satellite images at multiple observation angles of the same area to obtain satellite reflectance images at multiple observation angles; The observation angles include: observation zenith angle, observation azimuth angle, solar zenith angle, and solar azimuth angle; Calculating the solar-observation geometric data of slope pixels for the satellite reflectance images at each observation angle; the solar-observation geometric data of slope pixels includes: slope pixel observation zenith angle, slope pixel observation azimuth angle, slope pixel solar zenith angle, and slope pixel solar azimuth angle; Constructing a scattering kernel function library; the scattering kernel function library includes: a volume scattering kernel function library, a volume scattering kernel function coefficient library, a geometric optics scattering kernel function library, and a geometric optics scattering kernel function coefficient library; Constructing a vegetation morphological structure parameter library; Using a semi-empirical kernel-driven bidirectional reflectance model to select the optimal combination of scattering kernel functions and the optimal vegetation morphological structure parameters from the scattering kernel function library and the vegetation morphological structure parameter library; Based on the optimal combination of scattering kernel functions and the optimal vegetation morphological structure parameters, reconstructing the forest BRDF characteristics of satellite images at multiple observation angles; The expression of the vegetation morphological structure parameter library is as follows: D D Among them, b is the vertical radius of the tree crown, r is the horizontal radius of the tree crown, D , D are the corresponding tree crown morphological structure parameter libraries, and n is the maximum tree crown morphological structure parameter; The expression of the semi-empirical kernel-driven bidirectional reflectance model is as follows: wherein, is the reflectance of the pixel in the satellite reflectance images at multiple observation angles, is the observation zenith angle of the slope pixel, is the solar zenith angle of the slope pixel, is the relative azimuth angle between the sun and the line pushbroom sensor, is the wavelength, and are the corresponding canopy morphological structure parameters, is the isotropic scattering kernel coefficient, is the volume scattering kernel coefficient, is the volume scattering kernel, is the geometric optical scattering kernel coefficient, is the geometric optical scattering kernel.

2. The method for reconstructing the BRDF characteristics of a forest from multi-angle optical satellite images according to claim 1, wherein After performing geometric correction, radiometric calibration, and atmospheric correction on satellite images at multiple observation angles of the same area to obtain satellite reflectance images at multiple observation angles, it further includes: Performing spatial resampling on the satellite reflectance images at other observation angles according to the spatial resolution of the satellite reflectance image with the maximum observation zenith angle.

3. The multi-angle optical satellite image forest BRDF feature reconstruction method according to claim 1, wherein, Calculating the solar-observation geometric data of slope pixels for the satellite reflectance images at each observation angle, specifically including: Based on flat terrain, calculating the observation azimuth angle, observation zenith angle, solar azimuth angle, and solar zenith angle of pixels in the satellite reflectance images at each observation angle; Calculating the slope and aspect of pixels in the satellite reflectance images at each observation angle based on digital elevation model data; Based on the slope, aspect, observation azimuth angle, and observation zenith angle, calculating the slope pixel observation zenith angle and slope pixel observation azimuth angle; Based on the slope, aspect, solar azimuth angle, and solar zenith angle, calculating the slope pixel solar zenith angle and slope pixel solar azimuth angle.

4. The method for reconstructing the forest BRDF characteristics of multi-angle optical satellite images according to claim 1, wherein The expression of the volume scattering kernel function library is as follows: The expression of the volume scattering kernel function coefficient library is as follows: The expression of the geometric optics scattering kernel function library is as follows: The expression of the geometric optics scattering kernel function coefficient library is as follows: Among them, is the volume scattering kernel function library, is the volume scattering kernel function coefficient library, is the geometric optical scattering kernel function library, is the geometric optical scattering kernel function coefficient library, is the Ross-Thick kernel function, is the Ross-Thin kernel function, is the Ross-Thick-Maignan kernel function, is the Ross-Thick kernel function coefficient, is the Ross-Thin kernel function coefficient, is the Ross-Thick-Maignan kernel function coefficient, is the Li-Sparse-Reciprocal kernel function, is the Li-Dense-Reciprocal kernel function, is the Li-Transit-Reciprocal kernel function, is the Li-Sparse-Reciprocal kernel function coefficient, is the Li-Dense-Reciprocal kernel function coefficient, is the Li-Transit-Reciprocal kernel function coefficient.

5. A multi-angle optical satellite image forest BRDF feature reconstruction system, characterized in that, Including: A processing module for performing geometric correction, radiometric calibration, and atmospheric correction on satellite images at multiple observation angles of the same area to obtain satellite reflectance images at multiple observation angles; the observation angles include: observation zenith angle, observation azimuth angle, solar zenith angle, and solar azimuth angle; Slope pixel solar-observation geometry data calculation module, which is used to calculate the slope pixel solar-observation geometry data of the satellite reflectance image at each observation angle; the slope pixel solar-observation geometry data includes: slope pixel observation zenith angle, slope pixel observation azimuth angle, slope pixel solar zenith angle, and slope pixel solar azimuth angle; Scattering kernel function library construction module, which is used to construct a scattering kernel function library; the scattering kernel function library includes: volume scattering kernel function library, volume scattering kernel function coefficient library, geometric optical scattering kernel function library, and geometric optical scattering kernel function coefficient library; Vegetation morphological structure parameter library construction module, which is used to construct a vegetation morphological structure parameter library; Screening module, which is used to screen out the optimal scattering kernel function combination and the optimal vegetation morphological structure parameters from the scattering kernel function library and the vegetation morphological structure parameter library by using the semi-empirical kernel-driven bidirectional reflectance model; Reconstruction module, which is used to reconstruct the forest BRDF characteristics of the satellite image at multiple observation angles based on the optimal scattering kernel function combination and the optimal vegetation morphological structure parameters; The expression of the vegetation morphological structure parameter library is as follows: D D where b is the vertical radius of the tree crown, r is the horizontal radius of the tree crown, D and D are the corresponding tree crown morphological structure parameter libraries, and n is the maximum tree crown morphological structure parameter; The expression of the semi-empirical kernel-driven bidirectional reflectance model is as follows: Among them, is the reflectance of the pixel in the satellite reflectance images at multiple observation angles, is the observation zenith angle of the slope pixel, is the solar zenith angle of the slope pixel, is the relative azimuth angle between the sun and the line pushbroom sensor, is the wavelength, and are the corresponding crown morphological structure parameters, is the isotropic scattering kernel coefficient, is the volume scattering kernel coefficient, is the volume scattering kernel, is the geometric optical scattering kernel coefficient, is the geometric optical scattering kernel.

6. The multi-angle optical satellite image forest BRDF feature reconstruction system according to claim 5, characterized in that, It also includes: Spatial resampling module, which is used to perform spatial resampling on the satellite reflectance images at other observation angles according to the spatial resolution of the satellite reflectance image with the maximum observation zenith angle.

7. The multi-angle optical satellite image forest BRDF feature reconstruction system according to claim 5, characterized in that The slope pixel solar-observation geometry data calculation module specifically includes: Plane pixel solar-observation geometry information calculation unit, which is used to calculate the observation azimuth angle, observation zenith angle, solar azimuth angle, and solar zenith angle of the pixels in the satellite reflectance image at each observation angle based on flat terrain; Slope and aspect calculation unit, which is used to calculate the slope and aspect of the pixels in the satellite reflectance image at each observation angle based on digital elevation model data; Slope pixel observation geometry information calculation unit, which is used to calculate the slope pixel observation zenith angle and slope pixel observation azimuth angle based on the slope, aspect, observation azimuth angle, and observation zenith angle; Slope pixel solar geometry information calculation unit, which calculates the slope pixel solar zenith angle and slope pixel solar azimuth angle based on the slope, aspect, solar azimuth angle, and solar zenith angle.

Images