Bidirectional reflectivity factor calculation method for forest canopy high-resolution remote sensing image
By considering the influence of adjacent cell radiation at high spatial resolution, combining geometric optical model and spectral invariance theory, a forest canopy remote sensing mechanism model suitable for high spatial resolution is constructed, which solves the vegetation of traditional models under high spatial resolution. The problem of low parameter inversion accuracy achieves higher inversion accuracy and model applicability.
Patent Information
- Application Number
- CN202411013100.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-26
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2044-07-26
AI Technical Summary
The traditional vegetation remote sensing mechanism model is difficult to accurately describe the radiation transmission process of forest canopy at high spatial resolution, and does not consider the influence of adjacent cell radiation, resulting in a decrease in the inversion accuracy of vegetation parameters. The computer simulation model is highly complex and difficult to apply directly.
Using a bidirectional reflectivity factor calculation method for high spatial resolution, combining geometric optical model and spectral invariance theory, considering the influence of adjacent cell radiation, introducing path length information, and building a forest canopy remote sensing mechanism model suitable for high spatial resolution , by calculating the bidirectional reflectivity factor, improve the accuracy of vegetation parameter inversion.
Improve the accuracy of inversion of the parameters of forest canopy vegetation under high spatial resolution, and can more accurately describe the three-dimensional geometric structure and spatial heterogeneity of forest canopy, satisfying the ecology of the fine structure of small-scale clusters and community forest canopy Requirements, model is highly applicable.
Smart Images

Figure CN119000617A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of remote sensing and relates to constructing a bidirectional reflectance factor analysis calculation method for a high spatial resolution forest canopy. Background Art
[0002] With the increasing availability of remote sensing observation methods such as high-resolution satellites, drones, and lidar, it has become possible to obtain high-resolution remote sensing images, which provides an effective way to accurately describe the spatial changes in forest canopy vegetation parameters. Compared with low- and medium-resolution remote sensing images, high-resolution remote sensing images provide richer texture and structural information and are widely used in qualitative studies such as tree classification and crown extraction. In particular, with the development of drones and lidar in recent years, some studies have been devoted to successfully extracting structural parameters such as tree height, crown width, and coverage through stereo pairs of drone images and lidar point cloud data. However, while these high-spatial-resolution forest canopy qualitative studies are developing rapidly, related quantitative studies still use vegetation indices commonly used in low- and medium-spatial-resolution vegetation remote sensing, or directly migrate traditional mechanism models. With the improvement of the spatial resolution of remote sensing images, the forest canopy shows stronger spatial heterogeneity, and the individual geometric structure characteristics are more obvious. Compared with low- and medium-spatial-resolution remote sensing images, its mechanism of action with light is quite different. The light shielding and cross-radiation differences between adjacent pixels in high spatial resolution remote sensing images cannot be ignored. The reflected radiation of remote sensing pixels is not only affected by the incident sunlight, but also by the radiation from adjacent pixels. This means that in high-resolution remote sensing images, the radiation brightness information of pixels not only reflects the radiation information of the pixel itself, but also includes the radiation contribution from adjacent pixels. This radiation influence of adjacent pixels leads to the complexity of pixel radiation brightness information in remote sensing images, affecting the inversion accuracy of vegetation parameters.
[0003] Traditional vegetation remote sensing mechanism models, such as the 4-Scale model and the unified model, are mainly used for vegetation parameter inversion of low and medium spatial resolution remote sensing images. They focus on the radiation transfer process within the pixel, without considering the radiation effects caused by mutual shading of adjacent pixels and cross-radiation differences. With the improvement of the spatial resolution of remote sensing images, the forest canopy shows stronger spatial heterogeneity, and the radiation effects of adjacent pixels cannot be ignored, which makes it difficult for traditional models to accurately describe the radiation transfer process of the forest canopy, reducing its applicability and further limiting the inversion accuracy. In addition, although computer simulations such as the LESS model (LargE-Scale remote sensing data and image Simulation framework) can provide detailed radiation transfer analysis, they are difficult to directly apply to the actual vegetation parameter inversion due to their high complexity and lack of analytical solution expression.
[0004] The technical disadvantages of traditional solutions can be summarized as follows:
[0005] 1) Not considering the radiation effect of neighboring pixels: Traditional radiation transfer models mainly focus on the interior of pixels and ignore the radiation effect of neighboring pixels of the forest canopy in high spatial resolution remote sensing images, resulting in a decrease in the inversion accuracy of vegetation parameters at high spatial resolution.
[0006] 2) Lack of consideration of the complexity of the geometric structure of the forest canopy: Traditional radiation transfer models mostly use parameters with average statistical significance, which cannot accurately characterize and describe the three-dimensional geometric structure and spatial heterogeneity of the forest canopy at high spatial resolution, resulting in the model being unable to meet the needs of refined monitoring.
[0007] 3) Limited practicality: Although computer simulation models can simulate the radiation transfer process of forest canopies at high spatial resolution, their practicality is limited due to the lack of analytical expression and the complexity of the model input and simulation process, making it difficult to directly guide the extraction of vegetation parameters. Summary of the invention
[0008] In view of the problems existing in the prior art, the purpose of the present invention is to provide a method for calculating the bidirectional reflectance factor of the forest canopy with high spatial resolution. The present invention aims to fully consider the radiation influence of adjacent pixels, utilize the canopy height model (CHM) that can accurately describe the geometric structure information of the forest canopy, introduce path length information, and construct a forest canopy remote sensing mechanism model suitable for high spatial resolution, in order to give full play to the advantages of high-resolution images in depicting the geometric structure and spatial heterogeneity of trees, and improve the accuracy of inversion of high spatial resolution forest canopy vegetation parameters. The technical solution of the present invention considers the radiation influence of adjacent pixels, and combines the geometric optical model and the spectral invariance theory to give an analytical expression of the bidirectional reflectance factor of the high-resolution remote sensing pixel of the forest canopy, which can effectively improve the inversion accuracy of the forest canopy vegetation parameters at high spatial resolution, meet the needs of ecology for the fine structure of small-scale clusters and community forest canopies, and has broad application prospects.
[0009] The present invention establishes a high spatial resolution forest canopy remote sensing model that takes into account the radiation influence of adjacent pixels, thereby improving the accuracy of vegetation parameter inversion at high spatial resolution. The present invention introduces forest canopy geometric structure information to more accurately describe the forest canopy structure at high spatial resolution and meet ecological needs. The present invention provides an analytical expression for high spatial resolution forest canopy radiation transmission, which is conducive to the subsequent development of corresponding high-precision inversion methods based on this.
[0010] The technical solution of the present invention is:
[0011] A method for calculating the bidirectional reflectivity factor of a high-resolution remote sensing image of a forest canopy, the steps of which include:
[0012] 1) Determine the basic information of the remote sensing image and the corresponding canopy height model CHM, including the image range, image spatial resolution and canopy height model CHM spatial resolution; collect forest canopy structure parameters and forest spectral parameters, the forest canopy structure parameters include leaf area volume density FAVD, G function, canopy height model CHM, the spectral parameters include leaf reflectance r L ; Determine the observation geometry parameters and the sky scattered light ratio, the observation geometry parameters include the solar zenith angle θ s , solar azimuth Observation zenith angle θ v and the observation azimuth
[0013] 2) Using the forest canopy structure parameters, the spectral parameters and the observation geometry parameters, a bidirectional reflectance factor (BRF) is calculated that only the primary scattering occurring inside the target pixel is considered. self_sgl ;
[0014] 3) Using the forest canopy structure parameters, the spectral parameters and the observation geometry parameters, the first scattering bidirectional reflectance factor BRF of each target pixel affected by the adjacent pixels is calculated. adj_sgl ;
[0015] 4) Calculating the multiple scattering bidirectional reflectivity factor BRF considering only the target pixel internal multiple scattering by using the forest canopy structure parameters, the spectral parameters, the observation geometry parameters and the intermediate parameter leaf area index LAI obtained in the primary scattering process self_mul ;
[0016] 5) Calculate the multiple scattering bidirectional reflectance factor BRF of each target pixel affected by adjacent pixels using the forest canopy structure parameters, the spectral parameters, the observation geometry parameters and the intermediate parameter leaf area index LAI obtained from the primary scattering process adj_mul ;
[0017] 6) The bidirectional reflectivity factor BRF of the canopy layer that only considers the internal scattering process of the target pixel self_sgl and the canopy multiple scattering bidirectional reflectivity factor BRF which only considers the internal scattering process of the target pixel self_mul Add together to get the total bidirectional reflectivity factor BRF that only considers the internal scattering process of the target pixel self ;
[0018] 7) The canopy primary scattering bidirectional reflectivity factor (BRF) considering the radiation effect of adjacent pixels adj_sgl and the canopy multiple scattering bidirectional reflectivity factor (BRF) considering the radiation effect of adjacent pixels adj_mul Add together to get the total bidirectional reflectivity factor BRF of each target pixel affected by the neighboring pixelsadj .
[0019] Further, in step 2), according to ERF self_sgl =K g ·r g +K l ·r l Calculate the single scattering bidirectional reflectivity factor BRF self_sgl ; Among them, K g Represents the proportion of the illuminated soil area, K l represents the proportion of leaf area exposed to light, r g represents the reflectivity of the soil under illumination, r l represents the reflectivity of the illuminated leaves;
[0020] 21) Calculate the proportion of illuminated soil area K through the path length distribution model g , the proportion of the illuminated soil area K g The large pores between the tree crowns g1 , small pores in the shadow ground K g2 , the soil spot K observed through the pores in the tree canopy g3 Composition; among them, K g1 =P(l(Ω s =0),l(Ω v =0)) represents the path length l(Ω) in the direction of solar incidence v ) is 0, the path length in the sensor observation direction is l(Ω s ) is 0; K g2 =P(l(Ω s >0),l(Ω v =0))·P(Ω s |l(Ω s >0),l(Ω v =0)), P(l(Ω s >0),l(Ω v =0)) represents the path length l(Ω) in the sensor observation direction v ) is 0 and the path length in the direction of solar incidence is l(Ω s ) is greater than 0, P(Ω s |l(Ω s >0),l(Ω v =0)) refers to the porosity obtained by using the path length distribution based on the area ratio of the shadow soil. Where G(Ω s ) represents the projection ratio of the leaf in the direction of solar incidence, FAVD is the leaf area density, l g2 (Ω s ) represents the path length of the shadowed soil portion in the direction of solar incidence, pl_g2 (Ω s ) is the path length l g2 (Ω s ), the probability density distribution of max , l min Represent the path length l g2 (Ω s ) maximum and minimum values; K g3 =P(l(Ω v >0))·P(Ω v |l(Ω v >0)), P(l(Ω v >0))·P(Ω v |l(Ω v >0)) is the path length l(Ω) in the observation direction v ) is greater than 0, l g3 (Ω v ) represents the path length of the canopy in the vertical observation direction, p l_g3 (Ω v ) is the path length l g3 (Ω v ), G(Ω v ) represents the projection ratio of the blade in the vertical observation direction;
[0021] 22) Proportion of illuminated leaf area K l Bidirectional reflectivity factor BRF l =BRF lc +BRF lt ; By calculating the proportion of the illuminated crown area K c and the area ratio of the shaded tree crown K t , the proportion of the illuminated crown area K c Converted to bidirectional reflectance factor BRF of illuminated leaves lc , the area ratio K of the shadow crown t Converted to bidirectional reflectance factor BRF of illuminated leaves lt ; Among them, the proportion of illuminated crown area K c =P(l(Ω s =0),l(Ω v >0))·[1-P(Ω v |l(Ω s =0),l(Ω v >0))], the area ratio of the shadow canopy K t =P(l(Ω s >0),l(Ω v >0))·[1-P(Ω v |l(Ω s>0),l(Ω v >0))];P(l(Ω s =0),l(Ω v >0)) refers to the joint probability that the path length of the part in the observation direction is greater than 0 and the path length in the direction of solar incidence is 0.
[0022] Furthermore, in step 3), the shading factor S of the illumination crown component is first calculated c and the shading factor S of the soil component of the illumination g Then, the new illumination crown component K considering the influence of neighboring pixel radiation is derived c ′ and the new light soil component K′ g Then, the first scattering bidirectional reflectivity factor BRF of each target pixel affected by the adjacent pixels is calculated. adj_sgl .
[0023] Furthermore, the first scattering bidirectional reflectivity factor BRF of each target pixel affected by the adjacent pixels is calculated. adj_sgl The specific method is:
[0024] 31) Calculate the proportion of large pores P formed by the light canopy being blocked by adjacent pixels c (l=0,Ω s )=P(F c (l=0,Ω s )|K c ), the proportion of small pores formed by being blocked by adjacent pixels Then calculate the shading factor S on the light canopy c =[1-P c (l=0,Ω s )]·[1-P c (l>0,Ω s )]; then calculate the target pixel illumination crown component area ratio K after being affected by the radiation of the neighboring pixels c ′=K c ·(1-S c );l(Ω s |K c ) represents the path length of the target pixel in the direction of the sun’s incidence after the original illuminated canopy is affected by the neighboring pixels, p l (Ω s |K c ) represents the path length l(Ω s |K c )’s probability density distribution;
[0025] 32) Calculate the proportion of macropores P formed on the illuminated soil after being blocked by adjacent pixels g (l=0,Ω s)=P(F g (l=0,Ω s )|K g ), the proportion of small pores formed by being blocked by adjacent pixels and the shading factor S on the light soil g =[1-P g (l=0,Ω s )]·[1-P g (l>0,Ω s )], and then calculate the target pixel illumination soil component area proportion K′ after being affected by the radiation of the neighboring pixels g =K g ·(1-S g );l(Ω s |K g ) represents the path length of the target pixel in the direction of the sun’s incidence after the original part of the target pixel that illuminates the soil is affected by the neighboring pixels. l (Ω s |K g ) represents the path length l(Ω s |K g )’s probability density distribution;
[0026] 33) Calculate the first scattering bidirectional reflectivity factor BRF adj_sgl =(1-β)·(K′ g ·r s +BRF l ′); where β is the proportion of sky scattered light.
[0027] Further, in step 4), firstly according to the proportion of illuminated crown area K c , shadow crown area ratio K t And the vertical observation path length distribution F(l>0,Ω v ), calculated Then, based on the probability p of a photon scattering from a leaf and colliding with a leaf in the canopy again, 1 , multiple re-collision probability p m , calculate the bidirectional reflectance factor BRF by ignoring the soil and only considering the multiple scattering between the canopy layers self_vmul And consider the multiple scattering between soil and canopy interior bidirectional reflectivity factor BRF self_smul ; Then the canopy multiple scattering bidirectional reflectivity factor BRF is calculated when only the target pixel is considered self_mul =BRF self_vmul +BRF self_smul .
[0028] Furthermore, in step 5), the multiple scattering bidirectional reflectivity factor BRF of each target pixel affected by the adjacent pixels is calculated.adj_mul The method is as follows: First, according to the proportion of light crown area K c , shadow crown area ratio K t And the vertical observation path length distribution F(l>0,Ω v ), calculated Then calculate the bidirectional reflectivity factor of multiple scattering of neighboring pixels to the target pixel Among them, LAI i Represents the leaf area index of the surrounding i-th pixel, LAI self Represents the leaf area index of the current target pixel, BRF mul_i represents the multiple scattering BRF calculated for the surrounding i-th pixel when only internal multiple scattering is considered, W i is the normalized weight matrix related to the distance; then the multiple scattering bidirectional reflectivity factor BRF is calculated adj_mul =BRF self_mul +BRF others_mul .
[0029] Furthermore, p 1 =0.7exp(k1·LAI)-0.66exp(k2·LAI), Among them, k1 = 0.0045exp(1.2555cosθ), k2 = 0.1982lncosθ-0.7146, i(θ) is the interception probability of the canopy in the θ direction, and θ is the solar zenith angle.
[0030] Furthermore, Where i0 = β·i D +(1-β)·i S , represents the average interception probability of the canopy for direct sunlight and scattered light from the sky, ω is the primary scattering albedo of the leaves, r s represents soil reflectivity, R dn represents the canopy bottom albedo, T dn and T up represents the canopy downward and upward transmittance, p(Ω s ) and p(Ω v ) represent the porosity in the solar incidence direction and the sensor observation direction, respectively.
[0031] A server, characterized in that it comprises a memory and a processor, wherein the memory stores a computer program, the computer program is configured to be executed by the processor, and the computer program comprises instructions for executing each step in the above method.
[0032] A computer-readable storage medium stores a computer program, wherein the computer program implements the steps of the above method when executed by a processor.
[0033] The key points of the present invention include:
[0034] ① Combining the geometric optics model and spectral invariance theory, the BRF of high-resolution remote sensing pixels in the forest canopy is divided into primary scattering contribution and multiple scattering contribution, and then described using the geometric optics model and spectral invariance theory respectively, to more comprehensively describe the radiation transmission process of the forest canopy.
[0035] ② Considering the radiation influence of neighboring pixels, the shielding factor and neighborhood convolution algorithm are introduced to quantitatively express the contribution of neighboring pixels to the BRF of the target pixel and distinguish the BRF of the pixel itself and the neighboring pixels.
[0036] ③ The concept of shielding factor is used to quantitatively express the influence of neighboring pixels on the primary scattering of the target pixel.
[0037] ④ Use the neighborhood convolution algorithm to quantitatively express the impact of neighboring pixels on multiple scattering of target pixels.
[0038] ⑤Introduce the path length distribution model and combine it with high-resolution CHM data to more accurately describe the spatial distribution and three-dimensional structure of the forest canopy at high spatial resolution and improve the model accuracy.
[0039] The advantages of the present invention are as follows:
[0040] ① Improve the accuracy of vegetation parameter inversion under high spatial resolution: Considering the radiation influence of adjacent pixels, it can more accurately simulate the radiation transfer process of the forest canopy under high spatial resolution conditions, thereby improving the accuracy of forest canopy vegetation parameter inversion based on physical models.
[0041] ②The model effectively distinguishes the contribution of pixel's own reflected radiation and the radiation influence brought by neighboring pixels, so that the forest canopy vegetation parameters inverted based on this model can truly represent the vegetation conditions on high-resolution remote sensing pixels.
[0042] ③Meet ecological needs: Using high spatial resolution CHM data, the geometric structure information of the forest canopy is introduced, which improves the model's accuracy in depicting the three-dimensional geometric structure of the forest canopy and meets the ecological needs for the fine structure of small-scale clusters and community forest canopies.
[0043] ④The model is highly applicable: the model has a simple structure and is easy to implement. Compared with complex models such as LESS, it is more suitable for practical applications. BRIEF DESCRIPTION OF THE DRAWINGS
[0044] Figure 1 The present invention is a flow chart of the method.
[0045] Figure 2 Comparison chart of BRF value calculated by the model and BRF value simulated by LESS;
[0046] Figure 2 (a) Comparison of the calculated values of the APPLE-GO model in the red band and the simulated values of the LESS model in the red band when only the internal scattering of the target pixel is considered;
[0047] Figure 2 (b) Comparison of the calculated values of the APPLE-GO model in the near-infrared band and the simulated values of the LESS model in the near-infrared band when only the internal scattering of the target pixel is considered;
[0048] Figure 2 (c) Comparison of the calculated values of the APPLE-GO model in the red band and the simulated values of the LESS model in the red band considering the radiation influence of adjacent pixels;
[0049] Figure 2 (d) Comparison of the calculated values of the APPLE-GO model in the near-infrared band and the simulated values of the LESS model in the near-infrared band considering the influence of radiation from adjacent pixels.
[0050] Figure 3 This is a comparison chart of the BRF calculation value and LESS simulation value of the pixel close to the edge in scene 1 by the APPLE-GO model;
[0051] Figure 3 (a) Schematic diagram of the location of pixels close to the edge in scene 1.
[0052] Figure 3 (b) Polar coordinate comparison of the BRF calculated by the APPLE-GO model in the red band and the simulated value by the LESS model in the red band considering only the target pixel (the left side is the calculated result of the APPLE-GO model, and the right side is the simulated result of the LESS model);
[0053] Figure 3 (c) Polar coordinate comparison of the BRF calculated by the APPLE-GO model in the near-infrared band and the simulated value by the LESS model in the near-infrared band considering only the target pixel (the left side is the calculated result of the APPLE-GO model, and the right side is the simulated result of the LESS model);
[0054] Figure 3 (d) Polar coordinate comparison of the BRF calculated by the APPLE-GO model in the red band and the simulated value by the LESS model in the red band considering the radiation effect of adjacent pixels (the left side is the calculation result of the APPLE-GO model, and the right side is the simulation result of the LESS model);
[0055] Figure 3(e) Polar coordinate comparison of the BRF calculated by the APPLE-GO model in the near-infrared band and the simulated value by the LESS model in the near-infrared band considering the radiation effect of adjacent pixels (the left side is the calculation result of the APPLE-GO model, and the right side is the simulation result of the LESS model);
[0056] Figure 4 This is a comparison chart of reflectivity images;
[0057] Figure 4 (a) RGB reflectance composite image from Beijing-3 satellite;
[0058] Figure 4 (b) Reflectance RGB true color composite image calculated by the APPLE-GO model. DETAILED DESCRIPTION
[0059] The present invention is further described in detail below in conjunction with the accompanying drawings. The examples given are only used to explain the present invention but not to limit the scope of the present invention.
[0060] The high spatial resolution forest canopy radiation transfer model APPLE-GO proposed in this paper is established on the basis of in-depth analysis of the radiation impact of neighboring pixels of the forest canopy. The present invention can accurately simulate the radiation impact brought by neighboring pixels, effectively distinguish the target pixel's own BRF (bidirectional reflectivity factor) from the radiation impact of neighboring pixels, thereby improving the inversion accuracy of vegetation parameters such as leaf area index (LAI) and chlorophyll content (LCC).
[0061] The assumptions of the APPLE-GO model are as follows: ① Vertical sensor observation: The model assumes that the sensor observes the forest canopy vertically downward, which is consistent with the observation method of drones and most high-resolution satellites; ② The CHM of the forest canopy is known and has high spatial resolution: The model requires high-precision CHM data to describe the geometric structure of the forest canopy to ensure that the model can accurately calculate the path length distribution and the shielding factor; ③ The leaf area volume density is known at the modeling scale: The model assumes that the leaf area volume density is known at the modeling scale, which can be extracted through ground measurements; ④ The leaves follow a Poisson distribution at the modeling scale; ⑤ The trunk heights of trees in the forest canopy are approximately consistent: The model assumes that the trunk heights of trees in the forest canopy are approximately consistent, which simplifies the model calculations, but can be expanded by considering the differences in trunk heights; ⑥ The leaves and soil background are considered as Lambertian bodies; ⑦ The ground is considered as flat ground: The model assumes that the leaves and soil background are Lambertian bodies, that is, the reflectivity does not change with the observation angle, which simplifies the model calculations; ⑧ The wood components in the forest canopy scene are not considered.
[0062] Specific construction process and detailed formula of APPLE-GO model:
[0063] The APPLE-GO model divides the BRF of high spatial resolution remote sensing pixels of the forest canopy into primary scattering contribution and multiple scattering contribution, and models them separately.
[0064] For primary scattering, based on the geometric optics model, the path length distribution is introduced to calculate the shielding factors of the two components of illuminated soil and illuminated tree crown respectively to express the influence of neighboring pixels on the primary scattering of the target pixel.
[0065] For multiple scattering, based on the spectral invariance theory, the neighborhood convolution algorithm is used to consider the influence of neighboring pixels on the multiple scattering of the target pixel.
[0066] ① Primary scattering modeling
[0067] Primary scattering refers to the BRF presented by the sensor when the solar photon enters the vegetation canopy and hits the canopy element for the first time. Primary scattering is most affected by the vegetation canopy structure, which can be reflected from the hot spot effect. Primary scattering is the main reason for determining the directionality of the BRDF observed by the sensor. According to the theory of geometric optics, primary scattering is mainly determined by the BRF contribution of the illuminated soil and the illuminated leaves, as shown in the following formula (1).
[0068] BRF self_sgl =∑K(Ω s )·r=K g ·r g +K l ·r l (1)
[0069] Where K(Ω s ) represents the area ratio of the lighting component in the scene, represents the direction of the sun's incidence (θ s is the solar zenith angle, is the solar azimuth), r is the reflectivity of the corresponding component, K g Represents the proportion of the illuminated soil area, K l represents the proportion of leaf area exposed to light, r g and r l Represent the reflectance of these two light components. The light components include the light-irradiated soil and the light-irradiated leaves. This study assumes that both parts are Lambertian bodies, so only their area ratio changes with the direction of the sun's incidence, and the reflectance does not change with the angle.
[0070] When only the primary scattering inside the pixel is considered, the proportion of the illuminated soil area K g The large pores between the tree crowns g1 , small pores in the shadow ground K g2 , the soil spot K observed through the pores in the tree canopy g3The APPLE-GO model calculates the area ratios of these three parts through the path length distribution model, and the formulas are as follows (2) to (6).
[0071] K g1 =P(l(Ω s =0), l(Ω v =0)) (2)
[0072] Where l represents the path length, l(Ω s =0) means that the path length in the direction of solar incidence is 0, l(Ω v =0) indicates that the path length in the sensor observation direction is 0. This formula expresses the joint probability when the path lengths in both directions are 0. g2 For , it can be expressed based on the path length distribution as:
[0073] K g2 =P(l(Ω s >0),l(Ω v =0))·P(Ω s |l(Ω s >0),l(Ω v =0)) (3)
[0074] Where P(l(Ω s >0),l(Ω v =0)) represents the joint probability that the path length in the sensor observation direction is 0 and the path length in the sun incidence direction is greater than 0. This part represents the area proportion of the shadow soil in the traditional sense; where P(Ω s |l(Ω s >0),l(Ω v =0)) is the conditional probability, which refers to the expression of porosity obtained by using path length distribution based on the area ratio of the shadow soil, which can be expressed as:
[0075]
[0076] Where G(Ω s ) represents the projection ratio of the leaf in the direction of solar incidence, FAVD is the leaf area density, l g2 (Ω s )=l(Ω s |l(Ω s >0),l(Ω v =0)) represents the path length of the shadow soil part in the direction of solar incidence, p l_g2 (Ω s )=p l (Ω s |l(Ω s >0),l(Ωv =0)) is the probability density distribution of this part of the path length, and the upper and lower limits of the integral represent the maximum and minimum values of the path length respectively.
[0077] K g3 =P(l(Ω v >0))·P(Ω v |l(Ω v >0)) (5)
[0078] Where P(l(Ω v >0))·P(Ω v |l(Ω v >0)) means the proportion of porosity in the part (i.e., the crown part) with a path length greater than 0 in the observation direction (i.e., the vertical direction). v |l(Ω v >0)) is the conditional probability, which can be expressed as:
[0079]
[0080] Where G(Ω v ) represents the projection ratio of the leaf in the vertical observation direction, FAVD is the leaf area density, l g3 (Ω v )=l(Ω v |l(Ω v >0)) represents the path length of the canopy in the vertical observation direction, p l_g3 (Ω v )=p l (Ω v |l(Ω v >0)) is the probability density distribution of this part of the path length, and the upper and lower limits of the integral represent the maximum and minimum values of the path length respectively.
[0081] The illuminated leaves come from the illuminated crown and the shadowed crown. Therefore, the area ratio of the illuminated crown and the shadowed crown is calculated first, and then converted into the BRF contribution of the illuminated leaves. c The calculation formula is shown in formula (7), the area ratio of the shadow crown K t As shown in formula (8), the BRF contribution of the illuminated leaf is obtained after conversion, as shown in formulas (9) to (16).
[0082] K c =P(l(Ω s =0), l(Ω v >0))·[1-P(Ω v |l(Ω s =0), l(Ω v >0))] (7)
[0083] K t =P(l(Ω s >0), l(Ω v >0))·[1-P(Ω v |l(Ω s >0), l(Ω v >0))] (8)
[0084] Where P(l(Ω s =0),l(Ω v >0)) means the joint probability that the path length of the part (i.e., the canopy part) in the observation direction (i.e., the vertical direction) is greater than 0 and the path length in the direction of solar incidence is 0. v ) represents the portion actually illuminated by light after removing the internal porosity.
[0085] The formula for converting canopy illumination to leaf illumination (reference Mu X, Hu R, Zeng Y, et al. 2017. Estimating structural parameters of agricultural crops from ground-based multi-angular digital images with afractional model of sun and shade components. Agricultural and Forest Meteorology, 246: 162-177.):
[0086]
[0087] Where Γ(Ω s ,Ω v ) / π is the scattering phase function, which represents the proportion of light scattered to the observation direction after colliding with the leaves. L ) / 2π is the leaf inclination distribution function (based on the upper surface of the leaf), γ(Ω s →Ω v ) / π is the scattering phase function of the blade scale. When the emission and incidence occur on the same surface, it is multiplied by the blade reflectivity r L ; When the emission and incidence occur on different surfaces, multiply by the leaf transmittance t L . P(z,Ω s ,Ω v ) is the bidirectional porosity at height z, which is equal to the product of the porosity in the observation direction and the solar incidence direction, multiplied by the hotspot factor, expressed as follows:
[0088] P c(z, Ω s , Ω v )=exp(-G(Ω v )·FAVD·z)·exp(-G(Ω s )·FAVD·z / cosθ s )·C HS (12)
[0089]
[0090] Among them C HS is the hot spot factor. In formula (13), tanθ s It is a special form of expression for vertical observation, s L =π 2 ·d L / 16 is the expression form when the blades are spherically distributed, where d L is the blade size.
[0091] The formula for converting shadow canopy to illuminated leaves is:
[0092]
[0093] In formula (15), represents the average path length in the direction of the sun’s incidence. Therefore, the proportion of the leaf area illuminated by light, K l BRF Contribution l It can be expressed as the sum of the light crown and the shadow crown:
[0094] BRF l =BRF lc +BRF lt (16)
[0095] When considering the radiation blocking effect of neighboring pixels, the concept of shielding factor is introduced. In the APPLE-GO model, shielding factors of the illumination canopy and illumination soil components are defined respectively, and the proportion of the area reduction caused by the shielding effect of the two components from the neighboring pixels is calculated respectively. This proportion is called the shielding factor. The shielding factor is calculated by the path length distribution model, which takes into account the probability of removing large pores and small pores in the direction of the sun.
[0096] The proportion of large pores formed by the shade of adjacent pixels on the illuminated tree canopy is calculated as shown in formula (17), and the proportion of small pores formed by the shade of adjacent pixels is calculated as shown in formula (18). The shade factor S on the illuminated tree canopy is c It is expressed in the form of the probability of large pores and small pores, as shown in formula (19).
[0097] P c (l=0, Ω s )=P(Fc (l=0,Ω s )|K c ) (17)
[0098]
[0099] S c =[1-P c (l=0, Ω s )]·[1-P c (l>0, Ω s )] (19)
[0100] In formula (18), l(Ω s |K c ) represents the path length of the target pixel in the direction of the sun’s incidence after the original illuminated canopy is affected by the neighboring pixels, p l (Ω s |K c ) represents the probability density distribution of this part of the path length. The proportion of the target pixel illumination crown component area K after being affected by the radiation of the neighboring pixels c ′ is:
[0101] K c ′=K c ·(1-S c ) (20)
[0102] Similarly, the proportion of large pores formed by the shielding of adjacent pixels on the illuminated soil is calculated as shown in formula (21), and the proportion of small pores formed by the shielding of adjacent pixels is calculated as shown in formula (22). The shielding factor S on the illuminated soil is g It is expressed in the form of the probability of large pores and small pores, as shown in formula (23).
[0103] P g (l=0, Ω s )=P(F g (l=0,Ω s )|K g ) (twenty one)
[0104]
[0105] S g =[1-P g (l=0, Ω s )]·[1-P g (l>0, Ω s )] (twenty three)
[0106] In formula (22), l(Ω s |K g) represents the path length of the target pixel in the direction of the sun’s incidence after the original part of the target pixel that illuminates the soil is affected by the neighboring pixels. l (Ω s |K g ) represents the probability density distribution of this part of the path length. The proportion of the target pixel’s illuminated soil component area K′ after being affected by the radiation of the neighboring pixels g That is:
[0107] K′ g =K g ·(1-S g ) (twenty four)
[0108] The new illuminated tree crown area ratio and the new illuminated soil area ratio calculated by equations (20) and (24) are used to replace the original illuminated tree crown and illuminated soil area ratios when only the interior of the pixel is considered, respectively, and the primary scattering expression of the target pixel after being affected by the neighboring pixels can be obtained:
[0109] BRF adj_sgl =(1-β)·(K′ g ·r s +BRF l ′) (25)
[0110] ②Multiple scattering modeling
[0111] Multiple scattering refers to the BRF presented by solar photons after entering the vegetation canopy and colliding back and forth between various components multiple times before finally entering the sensor. As the number of collisions increases, multiple scattering tends to be approximately isotropic. Therefore, the heterogeneity of the canopy does not affect multiple scattering as much as the impact on primary scattering. However, if only primary scattering is considered and multiple scattering is ignored, the BRDF values simulated by the model will have a certain deviation. In the visible light band where the albedo of the primary scattering of leaves is low, the impact of multiple scattering is small, and primary scattering dominates the reflected radiation of the canopy. However, in the near-infrared band where the albedo of the primary scattering of leaves is high, the impact of multiple scattering cannot be ignored.
[0112] When only considering the multiple scattering inside the pixel, it can be divided into ignoring the soil, considering only the multiple scattering contribution between the canopy (black soil process), and considering the multiple scattering contribution between the soil and the canopy (white soil process). The model defines a re-collision probability p 1 , that is, the probability of a photon colliding with a leaf in the canopy again after being scattered by a leaf, is related to the angle of the incident direction. The calculation formulas are shown in (26) to (28). The model also defines the probability of multiple re-collisions p m , m is the number of collisions, m>1; that is, the probability of a photon colliding with a leaf in the canopy again after being scattered by a leaf is independent of the angle, as shown in equations (29) to (30).
[0113] p1 =0.7exp(k1·LAI)-0.66exp(k2·LAI) (26)
[0114] Where k1 and k2 are fitted expressions of the solar zenith angle, and the formula is as follows:
[0115] k1=0.0045exp(1.2555cosθ) (27)
[0116] k2=0.1982lncosθ-0.7146 (28)
[0117] Where θ is the solar zenith angle.
[0118]
[0119] where i D is the canopy hemispherical interception probability, that is, the spatial integral of the canopy interception probability in all directions:
[0120]
[0121] where i(θ) is the interception probability of the canopy in the θ direction.
[0122] In formulas (26) to (30), the leaf area index LAI is a parameter obtained from the primary scattering process: the proportion of illuminated crown area K c , shadow crown area ratio K t , and the vertical observation path length distribution (excluding the part with path length of 0, F(l>0,Ω v ))Calculation:
[0123]
[0124] Based on a re-collision probability p 1 and the probability of multiple re-collisions p m The expression of ignoring the soil and only considering the multiple scattering contribution between the canopy layers (black soil process) is shown in formula (32), and the expression of considering the multiple scattering contribution between the soil and the interior of the canopy layer (white soil process) is shown in formula (33).
[0125]
[0126] In formula (32), i0 = β·i D +(1-β)·i S , represents the average interception probability of the canopy for direct sunlight and scattered light from the sky, and ω is the leaf primary scattering albedo, which can be approximately equal to the sum of the leaf reflectivity and transmittance. s represents soil reflectivity, R dn represents the canopy bottom albedo, Tdn and T up represents the canopy downward and upward transmittance, p(Ω s ) and p(Ω v ) represent the porosity in the solar incident direction and the sensor observation direction respectively. dn , T dn and T up The expressions are:
[0127]
[0128]
[0129] Canopy multiple scattering BRF when only the target pixel is considered self_mul It is expressed as the sum of the multiple scattering contribution between the canopy layers and the multiple scattering contribution between the soil and the interior of the canopy layers:
[0130] BRF self_mul =BRF self_vmul +BRF self_smul (37)
[0131] When considering the radiation blocking effect of neighboring pixels, the model uses the neighborhood convolution algorithm to convolve the relative size of the target pixel's neighboring pixels' LAI and the plane Euclidean distance as weights to obtain the contribution of neighboring pixels to the multiple scattering of the target pixel:
[0132]
[0133] Among them, LAI i Represents the leaf area index of the surrounding i-th pixel, LAI self Represents the leaf area index of the current target pixel, BRF mul_i represents the multiple scattering BRF calculated for the surrounding i-th pixel when only internal multiple scattering is considered, W i is the normalized weight matrix associated with the distance.
[0134] By adding the multiple scattering contribution within the pixel and the multiple scattering increment under the influence of neighboring pixels, the expression of multiple scattering of the target pixel after being affected by neighboring pixels can be obtained:
[0135] BRF adj_mul =BRF self_mul +BRF others_mul (39)
[0136] ③Total BRF expression of target pixel:
[0137] The total BRF of the target pixel includes the total BRF contribution of the target pixel itself selfAnd the total BRF contribution BRF of the target pixel after considering the influence of neighboring pixels adj The target pixel's own BRF contributes to the BRF self Expressed as the canopy primary scattering BRF when only the target pixel is considered self_sgl And canopy multiple scattering BRF when only the target pixel is considered self_mul The sum of the two parts:
[0138] BRF self =BRF self_sgl +BRF self_mul (40)
[0139] The target pixel BRF contribution BRF after being affected by the radiation of neighboring pixels adj Expressed as canopy primary scattering BRF considering the radiation influence of adjacent pixels adj_sgl And canopy multiple scattering BRF considering the influence of neighboring pixel radiation adj_mul The sum of the two parts:
[0140] BRF adj =BRF adj_sgl +BRF adj_mul (41)
[0141] Summary of APPLE-GO model input parameters:
[0142] Forest canopy structure parameters: leaf area volume density (FAVD), G function (G), CHM, etc.
[0143] Spectral parameters: leaf reflectance (r L ), leaf transmittance (t L ), soil reflectivity (r s )wait.
[0144] Observation geometry parameters: solar zenith angle (θ s ), solar azimuth Observation zenith angle (θ v ), observation azimuth wait.
[0145] Other parameters: sky scattered light ratio (β), etc.
[0146] Summary of model output:
[0147] The target pixel’s own BRF contribution: only the BRF of the internal scattering process of the pixel is considered.
[0148] BRF contribution of the target pixel after being affected by neighboring pixels: BRF after considering the radiation influence of neighboring pixels.
[0149] The model process is as follows Figure 1 As shown, including:
[0150] 1. Data preparation:
[0151] Determine the basic information of the remote sensing image and the corresponding canopy height model (CHM), including image range, image spatial resolution, and CHM spatial resolution.
[0152] Collect necessary input data, including forest canopy structural parameters such as leaf area density FAVD, leaf reflectivity r L Spectral characteristic data.
[0153] At the same time, determine the solar zenith angle θ s , solar azimuth Observation zenith angle θ v and the observation azimuth Other parameters include observation geometry parameters and the proportion of sky scattered light.
[0154] 2. Calculate the BRF considering only the primary scattering inside the target pixel self_sgl :Using structural parameters such as leaf area density FAVD, G function, canopy height model CHM, leaf reflectivity r L Isospectral parameters and solar zenith angle θ s Based on the observation geometric parameters such as equations (1) to (16), the primary scattering BRF of the target pixel can be calculated when only the radiation inside the target pixel is considered and the radiation influence of the neighboring pixels is not considered. The G function is a well-known parameter in the field of quantitative remote sensing of vegetation, which means the projection coefficient of the unit leaf area on the plane perpendicular to the observation direction. Specifically, based on the geometric optical model (as shown in equation 1), the illuminated leaf component (as shown in equations 2 to 6) and the illuminated soil component (as shown in equations 7 to 16) are calculated respectively, and the sum of the two is the required BRF. self_sgl (As shown in formula 1).
[0155] 3. Calculate the primary scattering BRF of each target pixel affected by the neighboring pixels self_sgl :Using structural parameters such as leaf area density FAVD, G function, canopy height model CHM, leaf reflectivity r L Isospectral parameters and solar zenith angle θ sBased on the observation geometric parameters, the primary scattering BRF of each target pixel under the influence of the radiation of the neighboring pixels can be calculated based on equations (17) to (25). Specifically, the changes in the path lengths of the illumination crown component and the illumination soil component under the influence of the radiation of the neighboring pixels are considered, and the shielding factors of the illumination crown component (as shown in equations 17 to 19) and the shielding factors of the illumination soil component (as shown in equations 21 to 23) are calculated, and then the new illumination crown component (as shown in equation 20) and the new illumination soil component (as shown in equation 24) under the influence of the radiation of the neighboring pixels are derived, and the sum of the two is the desired value (as shown in equation 25).
[0156] 4. Calculate the multiple scattering BRF only considering the internal part of the target pixel self_mul :Using structural parameters such as leaf area density FAVD, G function, canopy height model CHM, leaf reflectivity r L Isospectral parameters, solar zenith angle θ s Based on the observed geometric parameters and the intermediate parameter leaf area index LAI obtained from the primary scattering process (as shown in formula 31), the multiple scattering BRF of the target pixel can be calculated based on formulas (26) to (37), which only considers the radiation inside the target pixel and does not consider the radiation influence of the adjacent pixels. Specifically, the leaf multiple scattering contribution can be calculated based on formulas (26) to (30) and (32), and the soil multiple scattering contribution can be calculated based on formulas (33) to (36). The sum of the two is the desired result (as shown in formula 37).
[0157] 5. Calculate the multiple scattering BRF of each target pixel affected by the neighboring pixels adj_mul :Using structural parameters such as leaf area density FAVD, G function, canopy height model CHM, leaf reflectivity r L Isospectral parameters, solar zenith angle θ s Based on the observed geometric parameters and the intermediate parameter leaf area index LAI obtained from the primary scattering process (as shown in formula 31), the multiple scattering BRF of each target pixel considering the radiation influence of the neighboring pixels can be calculated based on formulas (38) to (39). Specifically, first, on the basis of calculating the multiple scattering BRF considering only the inside of the target pixel, the neighborhood convolution algorithm is used to calculate the contribution of the neighboring pixels to the multiple scattering of the target pixel (as shown in formula 38), and then the sum of the two is the desired result (as shown in formula 39).
[0158] 6. Calculate the total BRF considering only the internal scattering process of the target pixel: the canopy primary scattering BRF contribution BRF considering only the internal scattering process of the target pixel self_sgl And the canopy multiple scattering BRF contribution BRF that only considers the internal scattering process of the target pixel self_mulThe total BRF considering only the internal scattering process of the target pixel is obtained by adding them together (as shown in Equation 40). The total BRF only represents the radiation information occurring inside the target pixel without considering the radiation influence of the neighboring pixels.
[0159] 7. Calculate the total BRF of each target pixel affected by neighboring pixels: the canopy primary scattering BRF contribution BRF considering the radiation influence of neighboring pixels adj_sgl and canopy multiple scattering BRF contribution BRF considering the radiation influence of adjacent pixels adj_mul By adding them together, we can get the total BRF of each target pixel affected by the neighboring pixels (as shown in equation 41). The total BRF represents the reflected radiation information received by the remote sensing sensor including the influence of the neighboring pixels.
[0160] Model forward calculation of high spatial resolution forest canopy remote sensing pixel BRF and verification example:
[0161] To test the performance of the APPLE-GO model under different conditions, it was compared with the simulation results of the computer simulation model LESS and real remote sensing images.
[0162] ①Comparison with LESS
[0163] Three different forest canopy scenarios were designed in the computer simulation LESS. The first forest canopy scenario consisted of an abstract canopy with a canopy height of 5m in a Poisson distribution. The second forest canopy scenario consisted of an abstract canopy with a canopy height of 10m in a manner closer to the actual distribution. The third forest canopy scenario was a mixed scenario consisting of canopies with a height of 5m and 10m in the first and second scenarios in a Poisson distribution. In the experiment, for each forest canopy scenario, three remote sensing image spatial resolutions were set, namely 2m, 5m and 10m. At the same time, the solar zenith angle was set at intervals of 10°, ranging from 0 to 60°, and the solar azimuth was set at intervals of 45°, ranging from 0 to 360°. The specific experimental design information is shown in Table 1.
[0164] Table 1 is the simulation scenario experiment design table
[0165]
[0166] Figure 2 A comparison chart of the model estimation and LESS simulation is shown. The scattered points in the comparison chart represent the results of random sampling of pixels in the three virtual forest canopy scenes under different conditions (sun incidence direction, pixel spatial resolution).
[0167] In order to analyze the change of BRF of high-resolution pixels under vertical observation conditions of different solar incidence directions, the pixels located at the edge of the scene in scene 1 are selected, and the polar coordinate diagram is drawn with the BRF calculated by the APPLE-GO model and the BRF simulated by the LESS model, as shown in the figure. Figure 3 shown.
[0168] ②Comparison with real remote sensing images
[0169] The APPLE-GO model was applied to a sample plot of larch forest in a certain area, and the spatial resolution of its multispectral remote sensing image was 1.2m. After inputting the basic parameters of remote sensing images, forest canopy structure parameters, spectral parameters, and observation geometry parameters into the APPLE-GO model, the calculated reflectance image and the Beijing No. 3 reflectance image were compared as shown in the figure below. Figure 4 As shown (all resampled to 3m spatial resolution).
[0170] The results of comparison with the LESS model simulation data show that ( Figure 2 The BRF results calculated by the APPLE-GO model are in good agreement with the LESS simulation values; the polar coordinate diagram shows (as shown in Figure 3 As shown in the figure, the changes in the values calculated by the APPLE-GO model in all directions are consistent with the LESS simulation values. The BRF of the pixels close to the left edge after being affected by the neighboring pixels shows obvious differentiation in direction, which is well reflected in the simulation results of the APPLE-GO model. Therefore, the APPLE-GO model has a good description of the influence of neighboring pixels.
[0171] Comparison results with the above satellite remote sensing images show that (such as Figure 4 As shown in Figure 2, the results calculated by the APPLE-GO model are generally close to the reflectivity of the Beijing-3 satellite.
[0172] In summary, the APPLE-GO model has a high simulation accuracy for BRF of high spatial resolution forest canopy remote sensing pixels.
[0173] Although the specific embodiments of the present invention are disclosed for the purpose of illustration, the purpose is to help understand the content of the present invention and implement it accordingly, those skilled in the art will understand that various substitutions, changes and modifications are possible without departing from the spirit and scope of the present invention and the appended claims. Therefore, the present invention should not be limited to the content disclosed in the best embodiment, and the scope of protection claimed by the present invention shall be subject to the scope defined in the claims.
Claims
1. A method for calculating the bidirectional reflectivity factor of a high-resolution remote sensing image of a forest canopy, the steps comprising: 1) Determine the basic information of the remote sensing image and the corresponding canopy height model CHM, including the image range, image spatial resolution and canopy height model CHM spatial resolution; collect forest canopy structure parameters and forest spectral parameters, the forest canopy structure parameters include leaf area volume density FAVD, G function, canopy height model CHM, the spectral parameters include leaf reflectance r L ; Determine the observation geometry parameters and the sky scattered light ratio, the observation geometry parameters include the solar zenith angle θ s , solar azimuth Observation zenith angle θ v and the observation azimuth 2) Using the forest canopy structure parameters, the spectral parameters and the observation geometry parameters, a bidirectional reflectance factor (BRF) is calculated that only the primary scattering occurring inside the target pixel is considered. self_sgl ; 3) Using the forest canopy structure parameters, the spectral parameters and the observation geometry parameters, the first scattering bidirectional reflectance factor BRF of each target pixel affected by the adjacent pixels is calculated. adj_sgl ; 4) Calculating the multiple scattering bidirectional reflectivity factor BRF considering only the target pixel internal multiple scattering by using the forest canopy structure parameters, the spectral parameters, the observation geometry parameters and the intermediate parameter leaf area index LAI obtained in the primary scattering process self_mul ; 5) Calculate the multiple scattering bidirectional reflectance factor BRF of each target pixel affected by adjacent pixels using the forest canopy structure parameters, the spectral parameters, the observation geometry parameters and the intermediate parameter leaf area index LAI obtained from the primary scattering process adj_mul ; 6) The bidirectional reflectivity factor BRF of the canopy layer that only considers the internal scattering process of the target pixel self_sgl and the canopy multiple scattering bidirectional reflectivity factor BRF which only considers the internal scattering process of the target pixel self_mul Add together to get the total bidirectional reflectivity factor BRF that only considers the internal scattering process of the target pixel self ; 7) The canopy primary scattering bidirectional reflectivity factor (BRF) considering the radiation effect of adjacent pixels adj_sgl and the canopy multiple scattering bidirectional reflectivity factor (BRF) considering the radiation effect of adjacent pixels adj_mul Add together to get the total bidirectional reflectivity factor BRF of each target pixel affected by the neighboring pixels adj .
2. The method according to claim 1, characterized in that In step 2), according to BRF self_sgl =K g ·r g +K l ·r l Calculate the single scattering bidirectional reflectivity factor BRF self_sgl ; Among them, K g represents the proportion of the illuminated soil area, K l represents the proportion of leaf area exposed to light, r g represents the reflectivity of the soil under illumination, r l represents the reflectivity of the illuminated leaves; 21) Calculate the proportion of illuminated soil area K through the path length distribution model g , the proportion of the illuminated soil area K g The large pores between the tree crowns g1 , small pores in the shadow ground K g2 , the soil spot K observed through the pores in the tree canopy g3 Composition; among them, K g1 =P(l(Ω s =0),l(Ω v =0)) represents the path length l(Ω) in the direction of solar incidence v ) is 0, the path length in the sensor observation direction is l(Ω s ) is 0; K g2 =P(l(Ω s >0),l(Ω v =0))·P(Ω s |l(Ω s >0),l(Ω v =0)), P(l(Ω s >0),l(Ω v =0)) represents the sensor The path length l(Ω v ) is 0 and the path length in the direction of solar incidence is l(Ω s ) is greater than 0, P(Ω s |l(Ω s >0),l(Ω v =0)) refers to the porosity obtained by using the path length distribution based on the area ratio of the shadow soil. Where G(Ω s ) represents the projection ratio of the leaf in the direction of solar incidence, FAVD is the leaf area density, l g2 (Ω s ) represents the path length of the shadowed soil portion in the direction of solar incidence, p l_g2 (Ω s ) is the path length l g2 (Ω s ), the probability density distribution of max , l min Represent the path length l g2 (Ω s ) maximum and minimum values; K g3 =P(l(Ω v >0))·P(Ω v |l(Ω v >0)), P(l(Ω v >0))·P(Ω v |l(Ω v >0)) is the path length l(Ω) in the observation direction v ) is greater than 0, l g3 (Ω v ) represents the path length of the canopy in the vertical observation direction, p l_g3 (Ω v ) is the path length l g3 (Ω v ), G(Ω v ) represents the projection ratio of the blade in the vertical observation direction; 22) Proportion of illuminated leaf area K l Bidirectional reflectivity factor BRF l =BRF lc +BRF lt ; By calculating the area proportion K of the illuminated tree crown c and the area ratio of the shaded tree crown K t , the proportion of the illuminated crown area K c Converted to bidirectional reflectance factor BRF of illuminated leaves lc , the area ratio K of the shadow crown t Converted to bidirectional reflectance factor BRF of illuminated leaves lt ; Among them, the proportion of illuminated crown area K c =P(l(Ω s =0),l(Ω v >0))·[1-P(Ω v |l(Ω s =0),l(Ω v >0))], the area ratio of the shadow crown K t =P(l(Ω s >0),l(Ω v >0))·[1-P(Ω v |l(Ω s >0),l(Ω v >0))]; P(l(Ω s =0),l(Ω v >0)) refers to the joint probability that the path length of the part in the observation direction is greater than 0 and the path length in the direction of solar incidence is 0.
3. The method according to claim 1, characterized in that In step 3), the shading factor S of the illumination canopy component is first calculated c and the shading factor S of the soil component of the illumination g Then, the new illumination crown component K considering the influence of neighboring pixel radiation is derived c ′ and the new light soil component K g ′, and then calculate the first scattering bidirectional reflectivity factor BRF of each target pixel affected by the adjacent pixels adj_sgl .
4. The method according to claim 3, characterized in that The first scattering bidirectional reflectivity factor BRF of each target pixel affected by the adjacent pixels is calculated adj_sgl The specific method is: 31) Calculate the proportion of large pores P formed by the light canopy being blocked by adjacent pixels c (l=0,Ω s )=P(F c (l=0,Ω s )|K c ), the proportion of small pores formed by being blocked by adjacent pixels Then calculate the shading factor S on the light canopy c =[1-P c (l=0,Ω s )]·[1-P c (l>0,Ω s )]; then calculate the target pixel illumination crown component area ratio K after being affected by the radiation of the neighboring pixels c ′=K c ·(1-S c );l(Ω s |K c ) represents the path length of the target pixel in the direction of the sun’s incidence after the original illuminated canopy is affected by the neighboring pixels, p l (Ω s |K c ) represents the path length l(Ω s |K c )’s probability density distribution; 32) Calculate the proportion of macropores P formed on the illuminated soil after being blocked by adjacent pixels g (l=0,Ω s )=P(F g (l=0,Ω s )|K g ), the proportion of small pores formed by being blocked by adjacent pixels and the shading factor S on the light soil g =[1-P g (l=0,Ω s )]·[1-P g (l>0,Ω s )], and then calculate the target pixel illumination soil component area ratio K after being affected by the radiation of the neighboring pixels g ′=K g ·(1-S g );l(Ω s |K g ) represents the path length of the target pixel in the direction of the sun’s incidence after the original part of the target pixel that illuminates the soil is affected by the neighboring pixels. l (Ω s |K g ) represents the path length l(Ω s |K g )’s probability density distribution; 33) Calculate the first scattering bidirectional reflectivity factor BRF adj_sgl =(1-β)·(K′ g ·r s +BRF l ′); where β is the proportion of sky scattered light.
5. The method according to claim 1, 2 or 3, characterized in that: In step 4), firstly, according to the proportion of the illuminated canopy area K c , shadow crown area ratio K t And the vertical observation path length distribution F(l>0,Ω v ), calculated Then, based on the probability p of a photon scattering from a leaf and colliding with a leaf in the canopy again, 1 , multiple re-collision probability p m , calculate the bidirectional reflectance factor BRF by ignoring the soil and only considering the multiple scattering between the canopy layers self_vmul And consider the multiple scattering between soil and canopy interior bidirectional reflectivity factor BRF self_smul ; Then the canopy multiple scattering bidirectional reflectivity factor is calculated when only the target pixel is considered BRF self_mul =BRF self_vmul +BRF self_smul 。 6. The method according to claim 5, characterized in that In step 5), the multiple scattering bidirectional reflectivity factor BRF of each target pixel affected by the adjacent pixels is calculated adj_mul The method is as follows: First, according to the proportion of light crown area K c , shadow crown area ratio K t And the vertical observation path length distribution F(l>0,Ω v ), calculated Then calculate the bidirectional reflectivity factor of multiple scattering of neighboring pixels to the target pixel Among them, LAI i Represents the leaf area index of the surrounding i-th pixel, LAI self Represents the leaf area index of the current target pixel, BRF mul_i represents the multiple scattering BRF calculated for the surrounding i-th pixel when only internal multiple scattering is considered, W i is the normalized weight matrix related to the distance; then the multiple scattering bidirectional reflectivity factor BRF is calculated adj_mul =BRF self_mul +BRF others_mul .
7. The method according to claim 5, characterized in that p 1 =0.7exp(k1·LAI)-0.66exp(k2·LAI), Where, k1 = 0.0045exp(θ1, k2 = 0.1982lncosθ-0.7146, i(θ) is the interception probability of the canopy in the θ direction, and θ is the solar zenith angle.
8. The method according to claim 7, characterized in that Where i0 = β·i D +(1-β)·i S , represents the average interception probability of the canopy for direct sunlight and scattered light from the sky, ω is the primary scattering albedo of the leaves, r s represents soil reflectivity, R dn represents the canopy bottom albedo, T dn and T up represents the canopy downward and upward transmittance, p(Ω s ) and p(Ω v ) represent the porosity in the solar incidence direction and the sensor observation direction, respectively.
9. A server, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program, the computer program is configured to be executed by the processor, and the computer program comprises instructions for executing each step in the method according to any one of claims 1 to 8.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.
Citation Information
Patent Citations
Aciculignosa canopy reflectivity calculation method and model
CN106874621A
Plant canopy three-dimensional radiation transmission analog method based on Lidar data
CN108196266A
Method for quantitatively describing influence of blade mirror reflection on simulation of canopy reflectivity
CN108896514A
Geometric optics-radiation transfer hybrid modeling method for row sowing aquatic vegetation canopy reflection
CN112784416A
Lidar system and method
WO2002065155A1
Cited By
Double-layer vegetation canopy full-wave band reflectivity radiation transmission calculation method and double-layer vegetation canopy full-wave band reflectivity radiation transmission calculation system
CN120847005A
Infrared radiation characteristic real-time simulation method for complex scene, medium and equipment
CN122221539A
System for inverting forest canopy physiological parameters through canopy photon re-collision probability
CN122336565A
System for retrieving forest canopy physiological parameters by canopy photon re-collision probability
CN122336565B
Unmanned aerial vehicle LAI inversion method and device based on high-precision canopy positioning
CN122415698A
