Chlorophyll hyperspectral inversion method based on curled plant leaf radiation transfer model

Abstract

The present invention discloses a chlorophyll hyperspectral inversion method based on a curled plant leaf radiation transmission model. The method is as follows: 1. For the curled leaf to be measured, the hemispherical reflectance spectrum DHR and the hemispherical transmission spectrum DHT are measured. The three-dimensional point cloud data of the curled leaf is collected. 2. The leaf structure parameters, leaf surface roughness, leaf average refractive index and leaf chlorophyll absorption coefficient of the curled leaf are obtained. 3. Based on the three-dimensional point cloud data, the spatial geometric model of the curled leaf is obtained. 4. The spatial geometric model of the curled leaf is spectrally simulated. 5. Chlorophyll concentration is inverted. The present invention takes into account the influence of the spatial morphology of the curled leaf on the multi-angle spectral simulation and the chlorophyll hyperspectral inversion results, improves the accuracy of chlorophyll inversion, and can obtain a high-precision canopy radiation transmission model based on this.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent perception and detection technology of agricultural ecological big data, and specifically relates to a chlorophyll hyperspectral inversion method based on a curled plant leaf radiation transmission model. Background Art

[0002] Photosynthesis of green plants is the most common and largest biochemical process on Earth. It plays a great role in the synthesis of organic matter, the conversion and storage of solar energy, the purification of the atmosphere, the maintenance of the dynamic balance of oxygen and carbon dioxide content in the atmosphere and the stability of the carbon cycle. Especially today, with the deepening process of globalization, plants in regional ecosystems play an important role in material (carbon) cycles and environmental early warning. As the basis of agricultural production, they are of great significance both in theory and practice.

[0003] Green plants carry out photosynthesis through their leaves. Crucial to photosynthesis are the various pigments within the leaves, particularly chlorophyll, which is closely linked to various ecological and physiological processes within the plant. Chlorophyll absorbs light in the 400-700nm wavelength range, while the high-energy wavelength of sunlight reaching the ground is 200-1100nm, occupying a large portion of the high-energy spectrum. Therefore, using hyperspectral remote sensing data for quantitative remote sensing of plant leaves has practical applications in monitoring the physiological and biochemical states of plants and studying their various biochemical components.

[0004] Leaf optical radiation transfer models simulate light transmission within leaves and infer the biochemical composition of plant leaves by quantitatively describing the factors influencing leaf optical properties. These models include the LEAFMOD (Leaf Experimental Absorptivity Feasibility) model (Ganapol et al. 1998), the FluorMOD (Chlorophyll Fluorescence Model) model (Zarco-Tejada et al. 2000), the PROSPECT (A Model of Leaf Optical Properties Spectra) model (Jacquemoud and Baret 1990), and the SLOP (Stochastic Model for Leaf Optical Properties) model (Maier et al. 1999). However, these leaf optical radiation transfer models all suffer from a problem. Because the spectral data used in these models is acquired using an integrating sphere, which typically has a single incident angle, the incident angle of the light source in these models is fixed. The latest model, the PROSEPCT-MA model proposed in patent CN111220552A, is a plant leaf radiation transfer model that takes light direction into account. The PROSPECT-MA model accounts for varying illumination directions but fails to factor leaf curl into the radiation transfer process. In natural settings, most plant leaves have a certain degree of natural curvature, and simplifying this with a flat leaf model inevitably introduces errors. Therefore, there is an urgent need to develop a plant leaf radiation transfer model that can account for both illumination direction and leaf curl to improve its accuracy. Summary of the Invention

[0005] The present invention aims to provide a method for constructing a plant leaf radiation transfer model that takes into account the influence of both light direction and leaf curling degree, and for performing chlorophyll inversion based on the model.

[0006] A chlorophyll hyperspectral inversion method based on a curled plant leaf radiation transfer model includes the following steps:

[0007] Step 1: Measure the hemispherical reflectance spectrum DHR and the hemispherical transmittance spectrum DHT of the curled blade to be tested. Collect the three-dimensional point cloud data of the curled blade.

[0008] Step 2: Obtain the leaf structure parameters N, leaf surface roughness σ, and leaf average refractive index of the curled leaf and leaf chlorophyll absorption coefficient K chls (λ).

[0009] Step 3: Based on the three-dimensional point cloud data, obtain the spatial geometric model of the curled leaf.

[0010] Step 4: Perform spectral simulation on the spatial geometric model of the curled leaf.

[0011] 4-1. Using the spatial geometric model of a curled leaf as the ray tracing scene, perform microplane segmentation to obtain multiple leaf microplanes.

[0012] 4-2. Take each blade microplane as the measured microplane in turn; perform ray tracing from the measured microplane to obtain the angle of light directly irradiated by the light source on the measured microplane, as well as the angle of light reflected from the other blade microplanes; for each angle of light, simulate the corresponding bidirectional reflectance distribution function (BRDF) through the blade reflectivity model to obtain the BRDF of the measured microplane under direct light, as well as the contribution of other blade microplanes to the BRDF of the measured microplane.

[0013] 4-3. After completing the traversal of all leaf microplanes, the bidirectional reflectance distribution function BRDF of all leaf microplanes is combined as the bidirectional reflectance distribution function BRDF of the curled leaf to obtain the functional relationship between the chlorophyll content of the curled leaf and the directional hemispherical reflection coefficient and directional hemispherical transmission coefficient at the wavelength corresponding to the maximum reflectance, maximum transmittance, and minimum absorptivity.

[0014] Step 5: Chlorophyll concentration inversion: Based on the functional relationship obtained in step 4 and the hemispherical reflectance spectrum DHR and hemispherical transmittance spectrum DHT measured in step 1, the chlorophyll concentration of the curled leaf is simulated.

[0015] Preferably, the blade reflectivity model uses the incident light source angle as one of the variables.

[0016] Preferably, the spatial geometric model obtained in step 3 is as follows:

[0017]

[0018]

[0019] Among them, (x, y, z) are the coordinates of any point on the curled leaf surface; (x0, y0, z0) are the coordinates of the center point of the curled leaf sample; a represents the characteristic parameter of the leaf curling degree; b1 and b2 represent the major axis and minor axis parameters of the curled leaf, respectively.

[0020] Preferably, the hemispherical reflectance spectrum DHR and the hemispherical transmittance spectrum DHT described in step 1 are measured by a UV-3600 spectrophotometer + integrating sphere.

[0021] Preferably, the blade structural parameter N is calculated based on the maximum reflectivity, maximum transmittance and minimum absorptivity within the 400-1000nm band using a spectral minimum fitting method. The blade structural parameter N is specifically as follows:

[0022]

[0023] Among them, λ′1, λ′2, and λ′3 are the wavelengths corresponding to the maximum spectral reflectance, maximum spectral transmittance, and minimum absorptivity of the curled blade respectively; DHR mea (·) and DHR mod (·) are the measured directional hemispherical reflection coefficient and the directional hemispherical reflection coefficient calculated by the model; DHT mea (·) and DHT mod (·) is divided into the measured directional hemispherical transmission coefficient and the directional hemispherical transmission coefficient calculated by the model.

[0024] Preferably, the blade surface roughness σ is obtained as follows:

[0025]

[0026] Wherein, λ is the wavelength in the 800-1400nm band; DHR mea (·) and DHR mod (·) are the measured directional hemispherical reflection coefficient and the directional hemispherical reflection coefficient calculated by the model; DHT mea (·) and DHT mod (·) is divided into the measured directional hemispherical transmission coefficient and the directional hemispherical transmission coefficient calculated by the model.

[0027] As a preference, the average refractive index of the blade and K chls (λ) is obtained as follows:

[0028]

[0029] Wherein, λ is the wavelength in the 500-800nm ​​band; DHR mea (·) and DHR mod (·) are the measured directional hemispherical reflection coefficient and the directional hemispherical reflection coefficient calculated by the model; DHT mea (·) and DHT mod (·) is divided into the measured directional hemispherical transmission coefficient and the directional hemispherical transmission coefficient calculated by the model.

[0030] The present invention has the following beneficial effects:

[0031] The present invention considers the influence of the spatial morphology of curled leaves on multi-angle spectral simulation and chlorophyll hyperspectral inversion results, improves the accuracy of chlorophyll inversion, and can obtain a high-precision canopy radiation transfer model based on this. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 This is a flow chart of the spectrum simulation in step 4 of the present invention.

[0033] Figure 2 This is a flowchart of ray tracing in step 4-2 of the present invention.

[0034] Figure 3 It is a simulation effect diagram of the blade surface curvature model and the blade boundary curve model used in step 3 of the present invention;

[0035] Figure 4 is a comparison diagram of the curled leaf BRDF calculated in step 4 of the present invention and the measured curled leaf BRDF;

[0036] Figure 5 It is a comparison diagram of the simulated spectrum curve and the measured spectrum curve obtained at different angles of the present invention;

[0037] Figure 6 This is the chlorophyll inversion result diagram for one of the curled leaves in the sample set at different light source zenith angles (15°, 30°, 40°, 55°);

[0038] Figure 7 This is a graph of the evaluation parameters (RMSE, IAS, PEC, V) of the chlorophyll inversion results for one of the curled leaves in the sample set at different light source zenith angles (15°, 30°, 40°, 55°). DETAILED DESCRIPTION

[0039] The present invention will be further described below with reference to the accompanying drawings.

[0040] A plant leaf radiation transfer model that considers the dual influence of light direction and leaf curling degree is constructed, and a chlorophyll inversion method based on the model is performed. The specific steps are as follows:

[0041] Step 1: Data acquisition and processing.

[0042] To ensure the freshness of the leaves and avoid changes in the properties and biochemical content of the detached leaves, the operation time for each leaf experiment should not exceed 1 hour.

[0043] 1-1. Data object selection.

[0044] The present invention is based on a curled leaf spatial geometry model, combined with a light source multi-angle flat leaf radiation transfer model PROSPECT-MA and a ray tracing algorithm to create a curled leaf multi-angle radiation transfer model that takes into account the dual factors of illumination direction and leaf curling degree. Therefore, different types of fresh plant leaves are selected as experimental samples according to the different degrees of curling of the plant leaves. The PROSPECT-MA model is specifically a leaf radiation transfer model that takes into account illumination direction, as described in publication number "CN111220552A." The ray tracing algorithm can simulate the entire process of light rays starting from a light source, intersecting with physical objects in a scene, undergoing multiple reflections and transmissions, and finally being collected by a detector.

[0045] Multi-angle spectral data and biochemical component content data of curled leaf samples were obtained to construct and verify a multi-angle radiation transfer model for curled leaves. Furthermore, since ray tracing algorithms are graphics algorithms in a 3D scene, real 3D point cloud data of the curled leaf samples was also required to reconstruct their 3D structure.

[0046] 1-2. Hemispherical spectral measurement of curled leaves

[0047] According to the SPADm value of the curled leaf sample, each curled leaf sample was collected three times according to five gradients, and its hemispherical reflectance spectrum (i.e., DHR) and hemispherical transmittance spectrum (i.e., DHT) were measured using a UV-3600 spectrometer + integrating sphere.

[0048] 1-3. Determination of biochemical components and surface roughness of curled leaves

[0049] The biochemical composition of curled leaves, including the determination of chlorophyll a and chlorophyll b, was measured using a 7500 visible light spectrophotometer. The corresponding leaf surface roughness was measured using a surface roughness meter (Ta200).

[0050] 1-4. Obtaining 3D Point Cloud Data of Curled Leaves

[0051] An in vitro leaf with only curled geometric features was selected as the object for point cloud data acquisition, and a three-dimensional laser scanner (model FastSCAN Cobra) was used to acquire three-dimensional point cloud data of the synthetic curled leaf.

[0052] Step 2: Selection of model parameters

[0053] 2-1. Obtaining model parameters related to blade samples

[0054] The blade structural parameter N was calculated using the spectral minimum fitting method in the three maximum bands (maximum reflectivity, maximum transmittance and minimum absorptivity bands) within the 400-1000 nm band.

[0055] Using the near-infrared range of the spectrum (800-1400nm), a sensitivity analysis was conducted on the blade surface roughness σ and the spectral reflectance and transmittance in the near-infrared band. The bands with higher sensitivity were screened out and the blade surface roughness was estimated through correlation analysis.

[0056] 2-2. Obtaining model parameters related to spectral bands

[0057] The spectral minimum fitting method was used in the entire spectral range to take the measured spectral characteristics (hemispherical reflectance spectrum and hemispherical transmittance spectrum), the content of corresponding leaf biochemical components and the obtained leaf structural parameters N and leaf surface roughness (σ) as model input variables to obtain the model parameters. and leaf chlorophyll absorption coefficient K chls (λ).

[0058] Using the spectral minimum fitting method, the blade structural parameter N in the dataset is calculated from the reflectance and transmission spectra of the ANGERS dataset in the three bands of maximum spectral reflectance, maximum spectral transmittance, and minimum absorbance. Then, the blade surface roughness σ in the dataset is calculated using the spectral minimum fitting method over the entire band:

[0059]

[0060]

[0061] Where χ(·) is the optimization objective function; λ′1, λ′2, and λ′3 are the wavelengths corresponding to the maximum spectral reflectance, maximum spectral transmittance, and minimum absorptivity of the curled blade, respectively; DHR mea (·) and DHR mod (·) are the measured directional hemispherical reflection coefficient and the directional hemispherical reflection coefficient calculated by the model; DHT mea (·) and DHT mod (·) is divided into the measured directional hemispherical transmission coefficient and the directional hemispherical transmission coefficient calculated by the model; DHR mod (λ i ,N),DHT mod (λ i ,N) is the directional hemispherical reflection coefficient and directional hemispherical transmission coefficient calculated by the model under the condition of the optimal blade structure parameter N determined by formula (2-2-1); λ is the wavelength, which represents the near-infrared band of 800-1400nm in formula (2-2-2).

[0062] Substituting the blade structural parameter N and blade surface roughness σ obtained above, the average refractive index of the blade is calculated in the range of 500-800nm ​​using the spectral minimum fitting method. and K chls(λ), the specific algorithm is:

[0063]

[0064] The average refractive index and pigment absorption coefficient obtained by the above algorithm were compared with the spectral characteristics of existing papers and leaves, and the results were found to be consistent, proving that the parameters obtained are true and reliable.

[0065] Step 3: Construct the spatial geometric model of the curled leaf

[0066] The spatial geometric model considering the curling of the blade is described by the blade surface curvature model and the blade boundary curve model. The accuracy of the curling blade multi-angle radiation transfer model is verified based on the measured 3D point cloud data of the actual curling blade.

[0067] The parameters of the blade surface curvature model and the blade boundary curve model are obtained by fitting the three-dimensional point cloud coordinates of the curled blade surface with the corresponding boundary curve equation of the curled blade model through the partial least squares method. The obtained three-dimensional point cloud coordinates of the curled blade space are converted into the relationship between Z and X using Rapidform software to represent the degree of curling of the blade, and the edge point cloud coordinates of the curled blade are projected onto the XOY coordinate plane to fit the boundary ellipse equation of the curled blade. In addition, when using a scanner to perform a three-dimensional point cloud scan, relative coordinates are obtained instead of absolute coordinates with the center of the blade as the origin. Therefore, the fitted blade surface curvature model and blade boundary curve model are expressed as follows:

[0068]

[0069]

[0070] Where (x, y, z) are the absolute coordinates of any point on the surface of the curled leaf sample; (x0, y0, z0) are the absolute coordinates of the center point of the curled leaf sample (i.e., the center of the main vein); a represents the characteristic parameter of the relative leaf curling degree; b1 and b2 represent the major and minor axis parameters of the corresponding curled leaf ellipse boundary equation, respectively.

[0071] The various parameters in these two function equations are fitted using the partial least squares method, and the optimization objective function of the fitting is:

[0072]

[0073] ψ mea is the measured point cloud coordinate; ψ mod The point cloud coordinates calculated for the model.

[0074] Obtain the optimal model parameters by minimizing the objective function; Figure 3The curl characteristic curvature model and the corresponding leaf boundary ellipse function simulation results for leaf number 1 among the 60 leaf samples are shown. Parameters of the 3D structural model were fitted for each of the 60 curled leaf samples. The model accuracy was verified using the 3D structure reconstructed from scanned point cloud data as the actual leaf structure. The fitting results and accuracy are shown in Table 1.

[0075] Table 1 Curled leaf curvature model and its corresponding boundary function fitting effect and fitting accuracy

[0076]

[0077]

[0078] The coefficient of determination for the curled leaf curvature model and boundary function fit averaged over 0.9 across 60 different leaves, and the mean root mean square error for all tested leaves was almost less than 1 mm. Therefore, the curled leaf curvature model and elliptical boundary function obtained through biomechanics are reliable.

[0079] Step 4: Figure 1 As shown in Figure 3, a PROSPECT-MART model was constructed and used for spectral simulation.

[0080] 4-1. Using the spatial geometric model of the curled leaf obtained in step 3 as the scene for ray tracing, perform microplane segmentation to obtain multiple leaf microplanes distributed across the spatial geometric model. This segmentation process segments the space along the X and Y axes with a defined step size. All leaf microplanes are set within the detector's field of view.

[0081] 4-2. Take each blade microplane as the measured microplane in turn; Figure 2 As shown in the figure, ray tracing is performed from the measured microplane to obtain the angle of light directly irradiated by the light source on the measured microplane, as well as the angle of light reflected from the other blade microplanes; for each angle of light, the corresponding bidirectional reflectance distribution function BRDF is simulated using the blade reflectivity model to obtain the bidirectional reflectance distribution function BRDF of the measured microplane under direct light, as well as the contribution of other blade microplanes to the bidirectional reflectance distribution function BRDF of the measured microplane.

[0082] The blade reflectivity model specifically adopts the blade radiation transfer model considering the illumination direction provided in steps 1-4 of claim 1 of the patent application with publication number "CN111220552A" (the model is described as the "PROSPECT-MA model" in the following record), which can construct a bidirectional reflectance distribution function BRDF based on light incident from different directions.

[0083] 4-3. After completing the traversal of all leaf microplanes, the bidirectional reflectance distribution function BRDF of all leaf microplanes is combined as the bidirectional reflectance distribution function BRDF of the curled leaf.

[0084] The PROSPECT-MART model constructed in this step couples the PROSPECT-MA model with a ray tracing algorithm. The ray tracing algorithm can simulate the entire process of light rays starting from the light source, intersecting with the physical objects in the scene, undergoing multiple reflections and transmissions, and finally being collected by the detector. In this process, for the blades in the scene, it is necessary to consider both the three-dimensional structure and optical properties of the blades. Based on the PROSPECT model, the PROSPECT-MA model constructs the BRDF and BTDF models of the internal unit layer of the blade and the BRDF and BTDF models of the top unit layer of the blade. By coupling the internal layer and the top layer, it constructs a multi-angle radiation transmission model for flat blades that can be used for any light source incident angle and any observation angle.

[0085] The input parameters of the PROSPECT-MA model include the incident light source zenith angle, the reflected light zenith angle in the hemispherical direction, and the reflected light azimuth angle. These three parameters can be obtained from spectral measurement, as well as the inherent optical properties of the blade: the average refractive index of the blade. Average refractive index of leaf epidermis Leaf chlorophyll absorption coefficient K chls (λ) needs to be obtained before the model is run. Among them, the average refractive index of the leaf epidermis is Related to the wax layer on the leaf surface, it is usually set to 1.45. The leaf surface roughness σ and the leaf structure coefficient N are physical properties that reflect the leaf surface geometry and the internal cellular structure, respectively, and are different for each leaf. and K chls (λ) is related to the band characteristics of the spectrum, and the ANGERS dataset is used to substitute the PROSPECT model for the leaf sample-independent parameter and K chls (λ) is obtained.

[0086] The ray tracing algorithm needs to consider the three-dimensional structure of the curled blade when calculating. The three-dimensional structure of the curled blade is described by equations (3-1) and (3-2) in step 3. In this embodiment, in addition to the blade-related parameters σ and N contained in the PROSPECT-MA model, the simulation scene parameters for ray tracing need to be set. The scene setting is based on the detector parameters and position in the actual measurement - the detector field of view angle is 4°, the center distance between the detector and the blade observation point is 70mm, and the three-dimensional spatial information of the blade is described using the blade surface curvature model and the blade boundary curve model.

[0087] In this embodiment, 46 reliable leaf samples were selected from the spectral data obtained from 60 experimental curled leaves, and the PROSPECT-MART model was further used for forward modeling to calculate the multi-angle spectra of the curled leaves and evaluate the model accuracy.

[0088] The PROSPECT-MART model constructed in this embodiment is used to input the measured chlorophyll content of the curled leaves, and the BRDF (bidirectional reflectance distribution function) of the curled leaves is calculated and compared with the measured BRDF. The results are as follows: Figure 4 shown.

[0089] In order to verify the spectral simulation performance of the model in the 500-800nm ​​band at different angles, the measured spectra and simulated spectra of the curled leaves with light source zenith angles of 15°, 30°, 40° and 55°, observation azimuth angles of 60° and 100°, light source azimuth angles of 180° and observation zenith angles of 45° were compared and analyzed. The results are as follows: Figure 5 As shown in the figure, the measured and simulated spectra are relatively close at four light source zenith angles and two observation azimuth angles. The spectral curves at eight angles in the 500-700 nm band are essentially identical. In the infrared band, there are some differences between the measured and simulated spectra, but these differences are minor, within 0.02. The measured spectra in the figure show some variation with light source and observation angle, influenced by the spatial structure of the blade. The simulated spectra also vary somewhat at the eight angles, but the magnitude of the variation is smaller than that of the measured values. Apart from errors caused by instrument jitter during measurement, the spectral differences are not significant because the 60° and 100° observation directions are far from the specular reflection direction, resulting in relatively uniform diffuse reflection. Furthermore, approximations used in constructing the blade's spatial geometric model and simplifications in the optical path calculation during ray tracing can also affect the spectral simulation results. However, the simulated and measured spectra show consistent trends across different light source and detection angles.

[0090] Step 5: Pigment inversion using the PROSPECT-MART model

[0091] The PROSPECT-MART model was used to invert and calculate the chlorophyll content of curled leaves and compared with the measured values ​​to verify the application performance of the model.

[0092] The PROSPECT-MART model supports multiple light sources and observation angles. The model inversion requires input parameters: the light source zenith angle, light source azimuth angle, observation zenith angle, and observation azimuth angle. The model operates in the 500-800 nm wavelength band, and the input variables are the spectra of curled leaves at these angles. Using the spectral minimum fit method, an iterative calculation is performed to invert the chlorophyll content of the corresponding leaf at the corresponding angle.

[0093] Figure 6 This is the multi-angle chlorophyll inversion result for one of the curled leaves in the sample set. Figure 6 Parts (a), (b), (c), and (d) are the inversion results at four light source zenith angles of 15°, 30°, 40°, and 55°, respectively. The measured chlorophyll content of the leaf is 38.67 μg / cm 2 From the results of pigment inversion, the chlorophyll content obtained at most angles is roughly 20-50 μg / cm 2 The model's inversion results are poor at some edge angles, primarily between 135° and 225° in azimuth. This is a range of angles close to the light source. When measuring spectra in this direction, the sensor partially blocks the light source, casting shadows on the leaves. This may affect the accuracy of spectral measurements. Overall, the PROSPECT-MART model used in this paper has good inversion capabilities at most angles.

[0094] The RMSE was used to evaluate the accuracy of the PROSPECT-MART model color number inversion results for 46 curled leaf samples. Figure 7 The pigment inversion accuracy evaluation parameters of various PROSPECT-MART models are shown. Figure 7 Parts (a), (b), (c), and (d) are the RMSE, BIAS, SPEC, and CV of the pigment inversion results, corresponding to light source zenith angles of 15°, 30°, 40°, and 55°, respectively. At most light source zenith angles and probe observation angles, the RMSE of the model pigment inversion is basically 25 μg / cm 2 Below, mainly concentrated in 12-20μg / cm 2 The average RMSE of pigment inversion at the zenith angle of the four light sources is 15.97 μg / cm 2 、16.87μg / cm 2 、16.65μg / cm 2 、19.21μg / cm 2 The overall average RMSE was 17.17 μg / cm 2 Compared to the flat leaf multi-angle model, the RMSE is slightly higher, but not significantly. The threshold processing in the ray tracing process and the approximation processing in the three-dimensional construction of the leaf have a certain impact on the results. Looking at the angles in the hemisphere, the RMSE is slightly higher at the edge of the hemisphere in the direction of the mirror and the light source. The BIAS value is roughly between -10 and 10 μg / cm 2The BIAS of the mirror direction inversion results at the light source zenith angle of 40° and 55° is larger than that at 15° and 30°, which has the same trend as the flat blade multi-angle model PROSPECT-MA model. The SPEC values ​​of the model are concentrated in the range of 10-20μg / cm 2 between, mainly between 14μg / cm 2 The CV trend is similar to the SEPC trend, with performance at 15°, 30°, and 40° zenith angles being similar, but generally increasing at 55°. The CV trend is similar to the SEPC trend, with performance being relatively uniform at around 30% at 15°, 30°, and 40° zenith angles. It fluctuates significantly at 55° zenith angle, fluctuating between 30% and 55%.

Claims

1. The chlorophyll hyperspectral inversion method based on the curled plant leaf radiation transfer model is characterized by: The following steps are included: Step 1: For the curled blade to be tested, measure the hemispherical reflectance spectrum DHR and the hemispherical transmittance spectrum DHT; collect the three-dimensional point cloud data of the curled blade; Step 2: Obtain the leaf structure parameters N, leaf surface roughness σ, and leaf average refractive index m of the curled leaf la and leaf chlorophyll absorption coefficient K chls (λ); Step 3: Based on the three-dimensional point cloud data, obtain the spatial geometric model of the curled leaf; Step 4: Perform spectral simulation on the spatial geometric model of the curled leaf; 4-1. Using the spatial geometric model of a curled leaf as the ray tracing scene, perform microplane segmentation to obtain multiple leaf microplanes. 4-2. Each blade microfacet is used as the measured microfacet in turn. Ray tracing is performed starting from the measured microfacet to obtain the angle of light directly irradiated on the measured microfacet by the light source, as well as the angle of light reflected from the other blade microfacets. For each light angle, the corresponding bidirectional reflectance distribution function (BRDF) is simulated using the blade reflectivity model. The BRDF of the measured microfacet under direct light is obtained, as well as the contribution of the other blade microfacets to the BRDF of the measured microfacet. 4-3. After traversing all leaf microfacets, combine the bidirectional reflectance distribution function (BRDF) of all leaf microfacets as the bidirectional reflectance distribution function (BRDF) of the curled leaf to obtain the functional relationship between the chlorophyll content of the curled leaf and the directional hemispherical reflectance and directional hemispherical transmittance at the wavelengths corresponding to the maximum reflectance, maximum transmittance, and minimum absorptance. Step 5: Chlorophyll concentration inversion: Based on the functional relationship obtained in step 4 and the hemispherical reflectance spectrum DHR and hemispherical transmittance spectrum DHT measured in step 1, the chlorophyll concentration of the curled leaves is simulated.

2. The chlorophyll hyperspectral inversion method based on the curled plant leaf radiation transfer model according to claim 1 is characterized by: The blade reflectivity model uses the incident light source angle as one of the variables.

3. The chlorophyll hyperspectral inversion method based on the curled plant leaf radiation transfer model according to claim 1 is characterized by: The spatial geometric model obtained in step 3 is as follows: Among them, (x, y, z) are the coordinates of any point on the curled leaf surface; (x0, y0, z0) are the coordinates of the center point of the curled leaf sample; a represents the characteristic parameter of the leaf curling degree; b1 and b2 represent the major axis and minor axis parameters of the curled leaf, respectively.

4. The chlorophyll hyperspectral inversion method based on the curled plant leaf radiation transfer model according to claim 1, characterized in that: The hemispherical reflectance spectrum DHR and the hemispherical transmittance spectrum DHT described in step 1 are measured by a UV-3600 spectrophotometer + integrating sphere.

5. The chlorophyll hyperspectral inversion method based on the curled plant leaf radiation transfer model according to claim 1, characterized in that: The blade structural parameter N is calculated using the spectral minimum fitting method based on the maximum reflectivity, maximum transmittance, and minimum absorptivity bands within the 400-1000nm band. The blade structural parameter N is specifically as follows: Among them, λ1′, λ2′, and λ3′ are the wavelengths corresponding to the maximum spectral reflectance, maximum spectral transmittance, and minimum absorptivity of the curled blade respectively; DHR mea (·) and DHR mod (·) are the measured directional hemispherical reflection coefficient and the directional hemispherical reflection coefficient calculated by the model; DHT mea (·) and DHT mod (·) is divided into the measured directional hemispherical transmission coefficient and the directional hemispherical transmission coefficient calculated by the model.

6. The chlorophyll hyperspectral inversion method based on the curled plant leaf radiation transfer model according to claim 1, characterized in that: The blade surface roughness σ is obtained as follows: Wherein, λ is the wavelength in the 800-1400nm band; DHR mea (·) and DHR mod (·) are the measured directional hemispherical reflection coefficient and the directional hemispherical reflection coefficient calculated by the model; DHT mea (·) and DHT mod (·) is divided into the measured directional hemispherical transmission coefficient and the directional hemispherical transmission coefficient calculated by the model.

7. The chlorophyll hyperspectral inversion method based on the curled plant leaf radiation transfer model according to claim 1, characterized in that: Average refractive index of blades and K chls (λ) is obtained as follows: Wherein, λ is the wavelength in the 500-800nm ​​band; DHR mea (·) and DHR mod (·) are the measured directional hemispherical reflection coefficient and the directional hemispherical reflection coefficient calculated by the model; DHT mea (·) and DHT mod (·) is divided into the measured directional hemispherical transmission coefficient and the directional hemispherical transmission coefficient calculated by the model.

Images