A method for correcting observation angle effect of high-resolution satellite large field angle remote sensing data

Abstract

The embodiment of the application provides a kind of observation angle effect correction method of high-resolution satellite No.1 large field angle remote sensing data, comprising: obtaining multiple observation angle reflectivity remote sensing data that four wide field sensors (WFV) of high-resolution satellite No.1 accumulates in a certain period;First, remote sensing data pixels are grouped according to land cover type (farmland, grassland and forest), NDVI and observation angle, determine multiple groups of BRDF parameters;In view of the BRDF parameter difference of three different vegetation surface cover types is smaller under the same NDVI value, further merge three land cover type pixels, determine multiple groups of BRDF parameters based on NDVI classification;Finally, according to the BRDF parameter of NDVI classification, the observation angle effect of remote sensing data to be processed is corrected.The application can better correct the angle effect caused by large angle observation in vegetation cover area such as farmland and grassland, and improve the vegetation parameter inversion accuracy of high-resolution satellite No.1 remote sensing data.

Description

Technical Field

[0001] This invention relates to the field of remote sensing image processing technology, and in particular to a method for correcting the observation angle effect of large field-of-view remote sensing data from the Gaofen-1 satellite. Background Technology

[0002] The Gaofen-1 (GF-1) satellite carries four 16m spatial resolution wide field-of-view (WFV) sensors, providing Earth observation data with a 4-day revisit period and a combined coverage width greater than 800km. This provides a valuable data source for high spatiotemporal resolution and large-area surface monitoring and ecosystem services. Vegetation is a crucial component of terrestrial ecosystems, playing a key role in the cycling of Earth's elements such as energy, carbon, and water. Currently, the spatial resolution of vegetation parameter products at global or large-area scales is only at the kilometer or hundred-meter level, which is insufficient for high-precision agricultural, forestry, and ecological environment monitoring, as well as global climate change research and applications. Therefore, high spatiotemporal resolution vegetation parameter products have become an urgent need for research in precision agriculture, high-quality ecological assessment, and other related fields, and GF-1 WFV data offers a potential solution.

[0003] However, due to the anisotropy of surface reflectance, the wide viewing angle of 0-48° in GF-1WFV data leads to the reflectance of most images far from the nadir being affected by the observation angle effect, thus impacting the accuracy of vegetation parameter inversion. Currently, MODIS 500m BRDF parameter products are the most commonly used BRDF parameter data, but directly using them to correct the observation angle effect of high spatial resolution data results in low correction accuracy. Therefore, some methods utilize MODIS pure pixels to extract BRDF parameters for different land cover types or crop types to correct the angle effect of higher resolution images, but this requires combining land cover type data or crop classification data, which is often difficult to obtain. Other methods use fixed BRDF parameters developed based on MODIS BRDF parameters to correct the angle effect of remote sensing data such as Landsat and Sentinel-2, but this method is suitable for situations with small observation angles.

[0004] Therefore, there is an urgent need to propose a set of BRDF model parameters applicable to medium and high spatial resolution and various vegetation conditions to correct the observation angle effect in GF-1WFV reflectance data. Summary of the Invention

[0005] To address the problems in the existing technology, this invention provides a method for correcting the observation angle effect of large field-of-view remote sensing data from the Gaofen-1 satellite.

[0006] Specifically, the embodiments of the present invention provide the following technical solutions:

[0007] In a first aspect, embodiments of the present invention provide a method for correcting the observation angle effect of large field-of-view remote sensing data from the Gaofen-1 satellite, comprising:

[0008] Acquire remote sensing data from multiple satellite sensors;

[0009] Pixels in remote sensing data are grouped according to three land cover types, Normalized Difference Vegetation Index (NDVI), and observation angle to determine multiple sets of BRDF parameters based on land cover type and NDVI classification; the three land cover types include farmland, grassland, and forest.

[0010] By merging three land cover types—farmland, grassland, and forest—and grouping pixels in remote sensing data according to NDVI and observation angle, multiple sets of BRDF parameters based on NDVI classification are determined.

[0011] The observation angle effect of the remote sensing data to be processed is corrected based on the multiple sets of BRDF parameters.

[0012] Furthermore, the process of grouping pixels in remote sensing data based on three land cover types, Normalized Difference Vegetation Index (NDVI), and observation angle to determine multiple sets of BRDF parameters based on land cover type and NDVI classification includes:

[0013] Based on the three land cover types, NDVI, and observation angle, the pixels in the remote sensing data are grouped, and the average reflectance and average observation angle of each group of pixels are calculated.

[0014] Based on the average reflectance, average observation angle, and RTLSR-Chen BRDF model of each group of pixels, multiple sets of BRDF parameters based on land cover type and NDVI classification are determined using the least squares method.

[0015] Furthermore, the process of merging farmland, grassland, and forest land cover types, grouping pixels in the remote sensing data according to NDVI and observation angle, and determining multiple sets of BRDF parameters based on NDVI classification includes:

[0016] The pixels of three land cover types—farmland, grassland, and forest—are merged. Based on NDVI and observation angle, the pixels in the remote sensing data are grouped, and the average reflectance and average observation angle of each group are calculated.

[0017] Based on the average reflectance, average observation angle, and RTLSR-Chen BRDF model of each group of pixels, multiple BRDF parameters based on NDVI classification are determined using the least squares method.

[0018] Furthermore, the step of correcting the observation angle effect of the remote sensing data to be processed based on the multiple sets of BRDF parameters includes:

[0019] Based on the multiple sets of BRDF parameters, the original reflectance in the remote sensing data to be processed is corrected to the nadir observation reflectance in order to correct the observation angle effect of the remote sensing data to be processed.

[0020] Further, the step of correcting the original reflectance in the remote sensing data to be processed to the nadir observation reflectance based on the multiple sets of BRDF parameters, in order to correct the observation angle effect of the remote sensing data to be processed, includes:

[0021] NDVI is calculated based on the remote sensing data to be processed. The BRDF parameters corresponding to the angle correction are determined based on the NDVI range. The reflectance correction coefficient is calculated using the C-factor correction method.

[0022] Based on the reflectance correction coefficient, the original reflectance in the remote sensing data to be processed is corrected to the nadir observation reflectance in order to correct the observation angle effect of the remote sensing data to be processed.

[0023] Secondly, embodiments of the present invention also provide an observation angle effect correction device for large field-of-view remote sensing data from the Gaofen-1 satellite, comprising:

[0024] The acquisition module is used to acquire remote sensing data from multiple satellite sensors;

[0025] The determination module is used to group pixels in remote sensing data according to three land cover types, Normalized Difference Vegetation Index (NDVI), and observation angle, and determine multiple sets of BRDF parameters based on land cover type and NDVI classification; the three land cover types include farmland, grassland, and forest.

[0026] The merging module is used to merge three land cover types: farmland, grassland, and forest. It groups pixels in remote sensing data according to NDVI and observation angle, and determines multiple sets of BRDF parameters based on NDVI classification.

[0027] The correction module is used to correct the observation angle effect of the remote sensing data to be processed based on the multiple sets of BRDF parameters.

[0028] Thirdly, embodiments of the present invention also provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the observation angle effect correction method for the large field-of-view remote sensing data of the Gaofen-1 satellite as described in the first aspect.

[0029] Fourthly, embodiments of the present invention also provide a non-transitory computer-readable storage medium storing a computer program thereon, wherein the computer program, when executed by a processor, implements the steps of the observation angle effect correction method for large field-of-view remote sensing data of the Gaofen-1 satellite as described in the first aspect.

[0030] Fifthly, embodiments of the present invention also provide a computer program product, including a computer program that, when executed by a processor, implements the steps of the observation angle effect correction method for large field-of-view remote sensing data of the Gaofen-1 satellite as described in the first aspect.

[0031] The observation angle effect correction method and device for large field-of-view remote sensing data from the Gaofen-1 satellite provided in this invention fully utilizes the multi-angle observation information accumulated by GF-1 WFV image data at different times, and characterizes vegetation status through NDVI grading. This allows for the extraction of BRDF information under different vegetation conditions, and the use of the RTLSR-Chen BRD model to more accurately simulate hotspot effects, ensuring accurate calculation of BRDF parameters. It boasts advantages such as strong universality, high efficiency, high precision, and ease of operation. Compared to traditional 500-meter MODIS BRDF parameters, it has higher precision; compared to surface classification-based methods, it is easier to operate; and it can effectively correct the angle effect caused by large-angle observations in vegetated areas such as farmland and grassland, improving the accuracy of vegetation parameter inversion from GF-1 remote sensing data. Attached Figure Description

[0032] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0033] Figure 1 This is a flowchart illustrating the method for correcting the observation angle effect of large field-of-view remote sensing data from the Gaofen-1 satellite, as provided in this embodiment of the invention.

[0034] Figure 2 This is another flowchart illustrating the method for correcting the observation angle effect of large field-of-view remote sensing data from the Gaofen-1 satellite provided in this embodiment of the invention.

[0035] Figure 3 This is a schematic diagram of the observation angle effect correction device for large field-of-view remote sensing data of Gaofen-1 satellite provided in an embodiment of the present invention;

[0036] Figure 4 This is a schematic diagram of the structure of the electronic device provided in an embodiment of the present invention. Detailed Implementation

[0037] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0038] The method of this invention can be applied to remote sensing image processing scenarios, and can effectively correct the angle effect caused by large-angle observation in vegetated areas such as farmland and grassland, thereby improving the accuracy of vegetation parameter inversion from GF-1 remote sensing data.

[0039] The following is combined with Figures 1-4 The technical solution of the present invention will be described in detail with reference to specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0040] Figure 1 This is a flowchart illustrating an embodiment of the observation angle effect correction method for large field-of-view remote sensing data from the Gaofen-1 satellite provided in this invention. Figure 1 As shown, the method provided in this embodiment includes:

[0041] Step 101: Acquire remote sensing data from multiple satellite sensors.

[0042] Specifically, in the existing technology, directly correcting the observation angle effect of high spatial resolution data based on MODIS 500m BRDF parameters has the problem of low correction accuracy.

[0043] To address the aforementioned issues, this embodiment first acquires remote sensing data from multiple satellite sensors. Optionally, the acquired remote sensing data can be data collected by the four 16m spatial resolution wide field-of-view (WFV) sensors carried by the Gaofen-1 (GF-1) satellite, or other types of remote sensing data; this embodiment does not impose any limitations.

[0044] For example, the remote sensing data used in this application embodiment includes GF-1WFV reflectance data and GlobeLand30 land cover products. The GF-1WFV reflectance data was downloaded from the China Resources Satellite Data and Application Center, comprising 48 high-quality GF-1WFV images of Northeast China and the North China Plain from April to September 2017. The GF-1WFV reflectance data includes four spectral bands, with a spatial resolution of 16m and a temporal resolution of 4 days. The main bands used in this application embodiment are green, red, and near-infrared. Preprocessing of the GF-1WFV data includes radiometric calibration, atmospheric correction, and geometric correction.

[0045] GlobeLand30 is a global land cover classification dataset with a spatial resolution of 30 meters, providing information on the Earth's land surface. Developed by the National Information Center of China (NGCC), it consists of three products: GlobeLand30 2000, GlobeLand30 2010, and GlobeLand30 2020. This dataset includes 10 land cover categories: farmland, forest, grassland, shrubland, wetland, water bodies, permafrost, artificial surfaces, wasteland, glaciers, and permanent snow cover. This application example collects 15 tile land cover data from the GlobeLand30 2020 version, covering Northeast China and the North China Plain. To ensure consistency, the GlobeLand30 data was resampled from 30-meter resolution to 16-meter to match the resolution of the WFV data.

[0046] This application embodiment uses the USGS spectral library to calculate spectral matching factors to address potential reflectance differences caused by variations in the spectral responses of different WFV sensors. This spectral library, sourced from the USGS Spectroscopy Laboratory website, covers over 1300 spectra from 380 nm to 2500 nm wavelengths. The library contains spectral samples from various surface features, such as minerals, vegetation, and man-made materials. Notably, most remote sensing spectral features have corresponding hyperspectral reflectance spectral features in the library. This application embodiment uses vegetation spectra from the USGS spectral library to calculate the spectral band matching factors for the four WFV sensors of GF-1.

[0047] Step 102: Group the pixels in the remote sensing data according to the three land cover types, the Normalized Difference Vegetation Index (NDVI), and the observation angle, and determine multiple sets of BRDF parameters based on land cover type and NDVI classification; the three land cover types include farmland, grassland, and forest.

[0048] Specifically, after acquiring remote sensing data from multiple satellite sensors, this embodiment of the application groups the pixels in the remote sensing data according to land cover type, NDVI, and observation angle, thus obtaining multiple groups of pixels. For example, grouping pixels in the remote sensing data based on land cover type, NDVI, and observation angle means grouping pixels with the same land cover type, the same NDVI, and the same observation angle into one group. Optionally, the land cover types include 10 types such as farmland, forest, grassland, shrubs, wetlands, water bodies, permafrost, artificial surfaces, wasteland, glaciers, and permanent snow cover. Data points for each land cover type can be classified into a data warehouse using specific NDVI intervals and angular intervals, thus achieving grouping and obtaining multiple groups of pixels. Optionally, the NDVI interval can be 0.1, and the angular intervals for VZA (observed zenith angle), SZA (solar zenith angle), and RAA (relative azimuth angle) are 1°, 2°, and 5°, respectively. Alternatively, groups can be formed based on different land cover types, NDVI intervals, and angular intervals, depending on actual needs. This application does not impose any restrictions on this.

[0049] Step 103: Merge the three land cover types of farmland, grassland and forest, group the pixels in the remote sensing data according to NDVI and observation angle, and determine multiple sets of BRDF parameters based on NDVI classification.

[0050] Specifically, after grouping the pixels in the remote sensing data according to three land cover types, Normalized Difference Vegetation Index (NDVI), and observation angle, multiple sets of BRDF parameters based on land cover type and NDVI classification are determined. That is, after grouping pixels with the same land cover type, the same NDVI, and the same observation angle, the embodiments of this application merge the three land cover types of farmland, grassland, and forest, and group the pixels in the remote sensing data according to NDVI and observation angle to determine multiple sets of BRDF parameters based on NDVI classification.

[0051] Optionally, after merging the three land cover types of farmland, grassland, and forest, and grouping the pixels in the remote sensing data according to NDVI and observation angle, and determining multiple sets of BRDF parameters based on NDVI classification, the reflectance of each set of pixels in this embodiment is analyzed. This yields the reflectance corresponding to each observation angle, and thus the BRDF parameters corresponding to each observation angle. Optionally, BRDF (Bidirectional Reflectance Distribution Function) is a function used to describe the reflection characteristics of light on the surface of an object.

[0052] Step 104: Correct the observation angle effect of the remote sensing data to be processed based on multiple sets of BRDF parameters.

[0053] Specifically, after grouping the pixels in the remote sensing data according to land cover type, NDVI and observation angle, and determining multiple sets of BRDF parameters, this embodiment of the application further corrects the observation angle effect of the remote sensing data to be processed based on the multiple sets of BRDF parameters, thereby improving the accuracy of vegetation parameter inversion of GF-1 remote sensing data.

[0054] The method described above fully utilizes the multi-angle observation information accumulated by GF-1 WFV image data at different times, and characterizes vegetation status through NDVI grading. This enables the extraction of BRDF information under different vegetation conditions. The RTLSR-Chen BRD model is used to more accurately simulate the hotspot effect, ensuring the accurate calculation of BRDF parameters. It has advantages such as strong universality, high efficiency, high precision, and ease of operation. Compared with the traditional 500-meter MODIS BRDF parameters, it has higher precision. Compared with the land surface classification method, it is easier to operate and can better correct the angle effect caused by large-angle observation in vegetated areas such as farmland and grassland, thereby improving the accuracy of vegetation parameter inversion from GF-1 remote sensing data.

[0055] In one embodiment, the reflectivity values ​​of multiple satellite sensors are normalized to the reflectivity value of a target sensor; the target sensor is any one of the multiple satellite sensors.

[0056] Specifically, the different spectral response functions of multiple sensors on a high-resolution satellite lead to variations in reflectance. In other words, for the same surface, sensors at different observation angles will observe different reflectance values. This application addresses this by normalizing the reflectance values ​​of multiple satellite sensors to the reflectance value of the target sensor. This effectively eliminates the differences between sensors at different observation angles, thereby significantly improving the accuracy of the BRDF (Browser Response Function) and consequently, the accuracy of the observation angle effect correction results.

[0057] For example, in this embodiment of the application, the four 16m spatial resolution wide field of view (WFV) sensors carried by the Gaofen-1 (GF-1) satellite exhibit differences in spectral response functions, leading to differences in reflectance. This application normalizes the reflectance values ​​of each sensor into a single, universal sensor, ensuring comparability and consistency among the reflectance values ​​of the four GF-1 WFV sensors. In this application, the target sensor is the second of the four sensors, but any sensor can also be used as the target sensor; then, adjustments are made to address the differences in reflectance across spectral bands. Specifically, the reflectance values ​​of the four WFV sensors are normalized to the reflectance values ​​of the WFV2 sensor. The spectral adjustment factor QE is determined based on vegetation hyperspectral data from the USGS spectral library, and the calculation method is as follows:

[0058] ρ′ WFV,i =ρ WFV,iQE i

[0059] For a given sensor i, the reflectivity ρ WFV,i Defined as:

[0060]

[0061] S i (λ) represents the spectral response function of the WFV,i sensor in the λ band, where [a,b] is the wavelength range and f is the continuous external atmospheric solar irradiance. The only unknown coefficient is ρ. i This is the incident continuous spectral reflectance. Assuming the incident spectral reflectances of the two sensors are identical, a spectral adjustment factor QE can be defined. i for:

[0062]

[0063] Considering the different spectral responses, the average QE is used. i The values ​​are used to normalize the differences in spectral bands. The QE values ​​for the four bands of the WFV1, WFV3, and WFV4 sensors are shown in the table below:

[0064]

[0065] The method described in the above embodiments takes into account that for the same surface, the reflectivity observed by sensors at multiple observation angles is different. Therefore, by normalizing the reflectivity values ​​of multiple satellite sensors to the reflectivity value of the target sensor, the differences between sensors at different observation angles can be effectively eliminated, thereby effectively improving the accuracy of BRDF and thus improving the accuracy of the observation angle effect correction results.

[0066] In one embodiment, pixels in remote sensing data are grouped according to three land cover types, Normalized Difference Vegetation Index (NDVI), and observation angle to determine multiple sets of BRDF parameters based on land cover type and NDVI classification, including:

[0067] Based on three land cover types, NDVI, and observation angle, the pixels in the remote sensing data are grouped, and the average reflectance and average observation angle of each group of pixels are calculated.

[0068] Based on the average reflectance, average observation angle, and RTLSR-Chen BRDF model of each group of pixels, multiple sets of BRDF parameters based on land cover type and NDVI classification are determined using the least squares method.

[0069] Specifically, in this embodiment, pixels in the remote sensing data can be grouped according to land cover type, NDVI, and observation angle. This fully utilizes multi-angle observation information accumulated at different times from the GF-1WFV high-resolution satellite imagery, characterizes vegetation status through NDVI grading, and considers land cover type, achieving more refined grouping. This allows for the calculation of BRDF parameters for different land cover types under different NDVI values, effectively improving the accuracy of BRDF parameters and the accuracy and precision of observation angle effect correction. Optionally, the land cover type data is 30-meter GlobalLand30 land cover classification data. Optionally, the land cover types include 10 land cover types: farmland, forest, grassland, shrubland, wetland, water body, permafrost, artificial surface, wasteland, glacier, and permanent snow cover.

[0070] Optionally, after considering land cover type, NDVI, and observation angle to group pixels in remote sensing data and obtain multiple groups of pixels, the reflectance of each group of pixels in this embodiment is analyzed, the average reflectance and average observation angle of each group of pixels are calculated, and the average reflectance and average observation angle of each group of pixels are fitted based on the least squares method and the RTLSR-Chen BRDF model to obtain the BRDF parameters corresponding to each group of observation angles. This can more effectively correct the observation angle effect of GF-1WFV data and improve the inversion accuracy of vegetation parameters.

[0071] For example, this application considers three land cover types—farmland, grassland, and forest—and only considers cases where the NDVI value is greater than 0.1. The pixel grouping process consists of three steps. First, cloud-free pixels for the three land cover types are extracted from WFV imagery; then, based on the NDVI range and observation angle, the land cover grouped pixels are further classified into smaller data bins. It is worth noting that directly using data with uneven distributions of VZA, SZA, and RAA may lead to fitting results biased towards the most common observation geometry. Therefore, this may limit the impact of equally important but less frequent data points on the estimation of BRDF model parameters. To obtain general fitting results, data points for each land cover category are classified into data bins using specific NDVI and angular intervals. During the fitting process, only one value is considered for each data bin. In this application, the NDVI interval is defined as 0.1, and the VZA (observed zenith angle), SZA (solar zenith angle), and RAA (relative azimuth angle) intervals are 1°, 2°, and 5°, respectively. Additionally, due to the low number of pixels in the NDVI 0.9–1.0 range, the NDVI 0.9–1.0 and NDVI 0.8–0.9 ranges were merged. Statistical preprocessing was performed on the data collected from each data warehouse to remove outliers. For each data warehouse, the data was sorted, and the lowest and highest 5% of data were removed from the sample before calculating the average. This statistical process generates an average for each data warehouse, representing the reflectance and observation geometry of the observed dataset.

[0072] The RTLSR-Chen BRDF model is a linear kernel-driven model that includes three fundamental scattering components: isotropic scattering, volumetric scattering, and geometric optics scattering. The general RTLSR equations are:

[0073] ρ λ (θ s ,θ υ ,φ)=f is (λ)+f vol (λ)K vol (θ s ,θ υ ,φ)+f geo (λ)K geo (θ s ,θ υ ,φ)

[0074] Where θs, θv, and φ are the solar zenith (SZA), field zenith (VZA), and relative azimuth (RAA), respectively, and ρ λ f is the reflectivity of the λ band. iso K is a constant representing isotropic reflectivity. vol (θ s ,θ υ ,φ) and Kgeo (θ s ,θ υ ,φ) represent the volume scattering kernel and the geometric scattering kernel, respectively, f vol and f geo These represent the weights of the two kernels. This application uses the Ross-Thick Chen kernel, which considers hotspot variations, to calculate volume scattering. The calculation formula is as follows:

[0075]

[0076] cosξ=cosθ v cosθ s +sinθ v sinθ s cosφ

[0077] Where ξ is the phase angle, This is the corrected hotspot function. Two free parameters, C1 and C2, allow for a large dynamic range of hotspot variations, facilitating the analysis of changes in hotspot height and width during BRF fitting. The C1 and C2 values ​​are the same for the blue, green, and red bands, set to 0.7 and 5.2 respectively, while the C1 and C2 values ​​for the near-infrared band are set to 0.5 and 4.5 respectively.

[0078] Calculate f for each group using the least squares method. iso f vo; and f geo Parameters. First, determine K based on the observed geometry. vol and K geo The value is then determined. Next, a combined equation is derived from the reflectance of different observation geometries. Finally, the least squares method is used to estimate f. iso f vol and F geo The value of .

[0079]

[0080]

[0081] ...

[0083]

[0084]

[0085] Among them, ρ1, ρ2, ρ3,…,ρ n This indicates the reflectance of different pixels within a certain group.

[0086] It should be noted that during the inversion process, f vol and fgeo The solutions can occasionally be negative, especially when these parameters are very small. This can be attributed to insufficient data or angle samples, or, more reasonably, to a lack of orthogonality between model kernels. Therefore, in the computation, when a parameter has a negative solution, it is set to zero, and the fitting process for the other parameters is repeated.

[0087] The method described above fully utilizes the multi-angle observation information accumulated at different times by the GF-1WFV image data of the high-resolution satellite. It characterizes vegetation status through NDVI grading and takes into account land cover type, achieving more refined grouping. This allows for the calculation of BRDF parameters for different land cover types under different NDVI values, thereby effectively improving the accuracy of BRDF parameters and enhancing the accuracy and precision of observation angle effect correction.

[0088] In one embodiment, three land cover types—farmland, grassland, and forest—are combined, and pixels in the remote sensing data are grouped according to NDVI and observation angle to determine multiple sets of BRDF parameters based on NDVI classification, including:

[0089] The pixels of three land cover types—farmland, grassland, and forest—are merged. Based on NDVI and observation angle, the pixels in the remote sensing data are grouped, and the average reflectance and average observation angle of each group are calculated.

[0090] Based on the average reflectance, average observation angle, and RTLSR-Chen BRDF model of each group of pixels, multiple BRDF parameters based on NDVI classification are determined using the least squares method.

[0091] Specifically, in the process of pixel grouping, pixels in remote sensing data can be grouped based solely on NDVI and observation angle. This means merging vegetation cover types such as farmland, grassland, and forest to calculate BRDF parameters under different NDVIs, which can greatly improve the calculation efficiency of BRDF parameters and thus enable more efficient correction of observation angle effects.

[0092] The method described in the above embodiments combines vegetation cover types such as farmland, grassland and forest to calculate BRDF parameters under different NDVI, thereby greatly improving the calculation efficiency of BRDF parameters and thus enabling more efficient correction of observation angle effects.

[0093] In one embodiment, the observation angle effect of the remote sensing data to be processed is corrected based on multiple sets of BRDF parameters, including:

[0094] Based on multiple sets of BRDF parameters, the original reflectance in the remote sensing data to be processed is corrected to the nadir observation reflectance in order to correct the observation angle effect of the remote sensing data to be processed.

[0095] Specifically, this embodiment fully utilizes multi-angle observation information accumulated from GF-1WFV image data at different times, and characterizes vegetation status through NDVI grading. This allows for the extraction of BRDF information under different vegetation conditions. The RTLSR-Chen BRDF model can more accurately simulate hotspot effects, ensuring accurate calculation of BRDF parameters. It also boasts advantages such as strong universality, high efficiency, high precision, and ease of operation, offering higher accuracy compared to traditional 500-meter MODIS BRDF parameters. Optionally, after determining the BRDF parameters, the observation angle effect of the remote sensing data to be processed can be corrected based on the BRDF parameters. Optionally, the land cover type, NDVI, and observation angle corresponding to the remote sensing data to be processed can be determined first, and then the corresponding grouped BRDF parameters can be selected for correction of the observation angle effect.

[0096] Optionally, in this embodiment, the NDVI can be calculated based on the remote sensing data to be processed, the BRDF parameters corresponding to the angle correction can be determined based on the NDVI range, and the reflectance correction coefficient can be calculated using the C-factor correction method. Based on the reflectance correction coefficient, the original reflectance in the remote sensing data to be processed is corrected to the nadir observation reflectance to correct the observation angle effect of the remote sensing data to be processed. That is, the BRDF parameters are determined based on the NDVI value of the remote sensing image, and the corresponding reflectance correction coefficient is calculated. Then, based on the reflectance correction coefficient, the original reflectance in the remote sensing data to be processed is corrected to the nadir observation reflectance to achieve the correction of the observation angle effect of the remote sensing data to be processed.

[0097] For example, the non-nadir reflectance of GF-1WFV is normalized to the nadir reflectance (θ). v =0°). The calculation formula is as follows:

[0098]

[0099] ρ λ (θ s ,0,φ)=c λ ×ρ λ (θ s ,θ υ ,φ)

[0100] Among them, C λ ρ is the reflectivity correction factor for the λ-band. λ (θs,0,φ) represents the lowest point reflectivity after BRDF correction in the λ-band, ρ λ (θs ,θ υ ,φ) is the actual directional reflectivity of the λ band.

[0101] The method described above first calculates the corresponding reflectance correction coefficient based on the NDVI value and BRDF parameters of the remote sensing image; then, based on the reflectance correction coefficient, the original reflectance in the remote sensing data to be processed is corrected to the nadir observation reflectance, thereby accurately and efficiently correcting the observation angle effect of the remote sensing data to be processed.

[0102] For example, the specific process of the observation angle effect correction method for large field-of-view remote sensing data of Gaofen-1 satellite in this application embodiment is as follows: Figure 2 As shown, the bidirectional reflectance distribution function (BRDF) parameters of vegetation surface are extracted from data from multiple observation angles of GF-1. The normalized difference vegetation index (NDVI) is used to indicate vegetation status characteristics, and NDVI is graded to extract BRDF parameter information with higher spatial resolution. The extracted NDVI-based BRDF parameter set can effectively correct the observation angle effect caused by large-angle observations of GF-1, and can effectively improve the accuracy of vegetation parameter inversion from GF-1 data.

[0103] For example, the correction effect and accuracy evaluation of the observation angle effect correction method based on the large field-of-view remote sensing data of the Gaofen-1 satellite in this application are as follows:

[0104] A total of 24 sets of BRDF model parameters were generated based on three land cover types and eight NDVI intervals. It can be seen that the fiso parameters for the three land cover types show similar results at different NDVI levels. Overall, the fiso parameters in the blue, green, and red bands decrease with increasing NDVI, while those in the near-infrared band decrease... iso The parameter increases with increasing NDVI. However, f vol and f geo The relationship between parameters and NDVI is not always so simple. Generally, the f-value of a forest... vol The parameter is the largest, followed by farmland. Within the NDVI range of 0.1–0.2, the f values ​​for the three land cover types are... vol Significant parameter differences were observed, particularly in the red and near-infrared bands, likely due to variations in underlying surface caused by low vegetation cover. Furthermore, when NDVI was in the range of 0.8–1, the f values ​​for the three land cover types... vol The parameters are significantly larger, due to the greater reflection from denser leaves. The f values ​​for the three land cover types... geo The parameters are influenced by vegetation structure (including leaf size, branches and height) and exhibit complex variations.

[0105] Regarding RMSE, it can be observed that farmland has the lowest RMSE overall, followed by grassland, while forest has the highest RMSE. This indicates that farmland has the best accuracy in BRDF parameter estimation, while forest has the worst. Overall, the root mean square error (RMSE) of the three visible light bands decreases with increasing NDVI. Furthermore, the RMSE in the 0.1–0.2 NDVI range is significantly higher than the RMSE in other high NDVI ranges. This can be explained by the fact that when NDVI is very low, vegetation cover may consist of sparse vegetation and bare ground or other land cover, which may increase the complexity of reflectivity characteristics. Retrieving accurate BRDF parameters for such heterogeneous canopies becomes more challenging, leading to higher RMSE. For the near-infrared band, except for the 0.8–0.1 NDVI range where the RMSE is significantly higher than other low NDVI ranges, the 0.1–0.2 RMSE is higher than other high NDVI ranges. When NDVI approaches 1, dense canopies with high LAI tend to exhibit more complex vegetation structures, such as overlapping leaves and multiple canopies. The complexity of these structures introduces greater uncertainty when accurately estimating BRDF parameters, resulting in higher RMSE values.

[0106] Analysis of the BRDF parameter calculation results for farmland, grassland, and forest shows that within the same NDVI range, the f values ​​for farmland, grassland, and forest are relatively similar. vol The parameters are similar, while the f values ​​for the three land cover types are different. vol The parameters are slightly different. On the other hand, f geo The differences in parameters are more significant. BRDF parameters differ somewhat across land cover types, but BRDF parameters based on land classification data depend on accurate land cover data. Practical applications of land cover data with spatial resolutions of 30 meters or higher typically involve temporal resolutions of 5 or 10 years. However, land cover classification using real-time GF-1 imagery presents challenges in terms of operability and accuracy. To simplify and facilitate BRDF correction of GF-1 imagery, farmland, grassland, and forest data were integrated, and a comprehensive set of BRDF parameters was calculated based on NDVI grading. The calculation results show that the integrated BRDF parameters and the individual BRDF parameters for the three land covers exhibit similar variation patterns, with little difference in BRDF parameter values. Table 1 shows the calculated BRDF model parameters and RMSE for farmland; Table 2 shows the calculated BRDF model parameters and RMSE for grassland; Table 3 shows the calculated BRDF model parameters and RMSE for forest; and Table 4 shows the calculated BRDF model parameters and RMSE for the integrated land cover data.

[0107] Table 1

[0108]

[0109]

[0110] Table 2

[0111]

[0112] Table 3

[0113]

[0114] Table 4

[0115]

[0116]

[0117] To evaluate the correction effect of the BRDF model parameters, the original and corrected reflectance of overlapping areas of image pairs from the same or adjacent dates were compared. Reflectance of image pairs from different viewpoints was normalized to the same observation value (θν and φ = 0°, θs = average SZA value of the image pair), and the mean absolute reflectance difference (MAD) and relative absolute percentage reflectance difference (RAPD) before and after correction were calculated for evaluation. To eliminate any 16m land cover and surface condition variations that may occur due to a one-day difference between the two WFV sensors, a filter was applied to the NDVI value.

[0118]

[0119]

[0120]

[0121] To evaluate the reflectance normalization effect for the three land covers, three pairs of images of different primary land covers were used to assess reflectance differences. The mean NDVI values ​​for farmland, grassland, and forest validation areas were 0.76, 0.59, and 0.83, respectively. We assumed that the true reflectance difference of the same surface over the course of day was small and negligible, in order to compare the reflectance differences of the same surface from different viewpoints. Table 5 shows the MAD and RAPD before and after reflectance correction for the three land covers.

[0122] Table 5

[0123]

[0124]

[0125] For the original image pairs, the three land cover types exhibited significant reflectance differences across the three visible light bands, with RAPD values ​​all exceeding 30%. However, the RAPD values ​​in the near-infrared bands showed significant similarity among the three land cover types, approximately 20%. Among the three land cover types, forest showed the greatest reflectance difference, with a RAPD value exceeding 122% in the blue light band and exceeding 50% in both the green and red light bands. Farmland followed, with a RAPD value exceeding 100% in the blue light band and exceeding 40% in both the green and red light bands. In contrast, grassland showed the smallest reflectance difference, with significantly lower RAPD values ​​in all three visible light bands compared to forest and farmland. After BRDF correction, the reflectance differences within the overlapping areas of the image pairs were significantly reduced, especially for farmland and grassland. After correction, the RAPD values ​​for farmland and grassland were below 20% in all three visible light bands and around 7% in the near-infrared band. Conversely, in the three visible light bands, the forest correction results were less effective than those for farmland and grassland, with RAPD values ​​ranging from 28% to 34%. However, the correction effect in the near-infrared band was comparable, with a RAPD value of approximately 10%. It is important to note that the relatively complex topography of forest areas can cause differences in reflectance, which angle correction cannot eliminate. Comparing the reflectance differences across the four bands after correction, it was found that the RAPD values ​​for the same land cover were similar in the three visible light bands. Furthermore, the near-infrared band had the lowest RAPD value, indicating the best overall correction performance in terms of reflectance consistency.

[0126] Overall, the reflectance of the original image is significantly affected by the observation angle effect, with the blue band showing the most pronounced effect, followed by the green and red bands. The near-infrared band is the least affected of the four bands. In terms of land cover type, forests are most affected by the observation angle effect, followed by farmland, while grassland is least affected. Using newly developed BRDF parameters to correct the original reflectance to nadir-observed reflectance effectively corrects for the observation angle effect. The observation angle correction significantly reduces reflectance differences, especially in farmland and grassland. Similar corrective effects were achieved for the reflectance in the three visible bands. The near-infrared band shows the most effective correction results.

[0127] To further evaluate the reflectance normalization effect across different NDVI ranges and bands, reflectance differences (MAD and RAPD) were calculated for assessment. The lowest reflectance of pixels used for evaluation included farmland, grassland, and forest. Table 6 shows the MAD and RAPD of band reflectance under different NDVI ranges.

[0128] Table 6

[0129]

[0130] It can be observed that as the NDVI range increases, the MAD of the blue, green, and red light bands generally decreases, while the MAD of the near-infrared band generally increases. As the vegetation density increases, the reflectance of the blue, green, and red light bands decreases, and the reflectance of the near-infrared band increases. Therefore, the overall MAD trend of the three visible light bands decreases, and the MAD trend of the near-infrared band increases. When the NDVI range is 0.1 - 0.2, the calibration effect of the four bands is relatively poor. At the same time, when the vegetation density is high and NDVI > 0.8, the calibration effect of the four bands is also poor. In the range of very low or very high NDVI, the calibration accuracy is poor, which is consistent with the accuracy result of BRDF parameter calculation. Among the four bands, when NDVI > 0.2, the RAPD value of the near-infrared band is always the lowest, and RAPD < 12%, indicating successful calibration with little difference. When 0.2 < NDVI < 0.8, the RAPD value of the NIR band decreases as the NDVI range increases, indicating that as the vegetation density increases, the reflectance calibration accuracy improves. Generally speaking, the red light band always shows the highest RAPD value, indicating a significant difference in the calibrated reflectance. For the red light band, the calibration accuracy is relatively stable within the NDVI range of 0.2 - 0.7, and the RAPD range is 13% - 16%. While in the range of 0.8 - 1, the RAPD value of NDVI is the highest, at 20.77%, which may be related to the saturation effect of the red light band under high vegetation cover. When NDVI > 0.1, the calibration accuracy of the blue and green light bands at different NDVI levels is relatively stable, and the RAPD values are between 13% - 19%.

[0131] In summary, the MAD and RAPD values provide the reflectance differences after angular calibration at different NDVI levels. For each NDVI range, the MAD and RAPD values of each band are different, indicating the variation of reflectance calibration. Among the four bands, the near-infrared band has the best calibration effect at all NDVI levels. When 0.2 <ndvi>At NDVI > 0.7, the correction effects are similar across the three visible light bands. When NDVI > 0.7, the reflectance correction results in the red light band are more unstable and less accurate compared to the blue and green bands. These observations indicate that reflectance correction accuracy is generally good within the moderate NDVI range.

[0132] The following describes the observation angle effect correction device for the large field-of-view remote sensing data of the Gaofen-1 satellite provided by the present invention. The observation angle effect correction device for the large field-of-view remote sensing data of the Gaofen-1 satellite described below and the observation angle effect correction method for the large field-of-view remote sensing data of the Gaofen-1 satellite described above can be referred to in correspondence with each other.

[0133] Figure 3 This is a schematic diagram of the observation angle effect correction device for large field-of-view remote sensing data from the Gaofen-1 satellite provided by the present invention. The observation angle effect correction device for large field-of-view remote sensing data from the Gaofen-1 satellite provided in this embodiment includes:

[0134] The acquisition module 310 is used to acquire remote sensing data from multiple satellite sensors;

[0135] Module 320 is used to group pixels in remote sensing data according to three land cover types, Normalized Difference Vegetation Index (NDVI), and observation angle, and to determine multiple sets of BRDF parameters based on land cover type and NDVI classification; the three land cover types include farmland, grassland, and forest.

[0136] The merging module 330 is used to merge three land cover types: farmland, grassland, and forest. It groups pixels in remote sensing data according to NDVI and observation angle, and determines multiple sets of BRDF parameters based on NDVI classification.

[0137] The correction module 340 is used to correct the observation angle effect of the remote sensing data to be processed based on multiple sets of BRDF parameters.

[0138] Optionally, the determining module 320 is specifically used to: group the pixels in the remote sensing data according to the three land cover types, NDVI and observation angle, and calculate the average reflectance and average observation angle of each group of pixels after grouping.

[0139] Based on the average reflectance, average observation angle, and RTLSR-Chen BRDF model of each group of pixels, multiple sets of BRDF parameters based on land cover type and NDVI classification are determined using the least squares method.

[0140] Optionally, the merging module 330 is specifically used to: merge pixels of three land cover types, namely farmland, forest and grassland; group the pixels in the remote sensing data according to NDVI and observation angle; and calculate the average reflectance and average observation angle of each group of pixels after grouping.

[0141] Based on the average reflectance, average observation angle, and RTLSR-Chen BRDF model of each group of pixels, multiple BRDF parameters based on NDVI classification are determined using the least squares method.

[0142] Optionally, the correction module 330 is specifically used to: correct the observation angle effect of the remote sensing data to be processed based on multiple sets of BRDF parameters, including:

[0143] Based on multiple sets of BRDF parameters, the original reflectance in the remote sensing data to be processed is corrected to the nadir observation reflectance in order to correct the observation angle effect of the remote sensing data to be processed.

[0144] Optionally, the correction module 330 is specifically used to: calculate NDVI based on the remote sensing data to be processed, determine the BRDF parameters corresponding to the angle correction based on the NDVI range, and calculate the reflectance correction coefficient using the C-factor correction method;

[0145] Based on the reflectance correction coefficient, the original reflectance in the remote sensing data to be processed is corrected to the nadir observation reflectance in order to correct the observation angle effect of the remote sensing data to be processed.

[0146] The apparatus of this invention is used to execute the method in any of the foregoing method embodiments, and its implementation principle and technical effect are similar, so they will not be described again here.

[0147] Figure 4 A schematic diagram of the physical structure of an electronic device is provided. This electronic device may include a processor 410, a communication interface 420, a memory 430, and a communication bus 440. The processor 410, communication interface 420, and memory 430 communicate with each other via the communication bus 440. The processor 410 can call logical instructions in the memory 430 to execute a method for correcting the observation angle effect of large field-of-view remote sensing data from the Gaofen-1 satellite. This method includes: acquiring remote sensing data from multiple satellite sensors; grouping pixels in the remote sensing data according to three land cover types, the Normalized Difference Vegetation Index (NDVI), and the observation angle, and determining multiple sets of BRDF parameters based on land cover type and NDVI classification; the three land cover types include farmland, grassland, and forest; merging the three land cover types (farmland, grassland, and forest), grouping pixels in the remote sensing data according to NDVI and the observation angle, and determining multiple sets of BRDF parameters based on NDVI classification; and correcting the observation angle effect of the remote sensing data to be processed based on the multiple sets of BRDF parameters.

[0148] Furthermore, the logical instructions in the aforementioned memory 430 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0149] On the other hand, the present invention also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium. The computer program includes program instructions, which, when executed by a computer, enable the computer to execute the observation angle effect correction method for the large field-of-view remote sensing data of the Gaofen-1 satellite provided by the above methods. The method includes: acquiring remote sensing data from multiple satellite sensors; grouping pixels in the remote sensing data according to three land cover types, Normalized Difference Vegetation Index (NDVI), and observation angle, and determining multiple sets of BRDF parameters based on land cover type and NDVI classification; the three land cover types include farmland, grassland, and forest; merging the three land cover types of farmland, grassland, and forest, grouping pixels in the remote sensing data according to NDVI and observation angle, and determining multiple sets of BRDF parameters based on NDVI classification; and correcting the observation angle effect of the remote sensing data to be processed according to the multiple sets of BRDF parameters.

[0150] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements the above-mentioned methods for correcting the observation angle effect of the large field-of-view remote sensing data from the Gaofen-1 satellite. The method includes: acquiring remote sensing data from multiple satellite sensors; grouping pixels in the remote sensing data according to three land cover types, Normalized Difference Vegetation Index (NDVI), and observation angle, and determining multiple sets of BRDF parameters based on land cover type and NDVI classification; the three land cover types include farmland, grassland, and forest; merging the three land cover types (farmland, grassland, and forest), grouping pixels in the remote sensing data according to NDVI and observation angle, and determining multiple sets of BRDF parameters based on NDVI classification; and correcting the observation angle effect of the remote sensing data to be processed according to the multiple sets of BRDF parameters.

[0151] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0152] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0153] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.< / ndvi>

Claims

1. A method for correcting the observation angle effect of large field-of-view remote sensing data from the Gaofen-1 satellite, characterized in that, include: Acquire remote sensing data from multiple satellite sensors; Pixels in remote sensing data are grouped according to three land cover types, Normalized Difference Vegetation Index (NDVI), and observation angle to determine multiple sets of BRDF parameters based on land cover type and NDVI classification; the three land cover types include farmland, grassland, and forest. By merging three land cover types—farmland, grassland, and forest—and grouping pixels in remote sensing data according to NDVI and observation angle, multiple sets of BRDF parameters based on NDVI classification are determined. Based on the multiple sets of BRDF parameters, the observation angle effect of the remote sensing data to be processed is corrected; The method involves grouping pixels in remote sensing data based on three land cover types, Normalized Difference Vegetation Index (NDVI), and observation angle to determine multiple sets of BRDF parameters based on land cover type and NDVI classification, including: Based on the three land cover types, NDVI, and observation angle, the pixels in the remote sensing data are grouped, and the average reflectance and average observation angle of each group of pixels are calculated. Based on the average reflectance, average observation angle, and RTLSR-Chen BRDF model of each group of pixels, multiple sets of BRDF parameters based on land cover type and NDVI classification are determined using the least squares method. The method merges three land cover types: farmland, grassland, and forest. Based on NDVI and observation angle, pixels in the remote sensing data are grouped to determine multiple sets of BRDF parameters based on NDVI classification, including: The pixels of three land cover types—farmland, grassland, and forest—are merged. Based on NDVI and observation angle, the pixels in the remote sensing data are grouped, and the average reflectance and average observation angle of each group are calculated. Based on the average reflectance, average observation angle, and RTLSR-Chen BRDF model of each group of pixels, multiple BRDF parameters based on NDVI classification are determined using the least squares method.

2. The method for correcting the observation angle effect of large field-of-view remote sensing data from the Gaofen-1 satellite according to claim 1, characterized in that, The step of correcting the observation angle effect of the remote sensing data to be processed based on the multiple sets of BRDF parameters includes: Based on the multiple sets of BRDF parameters, the original reflectance in the remote sensing data to be processed is corrected to the nadir observation reflectance in order to correct the observation angle effect of the remote sensing data to be processed.

3. The method for correcting the observation angle effect of large field-of-view remote sensing data from the Gaofen-1 satellite according to claim 2, characterized in that, The step of correcting the original reflectance in the remote sensing data to be processed to the nadir observation reflectance based on the multiple sets of BRDF parameters, in order to correct the observation angle effect of the remote sensing data to be processed, includes: NDVI is calculated based on the remote sensing data to be processed. The BRDF parameters corresponding to the angle correction are determined based on the NDVI range. The reflectance correction coefficient is calculated using the C-factor correction method. Based on the reflectance correction coefficient, the original reflectance in the remote sensing data to be processed is corrected to the nadir observation reflectance in order to correct the observation angle effect of the remote sensing data to be processed.

4. A device for correcting the observation angle effect of Gaofen-1 satellite large field-of-view remote sensing data, used to implement the method for correcting the observation angle effect of Gaofen-1 satellite large field-of-view remote sensing data as described in any one of claims 1 to 3, characterized in that, include: The acquisition module is used to acquire remote sensing data from multiple satellite sensors; The determination module is used to group pixels in remote sensing data according to three land cover types, Normalized Difference Vegetation Index (NDVI), and observation angle, and determine multiple sets of BRDF parameters based on land cover type and NDVI classification; the three land cover types include farmland, grassland, and forest. The merging module is used to merge three land cover types: farmland, grassland, and forest. It groups pixels in remote sensing data according to NDVI and observation angle, and determines multiple sets of BRDF parameters based on NDVI classification. The correction module is used to correct the observation angle effect of the remote sensing data to be processed based on the multiple sets of BRDF parameters.

5. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the observation angle effect correction method for the large field-of-view remote sensing data of the Gaofen-1 satellite as described in any one of claims 1 to 3.

6. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the observation angle effect correction method for the large field-of-view remote sensing data of the Gaofen-1 satellite as described in any one of claims 1 to 3.

7. A computer program product having executable instructions stored thereon, characterized in that, When executed by the processor, this instruction causes the processor to implement the steps of the observation angle effect correction method for the large field-of-view remote sensing data of the Gaofen-1 satellite as described in any one of claims 1 to 3.

Images