A rice growth process three-dimensional scene construction method based on a crop growth model and a rendering engine

Abstract

The present application relates to a kind of rice growth process three-dimensional scene construction method based on crop growth model and rendering engine, steps are as follows: through field experiment, plant density, distribution, leaf angle distribution and leaf spectrum, canopy spectrum and substrate reflectivity are obtained, and single rice 3D model is constructed;Coupling crop growth model, the canopy structure parameter and leaf BRDF are used to describe the reflection characteristics of vegetation canopy, and the canopy reflection factor is simulated;According to water biological optics model and wave water surface reflection model, water scattering phase function and water surface BRDF are calculated respectively;Three-dimensional water volume model is generated by AVRM model;By using Blender software, LuxCoreRender rendering engine is coupled into incident radiation, vegetation canopy reflection factor, water optical property and other related input parameters, and real rice scene is constructed.The method establishes high-precision three-dimensional rice scene model, improves the precision and efficiency of rice remote sensing radiation modeling, and further provides important technical support for agricultural remote sensing and wetland monitoring, with popularization and application prospect.

Description

(I) Field of the Invention

[0001] The present application relates to a rice growth process three-dimensional scene construction method based on a crop growth model and a rendering engine, belonging to the field of optical remote sensing, and is of great significance in wetland ecological research and crop quantitative monitoring applications. (II) BACKGROUND

[0002] Remote sensing monitoring plays a crucial role in precision agriculture. It not only enables long-term monitoring and crop yield estimation, but also assists in agricultural disaster monitoring and early warning, and aids in decision-making for precision management, thereby promoting the modernization of agriculture. However, remote sensing information reflects instantaneous physical conditions and cannot reveal the internal mechanisms of crop growth and development. Crop growth models, on the other hand, mathematically describe a series of physiological and biochemical processes during crop growth, development, and yield formation, enabling dynamic simulation of crop growth and development. Coupling remote sensing information with crop growth models can leverage the macro-monitoring capabilities of remote sensing technology and the mechanism simulation capabilities of crop growth models, thereby improving the accuracy of crop monitoring and prediction.

[0003] Three-dimensional radiative transfer models can simulate the three-dimensional light field distribution within crop canopies, providing a more accurate understanding of the absorption, reflection, and transmission of light by crop leaves. This is of great significance for studying physiological processes such as photosynthesis and light use efficiency. In this context, three-dimensional radiative transfer models have become a core tool for accurately inverting remote sensing data due to their outstanding performance in simulating complex heterogeneous remote sensing scenarios. The simulation of three-dimensional scenes not only captures subtle differences in the structure of different types of plants but also reflects the effects of multiple scattering and component proportion changes caused by specific regional and vegetation structures. This simulation approach is more realistic and provides more accurate descriptions of reflectance characteristics than traditional models, effectively supporting the monitoring and management of rice. Through three-dimensional radiative transfer models, researchers can recreate complex natural scenes in a virtual environment, thereby better understanding the ecological information contained in remote sensing data. This simulation method significantly improves the accuracy of remote sensing monitoring and provides a solid foundation for scientific research and management decisions.

[0004] The reflectance characteristics of objects are typically accurately described using the Bidirectional Reflectance Distribution Function (BRDF). The present application couples a plant leaf spectral model, a rice canopy three-dimensional radiative transfer model, a wave water surface reflection model, a shallow water biological optics model, and a physical rendering engine to construct a rice growth process three-dimensional scene construction method based on a crop growth model and a rendering engine, thereby improving the accuracy and efficiency of rice remote sensing modeling and providing important theoretical and practical references for agricultural remote sensing monitoring and ecological protection. (III)SUMMARY

[0005] The application relates to a rice growth process three-dimensional scene construction method based on a crop growth model and a rendering engine, and the steps are as follows: a single rice 3D model is constructed by obtaining plant density, distribution, leaf inclination angle distribution, leaf spectrum, canopy spectrum and substrate reflectivity through field experiments; a crop growth model is coupled, canopy structure parameters and leaf BRDF are used to describe the vegetation canopy reflection characteristics to simulate the canopy reflection factor; water body scattering phase functions and water surface BRDF are calculated according to a water body bio-optical model and a wave water surface reflection model respectively; a three-dimensional water volume model is generated through an AVRM model; and a real rice scene is constructed by coupling input parameters such as incident radiation, vegetation canopy reflection factor and water optical properties by using Blender software combined with a LuxCoreRender rendering engine, and the specific steps are as follows:

[0006] (1) Based on field experiment measurement, the statistical characteristics of the plant density, two-dimensional spatial distribution and leaf inclination angle distribution of rice, and the leaf reflectance spectrum, canopy reflectance spectrum and substrate reflectivity are obtained, a single rice 3D model is constructed, a crop growth model is further coupled, canopy structure parameters and leaf BRDF are used to describe the reflection characteristics of the vegetation canopy at different growth stages, and the reflection factor of the vegetation canopy is simulated;

[0007] (2) Water body scattering phase functions and water surface BRDF are calculated according to a water body bio-optical model and a wave water surface reflection model respectively, the water body scattering phase functions and the water surface BRDF are input into a water vegetation radiance model AVRM, and a complete three-dimensional water volume model is generated by comprehensively considering water optical parameters and observation geometric parameters;

[0008] (3) All input parameters, including incident radiation, vegetation canopy reflection factor, leaf BRDF, water body scattering phase function and water surface BRDF, are coupled by using Blender software combined with a LuxCoreRender rendering engine to construct a real rice vegetation scene under different canopy structures and observation geometries.

[0009] 5 The rice growth process three-dimensional scene construction method based on a crop growth model and a rendering engine according to claim 1

[0010] Method, characterized in that: in step (1), the "based on field experiment measurement, the statistical characteristics of the plant density, two-dimensional spatial distribution and leaf inclination angle distribution of rice, and the leaf reflectance spectrum, canopy reflectance spectrum and substrate reflectivity are obtained, a single rice 3D model is constructed, a crop growth model is further coupled, canopy structure parameters and leaf BRDF are used to describe the reflection characteristics of the vegetation canopy at different growth stages, and the reflection factor of the vegetation canopy is simulated"; the specific construction method is as follows:

[0011] Through field experiments, the morphological parameters of the plants, including plant height, leaf area index and leaf inclination angle distribution, are measured, a plurality of small facets are used to constitute the curved surface of the leaf, the iterative process of using finite elements to approximate and replace the smooth curved surface is used to simulate the real 3D morphology of the leaf; according to the measured plant height, leaf area index and leaf inclination angle distribution, the geometric structure model of the rice plant is established, the leaf spectral data calculated by the leaf radiation transfer model is assigned to the leaves of the geometric model, so that the three-dimensional model has real optical properties, and a single rice 3D model is constructed;

[0012] A three-dimensional structure of rice is generated through a rice spatial structure growth model based on measured data, including leaf length, node position, leaf angle and tillering angle information, while considering the influence of light period on growth; the spatial position of rice in the scene, the growth stage of interest and other parameters are specified, the three-dimensional structure model of the field rice is established by using a rendering software in combination with the single rice three-dimensional structure model; according to the plant density data, the spatial distribution mode of the single rice model in the whole field is set, and then the reflectance factor of the vegetation canopy is calculated; through simulation of the canopy reflection under different illumination and observation geometries, high-precision rice canopy reflectance spectrum is obtained.

[0013] 6. The method of claim 1, wherein the method further comprises: generating a three-dimensional scene of the rice growth process based on the crop growth model and the rendering engine.

[0014] The method is characterized in that: in step (2), the "water scattering phase function and water surface BRDF are calculated according to the water bio-optical model and the wave water surface reflection model respectively, the water scattering phase function and the water surface BRDF are input into the AVRM, and a complete three-dimensional water volume model is generated by integrating the water optical parameters and the observation geometry parameters", and the specific method is as follows:

[0015] First step: calculating the water scattering phase function according to the water bio-optical model: in the rice field scene, the scattering characteristics of the water volume element are anisotropic, the water body is regarded as a volume scattering medium, and therefore three main attributes need to be considered: absorption coefficient, scattering coefficient and scattering phase function; the first problem to be solved is the parameterization of the absorption coefficient and the scattering coefficient; the water bio-optical model calculates the absorption and scattering coefficients by considering the parameters of chlorophyll, suspended solids and CDOM (colored dissolved organic matter) in the water body; this method splits the absorption and scattering coefficients of the water body according to the contribution of different components, and the specific calculation method is as follows:

[0016]

[0017] Wherein, a pw , a ph , a NAP and a CDOMrespectively represent the absorption coefficients of pure water, phytoplankton, non-algal particulate matter and CDOM, b bw respectively represent the absorption coefficients of pure water, phytoplankton, non-algal particulate matter and CDOM, b p respectively represent the scattering coefficients of pure water and particulate matter, a NAP The calculation method of the term is as follows:

[0018] a NAP (λ) = a NAP (443) · 0.75e -0.0123(λ-443)

[0019] In the formula, the absorption coefficient of non-algal particulate matter at 443 nm is used as the reference absorption coefficient, where λ is the wavelength to be solved, and a NAP (443) and the concentration of suspended particulate matter SPM present a linear relationship, which satisfies a NAP (443) = 0.041m 2 · g -1 · SPM, and the absorption coefficient contributed by the CDOM component is:

[0020] a CDOM (λ) = a CDOM (375) e -0.0192(λ-375 )

[0021] Wherein, the value of a CDOM (375) is 1.25m -1 , and the unit absorption coefficient of phytoplankton is:

[0022]

[0023] Wherein, C is the concentration of chlorophyll a, and the scattering coefficient of pure water is represented as:

[0024]

[0025] The scattering coefficient of particulate matter is represented as:

[0026] b p (λ) = EC 0.62 550 / λ

[0027] In the above formula, C is the concentration of chlorophyll a, and E is an empirical coefficient, which is taken as 1.0 in the model;

[0028] Second step: calculate the water surface BRDF according to the Cox-Munk model: the Cox-Munk model regards the water surface as the geometry of different slope micro-facet, and calculates the BRDF of the wave water surface as:

[0029]

[0030] Where r(ω) is the Fresnel reflectivity at an incident angle of ω when the water surface is calm, and p(z x , z y ) is the probability distribution function of the wave surface slope, z x , z y is the small surface element coordinate system, θ n is the zenith angle direction of the normal line of the wave element where the mirror reflection occurs, θ s is the solar zenith angle, θ o is the observed zenith angle, σ is the water surface roughness factor to describe the degree of water surface fluctuation, and replaces the wind speed W as the input of the water surface reflectivity model;

[0031] Step 3: Input the water scattering phase function and the water surface BRDF into the AVRM model, and integrate the optical parameters and observation geometry parameters related to the water body to generate a complete three-dimensional water volume model. The specific calculation process is as follows:

[0032] In the absence of inelastic scattering, the parameters used to calculate the radiative transfer through a volume medium are the spectral absorption coefficient a(λ, p) and the spectral volume scattering function VSFβ(λ, p, ψ), where λ is the wavelength and p = (p x , p y , p z ) is a point in space, ψ is the scattering angle, 0≤ψ≤π, and the scattering coefficient b(λ, p) and attenuation coefficient c(λ, p) are derived from the absorption coefficient and VSF:

[0033]

[0034] c(λ,p)=a(λ,p)+b(λ,p)

[0035] The radiation distribution L(λ, p, v) at a point p in space is scattered by the VSF, which in turn produces the path radiation at that point:

[0036]

[0037] Among them, dΩ(v in ) is the same as v in The associated solid angle, integrated over it, Considering the radiation energy emitted from the inside of the medium, the radiation distribution L(λ, p, v) at point p is the radiation energy emitted from point p. S The sum of the received path attenuation radiation, the point p S The energy at the surface of the first voxel encountered when tracing backward from the direction v, plus any radiant energy received during the intervening path from the medium scattered or emitted from the volume:

[0038]

[0039] where L EX-G (λ, p S , v) denotes the outgoing radiance at point p S on the surface in the global coordinate system in the direction of the vector v, and the function K(λ, p1, p2) denotes the transmittance of the path radiance from point p1 to point p2 in the medium:

[0040]

[0041] where d = |p1-p2| is the distance between points p1 and p2, it is important to note that even if the first intersecting surface is transparent, the reverse path along the vector v does not need to be considered, since the influence of the volume surface distal radiance source can be considered as outgoing radiance of the surface itself, the volume discretization is initiated by subdividing the model domain into cubic voxels, where the viewpoint of each voxel is located at the voxel center, due to the cubic shape of the voxels, the incoming radiance at the voxel center has a voxel path length x(q) that depends on q, the distance x(q) can be numerically precomputed beforehand on all voxels of a given size by averaging over the solid angle, if the emission is zero, the source function based on these path integrals has the form:

[0042]

[0043] where

[0044]

[0045] where L'(λ, i, q in ) is the radiance distribution from the external source at the voxel center, i.e. not including the scattered radiance from the voxel itself, while the radiance at the center point is extrapolated forward and backward to the voxel boundaries and integrated along the path length:

[0046]

[0047] This equation accounts for internal scattering of the incoming radiance direction, while the current voxel path radiance is integrated and attenuated along the path in the voxel, since the voxel average VSF, β(λ, i, q in , q out ) is assumed to be constant within the voxel, the path integral can be analytically derived:

[0048]

[0049] For a given beam attenuation coefficient c(λ, i) and voxel size x(qi n ), in many model applications the range of voxel sizes and beam attenuation coefficient values will be very limited, it is important to note that while L *is cyclic, but the solution algorithm is iterative, so L * is updated from the current value at each iteration.

[0050] 7. The method according to claim 1, wherein the step (3) of constructing a real rice vegetation scene under different canopy structures and observation geometries by using Blender software coupled with LuxCoreRender rendering engine to couple all input parameters, including incident radiation, vegetation canopy albedo, leaf BRDF, water scattering phase function and water surface BRDF, is implemented as follows:

[0051] LuxCoreRender uses physical principles to calculate the interaction between light and material, adopts path tracing as its ray tracing integral algorithm, in the rendering process, light is emitted from the camera into the scene, and then interacts with the objects in the scene until the light source is found or absorbed; when establishing a three-dimensional rice scene, 128 rays are generated for each camera pixel, including direct scattering and diffuse scattering, the rendering results of single scattering and multiple scattering are obtained by setting the maximum bounce number of the light path in the engine, assuming that the leaf and rhizome are Lambertian surfaces with the same reflectivity, and the rhizome has no transmittance, the soil is set to be horizontal and flat, and the water body is set as a volume body in Blender software according to the water voxel parameters in AVRM; in order to simulate a finite size of the sun, the "sunlight" configuration is adopted, a perspective camera with a 5° field of view is selected, when the camera points to the center of the scene from a distance of 50m, the area of the 2m×2m experimental plot can be captured; at the same time, in order to avoid potential boundary effects, 8 times of replication are carried out around the central scene, by specifying the spatial position of rice in the scene, the growth stage of interest and other parameters, combined with the single rice model, the three-dimensional model of rice is established by using Blender software, the dynamic of the 3D rice growth model canopy is driven by the growth days, from the two-leaf stage to the beginning of male heading, six growth stages are simulated per 100℃d, by changing the six parameters of the model except the driving parameters, 36 rice canopy models at different stages can be simulated.

[0052] Compared with the prior art, the application has the advantages of:

[0053] (1) In the field of vegetation remote sensing, remote sensing modeling research currently lacks knowledge of the spatiotemporal variations of key ground parameters, which limits the ability to simulate continuous time phases. For example, in large-scale crop planting areas in China, the plant structure and biochemical parameters change significantly due to the crop growth process, thereby affecting the spatial distribution of crops. In order to achieve continuous time phase data simulation, the present invention introduces crop growth process knowledge and models, couples remote sensing information with crop growth models, and can fully utilize the macro-monitoring capabilities of remote sensing technology and the mechanism simulation capabilities of crop growth models. This method not only improves the understanding of the crop growth process, but also enhances the time phase continuous simulation capabilities of remote sensing models, which can more accurately reflect the actual situation, thereby improving the accuracy of crop monitoring and prediction.

[0054] (2) The present invention uses the aquatic vegetation canopy three-dimensional radiation transfer model AVRM and the rendering engine LuxCoreRender as the technical framework, and tightly couples and integrates the water surface reflection and transmission process, the absorption and scattering process inside the vegetation canopy, the absorption and scattering process of water components, and the bottom reflection process into a model, constructs real rice scenes under different canopy structures and observation geometries, and calculates the vegetation canopy reflectance. The model of the present invention has clear physical concepts, strong versatility, and reliable accuracy. (IV) Description of the accompanying drawings

[0055] Figure 1 The technical process of the present invention; Figure 2 Create a schematic diagram for the 3D blade; Figure 3 This is a single vegetation model in Blender; Figure 4 Rendering a scene for a 3D aquatic vegetation canopy. (V) Specific implementation methods

[0056] To better illustrate the method for constructing a three-dimensional scene of rice growth process based on a crop growth model and a rendering engine, the scene established by the present invention was tested and analyzed, and good results were achieved. The specific implementation method is as follows:

[0057] (1) Based on the field experimental measurements of rice plant density, two-dimensional spatial distribution, and leaf inclination distribution statistical characteristics, as well as leaf reflectance spectrum, canopy reflectance spectrum, and substrate reflectance, a single plant 3D model was constructed. At the same time, the reflectance characteristics of the vegetation canopy were described by canopy structure parameters and leaf BRDF, and the reflectance factor of the vegetation canopy was simulated.

[0058] (2) The water body scattering phase function and the water surface BRDF are calculated according to the water body bio-optical model number and the Cox-Munk model respectively. The water body scattering phase function and the water surface BRDF are input into the Aquatic Vegetation Radiosity Model (AVRM model), which integrates the basic water body parameters and observation geometric parameters to form a complete three-dimensional water body volume model;

[0059] (3) Blender software was used in combination with the LuxCoreRender rendering engine to couple all input parameters, including incident radiation, vegetation canopy reflectance factor, leaf BRDF, water scattering phase function, and water surface BRDF, to construct realistic rice vegetation scenes under different canopy structures and observation geometries.

[0060] Figure 2 The 3D morphological modeling process of the leaf is shown. The leaf is naturally curved, so when maintaining the basic shape of the leaf, multiple small facets are needed to form the surface of the leaf. This is a process of using finite elements to approximate or replace smooth surfaces, such as Figure 2 -a, which can simulate a more realistic 3D shape of leaves. Figure 2 In -b, the green curve represents the side edge of a leaf that curves naturally. As the number of facets is gradually reduced while creating the leaf 3D object, the curvature of the leaf's side edge becomes less smooth, transitioning from a green curve to a red, segmented curve and finally to a black, segmented curve. This trend also applies to progressively simpler 3D objects, including the most simplified.

[0061] Figure 3 Shows a single vegetation model in Blender, and Figure 4 Shows the final rice scene rendered by LuxCoreRender.

[0062] This method, which creates a realistic rice scene, facilitates a deeper and more scientific exploration of the relationship between the rice canopy reflectance spectrum and solar radiation, vegetation structural parameters, water body components, and bottom reflectance characteristics. It also significantly enhances the spatiotemporal continuous simulation capabilities of remote sensing modeling, significantly improving the accuracy of wetland vegetation reflectance simulations. Furthermore, the use of an advanced three-dimensional radiation transfer model and rendering engine enables highly accurate rice scene modeling, ensuring the model's versatility and reliability.

Claims

1. A method for constructing a three-dimensional scene of rice growth process based on a crop growth model and a rendering engine, characterized in that The following steps are involved: (1) Based on the statistical characteristics of rice plant density, two-dimensional spatial distribution, and leaf inclination angle distribution obtained from field experiments, as well as leaf reflectance spectrum, canopy reflectance spectrum, and substrate reflectance, a single rice plant 3D model was constructed. Furthermore, combined with the crop growth model, the reflectance characteristics of the vegetation canopy at different growth stages were described by canopy structural parameters and the leaf bidirectional reflectance distribution function (BRDF), and the reflectance factor of the vegetation canopy was simulated. (2) Based on the water body bio-optical model number and the wave surface reflection model, the water body scattering phase function and the water surface bidirectional reflectance distribution function (BRDF) are calculated respectively. The water body scattering phase function and the water surface BRDF are input into the aquatic vegetation radiosity model (AVRM). The water body optical parameters and observation geometric parameters are integrated to generate a complete three-dimensional water volume model. The specific method is as follows: Step 1: Calculate the water scattering phase function based on the water bio-optical model. In the rice field scenario, since the scattering characteristics of water voxels are anisotropic, water is considered a volume scattering medium. Therefore, we need to focus on its three main properties: absorption coefficient, scattering coefficient, and scattering phase function. The first problem to be solved is the parameterization of the absorption and scattering coefficients. The water bio-optical model calculates the absorption and scattering coefficients by considering parameters such as chlorophyll, suspended matter, and colored dissolved organic matter (CDOM). This method splits the absorption and scattering coefficients of water according to the contributions of different components. The specific calculation method is as follows: Among them, a pw 、a ph 、a NAP with a CDOM are the absorption coefficients of pure water, phytoplankton, non-algae particles and CDOM, respectively, and b bw with b p denote the scattering coefficients of pure water and particulate matter, respectively, a NAP The calculation method of the term is: the NAP (λ)=a NAP (443)·0.75e -0.0123(λ-443) In this formula, the absorption coefficient of non-algae particles at 443 nm is used as the reference absorption coefficient, where λ is the wavelength to be solved, and a NAP (443) shows a linear relationship with the concentration of suspended particles SPM, satisfying a NAP (443) = 0.041m 2 ·g -1 For SPM, the absorption coefficient contributed by the CDOM component is: a CDOM (λ)=a CDOM (375)e -0.0192(λ-375) Among them, a CDOM The value of (375) is 1.25m -1 , the unit absorption coefficient of phytoplankton is: Where C is a parameter related to the concentration of chlorophyll a, A(λ), and B(λ). The scattering coefficient of pure water is expressed as: The scattering coefficient of particles is expressed as: b p (λ)=EC 0.62 550 / λ In the above formula, C is the concentration of chlorophyll a, and E is an empirical coefficient, which is taken as 1.0 in the model; Step 2: Calculate the water surface BRDF according to the Cox-Munk model: The Cox-Munk model regards the water surface as a geometry of small surface elements with different slopes, and calculates the BRDF of the wavy water surface as: Where r(ω) is the Fresnel reflectivity at an incident angle of ω when the water surface is calm, and p(z x ,z y ) is the probability distribution function of the wave surface slope, z x ,z y is the small surface element coordinate system, θ n is the zenith angle direction of the normal line of the wave element where the mirror reflection occurs, θ s is the solar zenith angle, θ o is the observed zenith angle, σ is the water surface roughness factor to describe the degree of water surface fluctuation, and replaces the wind speed W as the input of the water surface reflectivity model; Step 3: Input the water scattering phase function and the water surface BRDF into the AVRM model, and integrate the optical parameters and observation geometry parameters related to the water body to generate a complete three-dimensional water volume model. The specific calculation process is as follows: In the absence of inelastic scattering, the parameters used to calculate the radiative transfer through a volume medium are the spectral absorption coefficient a(λ,p) and the spectral volume scattering function VSFβ(λ,p,ψ), where λ is the wavelength and p = (p x ,p y ,p z ) is a point in space, ψ is the scattering angle, 0≤ψ≤π, and the scattering coefficient b(λ,p) and attenuation coefficient c(λ,p) are derived from the absorption coefficient and VSF: c(λ,p)=a(λ,p)+b(λ,p) The radiation distribution L(λ,p,v) at a point p in space is scattered by the VSF, which in turn produces the path radiation at that point: Among them, dΩ(v in ) is the same as v in The associated solid angle, integrated over it, Considering the radiation energy emitted from the inside of the medium, the radiation distribution L(λ,p,v) at point p is the radiation energy from point p. S The sum of the received path attenuation radiation, the point p S The energy at the surface of the first voxel encountered when tracing backward from the direction v, plus any radiant energy received during the intervening path from the medium scattered or emitted from the volume: Among them, L EX-G (λ,p S ,v) represents the position at point p S The outgoing radiation of the direction vector v of the surface in the global coordinate system, the function K(λ,p1,p2) represents the transmittance of the path radiation transmitted from point p1 to point p2 in the medium: Where d = |p1-p2| is the distance between points p1 and p2. Note that even if the first intersection surface is transparent, there is no need to consider the reverse path along the vector v, because the influence of the radiation source on the far side of the volume surface can be regarded as the outgoing radiation of the surface itself. The volume discretization is started by subdividing the model domain into cubic voxels, where the viewpoint of each voxel is located at the center of the voxel. Since the voxel has a cubic shape, the incident radiation at the center of the voxel has a voxel path length x(q) that depends on q. The distance x(q) can be numerically pre-calculated on all voxels of a given size by taking the solid angle mean. If the emission is zero, the source function based on these path integrals is as follows: in Where, L′(λ,i,q in ) is the radiation distribution at the center of the voxel from the external source, that is, excluding the scattered radiation from the voxel itself, while the radiation at the center point is extrapolated forward and backward to the voxel boundary and integrated along the path length: This formula handles the internal scattering in the direction of the incident radiation, and integrates and attenuates the radiation of the current voxel path along the path in the voxel. Since the voxel average VSF is assumed, β(λ,i,q in ,q out ), is constant within the voxel, and the path integral can be derived analytically: For a given beam attenuation coefficient c(λ,i) and voxel size x(q in ), in many model applications, the range of voxel size and beam attenuation coefficient values ​​will be very limited. It should be noted that although L * is cyclic, but the solution algorithm is iterative, so L * It is updated from the current value in each iteration; (3) Blender software was used in combination with the LuxCoreRender rendering engine to couple all input parameters, including incident radiation, vegetation canopy reflectance factor, leaf BRDF, water scattering phase function, and water surface BRDF, to construct realistic rice vegetation scenes under different canopy structures and observation geometries.

2. The method for constructing a three-dimensional scene of rice growth process based on a crop growth model and a rendering engine according to claim 1, characterized in that: The specific requirements for step (1) are as follows: "Based on the field experimental measurements of rice plant density, two-dimensional spatial distribution, and statistical characteristics of leaf inclination angle distribution, as well as leaf reflectance spectrum, canopy reflectance spectrum, and substrate reflectance, a single rice plant 3D model is constructed. Further, combined with the crop growth model, the reflectance characteristics of the vegetation canopy at different growth stages are described by canopy structure parameters and leaf BRDF, and the reflectance factor of the vegetation canopy is simulated." Through field experiments, plant morphological parameters, including plant height, leaf area index, and leaf inclination angle distribution, were measured. Multiple small facets were used to construct the leaf surface, and finite elements were used to approximate and replace the smooth surface in an iterative process to simulate the true 3D leaf morphology. Based on the measured plant height, leaf area index, and leaf inclination angle distribution, a geometric structural model of the rice plant was established. Leaf spectral data calculated using a leaf radiation transfer model was assigned to the leaves of the geometric model, ensuring that the three-dimensional model possessed realistic optical properties, thus constructing a 3D model of a single rice plant. The three-dimensional structure of rice is generated through a rice spatial structure growth model based on measured data, including information such as leaf length, node position, leaf angle and tiller angle, while taking into account the impact of the light cycle on growth; the spatial position of rice in the scene, the growth stage of interest and other parameters are specified, and combined with the three-dimensional structure model of a single rice plant, a three-dimensional structure model of rice in the field is established using rendering software; according to the plant density data, the spatial distribution pattern of the single rice plant model in the entire field is set, and the reflectance factor of the vegetation canopy is calculated; by simulating the canopy reflectance under different lighting and observation geometries, a high-precision rice canopy reflectance spectrum is obtained.

3. The method for constructing a three-dimensional scene of rice growth process based on a crop growth model and a rendering engine according to claim 1, characterized in that: The specific implementation method of step (3) of "using Blender software combined with the LuxCoreRender rendering engine to couple all input parameters, including incident radiation, vegetation canopy reflectance factor, leaf BRDF, water scattering phase function and water surface BRDF, to construct a realistic rice vegetation scene under different canopy structures and observation geometries" is as follows: LuxCoreRender uses physical principles to calculate the interaction between light and materials, and adopts path tracing as its ray tracing integration algorithm. During the rendering process, light is emitted from the camera into the scene, and then interacts with objects in the scene until it finds a light source or is absorbed; when establishing a three-dimensional rice scene, 128 rays are generated for each camera pixel, including direct scattering and diffuse scattering. By setting the maximum number of bounces of the light path in the engine, single scattering and multiple scattering rendering results are obtained, assuming that the leaves and rhizomes are both Lambertian surfaces with the same reflectivity. The soil was set to be horizontal and flat, and the water volume was set in Blender software based on the water voxel parameters in AVRM. To simulate a finite-sized solar source, a "sunlight" configuration was used. A perspective camera with a 5° field of view was selected. When the camera was pointed at the center of the scene from a distance of 50 m, it could capture a 2 m × 2 m area of ​​the experimental plot. To avoid potential boundary effects, eight replications were performed around the central scene. By specifying parameters such as the spatial position of the rice plant in the scene and the growth stage of interest, combined with a single rice plant model, a 3D rice model was constructed using Blender software. The dynamics of the 3D rice canopy were driven by the number of growing days. Six growth stages were simulated every 100°C / d, from the two-leaf stage to the male heading stage. By varying six model parameters other than the driving parameters, 36 rice canopy models at different stages were simulated.

Images