A method for calculating leaf spectral reflectance
By dividing the blade into two-layer structures and optimizing parameters using the vulture optimization algorithm, the PISL-2 model solves the problem of insufficient assumptions of blade hierarchical structure and pigment distribution in the prior art, and improves the calculation accuracy of the spectral reflectivity of the blade.
Patent Information
- Application Number
- CN202310325122.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-03-30
- Publication Date
- 2024-11-26
- Estimated Expiration
- 2043-03-30
AI Technical Summary
The existing PROSPECT model assumes the isotropic nature of the blades and cannot effectively invert the hierarchical structure and non-uniform pigment distribution in non-density blades, resulting in limited model accuracy.
The PISL-2 model was proposed, and the blades were divided into two-layer structures. The condor optimization algorithm was used to optimize the structural parameters and biochemical content ratio. The total reflectivity and transmittance of the blades were calculated by superimposing multiple PROSPECT models, which was more in line with the actual stratification of the blades. structure.
The root mean square error between the measured value of the blade optical characteristics and the simulated value is improved, the applicability and accuracy of the model are enhanced, and the optical characteristics of the actual blade are more in line with the optical characteristics.
Smart Images

Figure CN116341385B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of vegetation remote sensing, and in particular to a method for calculating leaf spectral reflectance. Background Art
[0002] The radiation transfer theory of plant leaves is the research basis in the field of vegetation remote sensing. By simulating the radiation transfer process and mechanism of light inside the vegetation, information such as vegetation biochemical parameters and vegetation structure can be obtained. Compared with empirical / semi-empirical methods that lack clear physical meaning, the radiation transfer model shows stronger stability and portability in inverting vegetation physical and chemical parameters. PROSPECT is a radiation transfer model proposed by Jacquemoud based on Allen's generalized PLATE model. The model assumes that the leaves are composed of N layers of flat plates of the same nature, separated by N-1 layers of air. There are three initial input parameters: structural parameter N, chlorophyll content Cab, and equivalent water thickness Cw. Scholars have shown in the short infrared band that cellulose, lignin and protein in leaves also have a certain absorption effect on electromagnetic waves. Therefore, Fourty, Jacquemoud, Fourty and Baret added them to the model. Due to the strong absorption of electromagnetic waves by water in fresh leaves, the protein content cannot be inverted, and currently cellulose and lignin cannot be separated and quantified, so they are collectively referred to as dry matter. In this way, the PROSPECT model has four parameters: structural parameter N, chlorophyll content Cab, equivalent water thickness Cw and dry matter Cm. As long as the input parameters and the corresponding special spectral absorption coefficients are input, the reflectivity and transmittance of plant leaves at 400-2500nm can be accurately simulated, and the inversion can estimate the biochemical parameter content of the leaves.
[0003] A large number of scholars have applied and improved the PROSPECT model. Bousquet et al. described the physical properties of directional reflectance of leaves and introduced the specular reflection caused by the leaf surface into the PROSPECT model. Feret et al. successfully inverted the chlorophyll and carotenoid contents separately based on the improved PROSPECT5 model. Pedrós et al. estimated the fluorescence of chlorophyll a based on the PROSPECT model. Jay et al. proposed a physical model called ClOSINE (range Spectral Imaging of lEaves). By combining the PROSPECT model with the COSINE model, the biochemical characteristics and optical properties of leaves can be extracted from close-range hyperspectral images at the same time. Furthermore, Li et al. proposed an inversion method combining PROSPECT with continuous wavelet transform (CWT), called PROCWT, based on the assumption that the surface reflectance of rice and wheat leaves is independent of wavelength, to enhance the absorption characteristics of chemical components in rice and wheat and reduce the reflectance effect of the leaf surface. The PROSECT model simulates the radiation transfer process of light in leaves relatively accurately. Many researchers have used different types of leaves to test the performance of the PROSECT model, proving that the model is one of the best leaf radiation transfer models currently available.
[0004] With the continuous development of the PROSPECT model, the model's assumptions restrict the model's accuracy and application. The PROSPECT model is based on the assumption of isotropic properties inside the leaf. The non-dense leaf is regarded as consisting of N dense layers and N-1 air interlayers. Each dense layer is an optically isotropic layer. The incident solid angle replaces the assumption of the PLATE model that the incident light is isotropic parallel light. However, this assumption of isotropy inside the leaf is not completely consistent with the structure of the actual leaf. The pigments inside some leaves are not evenly distributed. AM Donato, Arimy, NQ, Bugg, C., CABBONE, AV, etc. observed slices of leaves of various plants such as myrtle, pitcher plant, moss and clover, and found that different plant leaf tissues have obvious differences, and the distribution of various elements has obvious stratification. The stratification of these leaves cannot be regarded as isotropic homogeneous layers, and the optical properties of different strata are quite different. Therefore, the existing PROSPECT model still has certain defects. Summary of the invention
[0005] The embodiment of the present invention provides a method for calculating the spectral reflectance of a blade, which can solve the problems existing in the prior art.
[0006] The present invention provides a method for calculating the spectral reflectance of a blade, comprising the following steps:
[0007] Multiple PROSPECT models were superimposed to construct the PISL-2 model;
[0008] The leaves were divided into two different layers to obtain the structural parameters, biochemical content and spectral absorption coefficient of the leaves;
[0009] The vulture optimization algorithm was used to optimize the structural parameters and biochemical content ratio of the leaves;
[0010] The spectral absorption coefficient of the leaves as well as the optimized structural parameters and biochemical content ratio were input into the PISL-2 model, and the total reflectance of the leaves was calculated using the PISL-2 model.
[0011] Preferably, the structural parameters of the leaf include the structural parameters N1 of the first layer and the structural parameters N2 of the second layer, and the biochemical content includes the chlorophyll content Cab of the first layer. 1 and the chlorophyll content of the second layer Cab 2 , the equivalent water thickness of the first layer Cw 1 and the equivalent water thickness ratio Cw of the second layer 2 And the first layer dry matter content Cm 1 The dry matter content of the second layer Cm 2 .
[0012] Preferably, the spectral absorption coefficient of the blade uses the spectral absorption coefficient in the PROSPECT model, and the wavelength range is 400nm to 2500nm, with an interval of 5nm.
[0013] Preferably, the total reflectivity is calculated by the following formula:
[0014]
[0015] Where R is the total reflectivity of a beam of light after it is incident on the blade surface and then reflected back to the upper surface of the blade multiple times, r 10 is the reflectivity of the first layer, t 10 is the transmittance of the first layer, τ 1 is the transmission coefficient of the first layer, τ 2 is the transmission coefficient of the first layer, t 01 is the transmittance of light on the upper surface of the leaf, r 21 is the reflectivity of light at the interface between the first layer and the second layer, t 12 is the transmittance of light at the interface between the first layer and the second layer, r 02 is the reflectivity of the light on the lower surface of the blade when it passes through the lower surface of the blade from the second layer, t 21 is the transmittance of light at the interface where the second layer passes through the first layer, r 01 Yes 21 The reflectivity produced on the upper surface of the blade.
[0016] Preferably, the total transmittance of the blade is calculated by the PISL-2 model, and the total transmittance is calculated by the following formula:
[0017]
[0018] In the formula, T represents the total transmittance of a beam of light after passing through the transmittance of the first layer of medium, entering the second layer of medium, and then undergoing multiple reflections on the lower surface of the blade. 20 It is the transmittance of light when it passes through the lower surface of the blade from the second layer.
[0019] Preferably, the calculated total reflectivity is evaluated using a root mean square error;
[0020] The root mean square error is calculated by the following formula:
[0021]
[0022] Where, X meas,j and X mod,j is the measured reflectivity and simulated reflectivity of a leaf j, and n is the number of samples.
[0023] Preferably, the processing process of the vulture optimization algorithm includes three parts: selecting a search space, searching in the selected search space, and diving hunting.
[0024] Preferably, the vulture optimization algorithm expresses the selection search space by the following formula:
[0025] P i,new =P best +α·r(P mean -P i )
[0026] Where P I,new is the current position of the vulture, α is the parameter that controls the position change, and its value range is [1.5, 2], r is any value between [0, 1], and P best is the optimal position identified by the vulture in the last search, P mean is the location data of the vulture at each previous time point; P i is the position of the i-th vulture.
[0027] Preferably, the vulture optimization algorithm represents the search in the selected search space by the following formula:
[0028] P i,new =P i +y(i)×(P i -P i+1 ) best+x(i)×(P i -P mean )
[0029] Where P i+1 is the position of the i+1th vulture, and the values of x(i) and y(i) are the positions of the vulture in the coordinates of the search phase.
[0030] Preferably, the vulture optimization algorithm represents the dive hunting by the following formula:
[0031] P i,new =rand*P best +x 1 (i)×(P i -c 1 *P mean )+y 1 (i)×(P i -c 2 *P best )
[0032] In the formula, c 1 、c 2 is the moving speed of the vulture when it reaches the optimal position, c 1 、c 2 ∈[1,2], rand is any value in [0,1], x 1 (i) and y 1 (i) is the position of the vulture in coordinates during the dive phase.
[0033] Compared with the prior art, the present invention has the following beneficial effects:
[0034] The present invention proposes a PISL-2 model, which is based on the assumption that leaves can be divided into two isotropic mean layers. According to the input leaf biochemical parameters and absorption coefficients, the material content ratio of each layer can be optimized to simulate the spectral reflectance. Compared with the traditional PROSPECT model, the PISL-2 model of the present invention is more in line with the actual situation of leaf stratification, and the root mean square error between the measured value and the simulated value of the leaf optical property is reduced to a certain extent compared with the PROSPECT model. BRIEF DESCRIPTION OF THE DRAWINGS
[0035] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0036] Figure 1A flow chart of a method for calculating leaf spectral reflectance of the present invention;
[0037] Figure 2 Schematic diagram of the light transmission process in the PLATE model;
[0038] Figure 3 Schematic diagram of multiple reflection and transmission of two groups of media in the Stokes optical system;
[0039] Figure 4 (a) is a schematic diagram of multiple reflection and transmission transmission of Jacquemond's first iterative method;
[0040] Figure 4 (b) is a schematic diagram of multiple reflection transmission of Jacquemond's second iterative method;
[0041] Figure 5 This is a cross-section of a plant leaf;
[0042] Figure 6 It is a schematic diagram of the light transmission process of the PISL-2 model of the present invention;
[0043] Figure 7 (a) is a schematic diagram showing the variation of the structural parameter N1 of the first layer on four data sets in an embodiment of the present invention;
[0044] Figure 7 (b) is a schematic diagram of the variation of the structural parameter N2 of the second layer on four data sets in an embodiment of the present invention;
[0045] Figure 7 (c) is a schematic diagram showing the variation of the proportional coefficient of the double-layer chlorophyll content in four data sets according to an embodiment of the present invention;
[0046] Figure 7 (d) is a schematic diagram showing the variation of the double-layer equivalent water thickness proportional coefficient on four data sets in an embodiment of the present invention;
[0047] Figure 7 (e) is a schematic diagram showing the variation of the proportional coefficient of the double-layer dry matter content in four data sets according to an embodiment of the present invention;
[0048] Figure 8 (a) is a comparison diagram of the simulated spectrum and the measured spectrum of a sample in Dataset 1 based on PISL-2 or PROSPECT in an embodiment of the present invention;
[0049] Figure 8 (b) is a comparison diagram of the simulated spectrum and the measured spectrum of another sample in Dataset 1 based on PISL-2 or PROSPECT in an embodiment of the present invention;
[0050] Figure 8 (c) is a comparison diagram of the simulated spectrum and the measured spectrum of a sample in data set 2 based on PISL-2 or PROSPECT in an embodiment of the present invention;
[0051] Figure 8 (d) is a comparison diagram of the simulated spectrum and the measured spectrum of a sample in data set 2 based on PISL-2 or PROSPECT in an embodiment of the present invention;
[0052] Figure 8 (e) is a comparison diagram of the simulated spectrum and the measured spectrum of a sample in data set 3 based on PISL-2 or PROSPECT in an embodiment of the present invention;
[0053] Figure 8 (f) is a comparison diagram of the simulated spectrum and the measured spectrum of a sample in data set 3 based on PISL-2 or PROSPECT in an embodiment of the present invention;
[0054] Figure 8 (g) is a comparison diagram of the simulated spectrum and the measured spectrum of a sample in data set 4 based on PISL-2 or PROSPECT in an embodiment of the present invention;
[0055] Figure 8 (h) is a comparison diagram of the simulated spectrum and the measured spectrum of a sample in data set 4 based on PISL-2 or PROSPECT in an embodiment of the present invention;
[0056] Fig. 9 (a) is a comparison diagram of the root mean square error between PISL-2 and PROSPECT of data set 1 in an embodiment of the present invention;
[0057] Fig. 9 (b) is a comparison diagram of the root mean square error between PISL-2 and PROSPECT of data set 2 in an embodiment of the present invention;
[0058] Fig. 9 (c) is a comparison diagram of the root mean square error between PISL-2 and PROSPECT of data set 3 in an embodiment of the present invention;
[0059] Fig. 9 (d) is a comparison chart of the root mean square error between PISL-2 and PROSPECT for data set 4 in an embodiment of the present invention. DETAILED DESCRIPTION
[0060] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0061] Reference Figure 1 The present invention provides a method for calculating the spectral reflectance of a leaf, comprising the following steps:
[0062] Step 1: Superimpose multiple PROSPECT models to construct the PISL-2 model.
[0063] Original PROSPECT model principle:
[0064] Dense leaves with structural parameter N = 1. The leaves are regarded as translucent isotropic absorbing plates with uniform rough surfaces, and the light is assumed to be isotropic and parallel. This is the PLATE model, which successfully simulates the reflectivity of dense corn leaves. The assumption of dense leaves in the PLATE model is extended to non-dense leaves. This non-dense leaf is regarded as consisting of N dense layers and N-1 air interlayers. Each dense layer is an optically isotropic layer with a refractive index of n and an absorption coefficient of k. The parameter N is extended from an integer to a real number using Stokes theory. The extended PLATE model is called the Generalized PLATE model. The PROSEPCT model is a model evolved from the Generalized PLATE model to simulate the hemispherical reflectivity and transmittance of leaves, which scatters light isotropically.
[0065] For N = 1, Figure 2 The transmission process of a beam of light passing through a rough surface plate in the PLATE model is shown. The refractive index of the blade is n, and the absorption coefficient is k. According to the reflectivity and transmittance formulas of the PLATE model,
[0066]
[0067]
[0068] Among them, r ij ,t ij represents the average reflectivity and average transmittance from medium i to medium j; τ represents the transmittance of light when it passes through medium 2 at one time; R α (1) and T α (1) represents the reflectivity of a beam of light after it is incident on medium 2 at an angle α and then reflected back to medium 1 after multiple reflections, T α (1) represents the total transmittance of a beam of light after it is incident on medium 2 at an angle α and passes through medium 2 multiple times.
[0069] The incident solid angle α is introduced to replace the assumption of isotropic light in the PLATE model, and t 12 =t av (α,1,n), r 12 =1-t12 , r 21 =1-t 21 , t 21 =t av (90,n,1), t av (90,n,1)=n -2 t av Substituting (90,1,n) into (13) and (14), we obtain:
[0070]
[0071]
[0072] Formulas (15) and (16) are the reflectivity and transmittance of a layer of dense medium in the PROSPECT model. Compared with the PLATE model, the assumption that the incident light is isotropic is no longer required.
[0073] Reference Figure 3 , non-dense leaves with N>1. When N>1, Stokes arbitrarily divides N into m layers and n layers, satisfying N=m+n. Suppose a beam of light with energy 1 is incident on the first layer of the m layers.
[0074] The expressions for reflected energy and transmitted energy are:
[0075]
[0076]
[0077] Where φ(m) is the energy reflected from the first group of m layers, Ψ(m) is the energy transmitted from the first group of m layers, and a, b are expressions containing the meanings of φ(m) and φ(n). When m=1, it is the PLATE model.
[0078] Jacquemond used an iterative method to solve the reflectivity and transmittance of light passing through N layers of isotropic media. There are two methods: one is to know the reflectivity and transmittance of the first layer, and then deduce the reflectivity and transmittance of the remaining layers; the other is to know the reflectivity and transmittance of the first N-1 layers, and then add a layer each time to calculate the reflectivity and transmittance.
[0079] Reference Figure 4 (a), the first layer is covered on the N-1 layer, and the light reflected from the first layer is reflected by R α (1) indicates that the transmission is represented by T α (1). Since the transmission of light in the medium is isotropic, the emission and transmittance of the N-1 layer are expressed by R 90 (N-1) and T 90 (N-1) indicates.
[0080] The total reflectivity and transmittance of the N layers are:
[0081]
[0082]
[0083] Reference Figure 4 (b) When N-1 layers are placed on top of 1 layer,
[0084]
[0085]
[0086] Using Stokes's formula for incident light reflectivity:
[0087]
[0088] Let m = N and we get:
[0089]
[0090] From formula (24), we get:
[0091] Substituting into formula (19) and (20) we get:
[0092]
[0093]
[0094] in, or
[0095] Firstly, the reflectivity and transmittance of the N-layer blade are calculated by using formulas (15) and (16) for the interaction between isotropic light and a layer of medium within a certain solid angle α. Then, the reflectivity and transmittance of the N-layer blade are calculated by using formulas (25) and (26) based on the iteration principle.
[0096] PISL-2 model construction principle:
[0097] Reference Figure 5The distribution of pigments in leaves is uneven. From the cross-section of plant leaves, it can be seen that leaves are composed of multiple structures including epidermis, palisade tissue, spongy tissue, phloem, etc. The distribution of various elements in various tissues is not uniform. The leaf layers cannot be regarded as homogeneous layers of optical properties according to the traditional PROSPECT model theory. In the present invention, the leaves are regarded as multiple layers of homogeneous layers of optical properties, and the element content and light absorption coefficient of each layer are different. The PROSPECT model is applied multiple times to establish the PROPSECT superposition model, namely the PISL-2 model. Finally, the total reflectance and total transmittance are obtained by cumulative superposition.
[0098] Reference Figure 6 , the present invention divides the blade into two layers, and the transmission coefficient of the first layer is τ 1 , the structural parameter is N 1 , the absorption coefficient is k 1 ; The transmission coefficient of the second layer is τ 2 , the structural parameter is N 2 , the absorption coefficient is k 2 .
[0099] When a beam of light with energy 1 is incident from air (medium 0) to the surface of the first layer (medium 1), reflection and transmission will occur. Suppose the reflectivity is r 10 , the transmittance is t 10 , without considering the absorption of light by the leaf surface, then r 10 =1-t 10 The reflectivity of this beam of light in the first layer is expressed as x 1 =r 10 +t 01 τ 1 r 21 τ 1 t 20 The transmittance t generated by this beam of light in the first layer 12 As the incident light of the second layer, it will produce a reflectivity and a transmittance after passing through the second layer. The reflectivity will react to the first layer, producing a transmittance on the first layer. At the same time, the reflectivity produced by the first layer will react to the second layer, and this iteration will be carried out until the light intensity gradually weakens to 0.
[0100] The total reflectivity is:
[0101]
[0102] Where R is the total reflectivity of a beam of light after it is incident on the blade surface and then reflected back to the upper surface of the blade multiple times, r 10 is the reflectivity of the first layer, t 10 is the transmittance of the first layer, τ 1 is the transmission coefficient of the first layer, τ2 is the transmission coefficient of the first layer, t 01 is the transmittance of light on the upper surface of the leaf, r 21 is the reflectivity of light at the interface between the first layer and the second layer, t 12 is the transmittance of light at the interface between the first layer and the second layer, r 02 is the reflectivity of the light on the lower surface of the blade when it passes through the lower surface of the blade from the second layer, t 21 is the transmittance of light at the interface where the second layer passes through the first layer, r 01 Yes 21 The reflectivity produced on the upper surface of the blade.
[0103] The total transmittance is:
[0104]
[0105] Among them, r ij ,t ij (i=1,2,j=1,2) represents the average reflectivity and average transmittance from medium i to medium j; τ i (i=1,2) represents the transmittance of light when it passes through medium i once; R represents the total reflectance of a beam of light after it is incident on the surface of the blade and reflected back to the upper surface of the blade multiple times; T represents the total transmittance of a beam of light after it passes through the transmittance of the first layer of medium and then enters the second layer of medium and is reflected multiple times on the lower surface of the blade.
[0106] Step 2: Divide the leaves into two different layers, and obtain the structural parameters, biochemical content and spectral absorption coefficient of the two layers. The two-layer structural parameters of the leaves include the structural parameters N1 of the first layer and the structural parameters N2 of the second layer. The two-layer biochemical content includes the chlorophyll content Cab of the first layer. 1 and the chlorophyll content of the second layer Cab 2 , the equivalent water thickness of the first layer Cw 1 Ratio of equivalent water thickness to the second layer Cw 2 And the first layer dry matter content Cm 1 The dry matter content of the second layer Cm 2 The spectral absorption coefficient uses the absorption coefficient in the PROSPECT model, with a wavelength range of 400nm to 2500nm and an interval of 5nm.
[0107] The structural parameter N of the leaf describes the complexity of the internal structure of the leaf. A low value (1-1.5) indicates that the mesophyll cells are simpler and more compact, while a high value (1.5-2) indicates that the mesophyll is more complex and contains more intracellular air space. The structural parameter N is indirectly estimated by measuring the reflectance and transmittance of the leaf in the near-infrared band. The traditional PROSPECT model is based on the assumption that the leaf is composed of isotropic plates. In the model, a structural parameter N is used to represent the structural characteristics of the leaf. However, in fact, the internal structure of the leaf is not uniform. The mesophyll cells of most dicotyledons are differentiated into palisade tissue and spongy tissue. This structure is usually called a two-sided leaf. However, the mesophyll of some monocotyledons does not have the differentiation of palisade tissue and spongy tissue, or there is palisade tissue on the upper surface. This leaf is called an isoplanatic leaf. The structural characteristics of leaves of different plants are different, which leads to different distributions of material content inside the leaf and a stratified phenomenon.
[0108] Step 3: Use the vulture optimization algorithm to optimize the parameters of the blade.
[0109] In order to analyze the structural parameters N1, N2 and material content ratio parameter Cab of PISL-2 model 12 , Cw 12 , Cm 12 For optimization, the present invention uses the vulture optimization algorithm BES. The BES algorithm simulates the hunting behavior of vultures and is divided into three stages, namely, selecting a search space, searching in the selected search space, and diving to hunt.
[0110] In the search phase, the vulture identifies and selects an optimal area (in terms of the amount of food) in the selected search space and hunts prey in this area. This phase can be expressed by equation (1):
[0111] P i,new =P best +α·r(P mean -P i ) (1)
[0112] α is a parameter that controls the position change, r is any value between 0 and 1, and P best is the optimal position identified by the vulture in the last search. The vulture arbitrarily searches every point in the previously selected search space, and Pmean defines each position data of these vultures at the previous time point. During the search phase, the vulture searches for prey in the selected search space and moves in different ways in the spiral space to increase speed.
[0113] The mathematical expression of the optimal dive position is (2):
[0114] P i,new =Pi +y(i)×(P i -P i+1 ) best +x(i)×(P i -P mean ) (2)
[0115] in,
[0116]
[0117] xr(i)=r(i)×sin(θ(i)), yr(i)=r(i)×cos(θ(i)) (4)
[0118] θ(i)=α×π×rand (5)
[0119] r(i)=θ(i)×R×rand (6)
[0120] In the formula, α takes a value between 5 and 10 to define the rotation angle of the search center point, and R takes a value between 0.5 and 2 to define the number of search cycles. During the search, the dispersion of the local optimal solution can be enhanced according to the different parameters α and R, and effective solutions can be continuously obtained. During the dive, the vulture moves from the search space to the best position of the target prey, and finally moves to the vicinity of the optimal point.
[0121] P i,new =rand*P best +x 1 (i)×(P i -c 1 *P mean )+y 1 (i)×(P i -c 2 *P best ) (7)
[0122]
[0123] xr(i)=r(i)×sinh(θ(i)), yr(i)=r(i)×cosh(θ(i)) (9)
[0124] θ(i)=α×π×randr(i)=θ(i) (10)
[0125] In the formula, c 1 、c 2 is the moving speed of the vulture when it reaches the optimal position, c 1 、c 2 ∈[1,2], rand is any value in [0,1], x 1 ( i ) and y1 (i) is the position of the vulture in the dive phase coordinates. The eagle's motion trajectory can be expressed in polar coordinates. The optimal point is calculated by multiplying the variance between the current position and the center point by the polar coordinate point on the x-axis. The error rate is treated as a fitness function, and the solution with the smallest error is regarded as the optimal solution, resulting in equation (11):
[0126]
[0127] The root mean square error (RMSE) was calculated to evaluate the difference between the simulated reflectance obtained by the PISL-2 model and the true reflectance of the leaf elements measured from different data sets:
[0128]
[0129] Among them, X meas,j and X mod,j is the measured reflectivity and simulated reflectivity of a leaf j, and n is the number of samples.
[0130] Step 4: Input multiple parameters of the optimized blade into the PISL-2 model to calculate the total reflectivity and total transmittance of the blade.
[0131] Example
[0132] The present invention has collected a total of 4 data sets from different researchers, namely LOPEX, ANGERS, KIT data and CABO data, as shown in Table 1. The LOPEX used in the present invention contains 46 plant species and a total of 330 samples, which are divided into two types: monocotyledon and dicotyledon. Among them, there are 80 monocotyledon samples and 250 dicotyledon samples, and each five samples are a group. Since the samples of No. 176-180 and No. 196-200 have abnormal parameters, they are deleted, so there are 74 monocotyledon samples and 245 dicotyledon samples. During verification, verification was performed in two cases: monocotyledon and dicotyledon. The ANGERS data set includes 276 samples of 44 plant species, all of which are dicotyledonous plant types. The KIT data set includes 38 different types of herbaceous plants, and a total of 630 samples are obtained after removing abnormal data. The fourth data set includes 610 samples of 73 trees and herbaceous plants, and 558 samples remain after removing abnormal data.
[0133] Table 1 Dataset sample statistics
[0134]
[0135] The structural parameters N1, N2 and material content ratio Cab of the two layers were optimized using the condor optimization algorithm. 12 , Cw 12 , Cm 12The optimization results are shown in Table 2.
[0136] Table 2 Optimization results of PISL-2 model parameters using BES
[0137]
[0138] In general, the structural parameters of the four data sets all have the characteristics of thicker first layer and thinner second layer. The chlorophyll ratio coefficient is close to 1, indicating that most of the chlorophyll is concentrated in the first layer and the second layer is almost 0. 12 The ratio is relatively high. The Cw of the fourth data set 12 The ratio is relatively low at 0.0129. The Cm of the first and third data sets 12 The ratios are 0.7401 and 0.5894 respectively, while the Cm of the second and fourth data sets are 12 The ratios are 0.2781 and 0.0027 respectively. It can be seen that the equivalent water thickness of the leaves of the first three data sets is concentrated in the first layer, while the equivalent water thickness of the fourth data set is concentrated in the second layer. The distribution of dry matter in the first and third data sets is relatively large, which is very different from the other two data sets. Therefore, the material content distribution of leaves of different plants in different data sets is very different, and the distribution of elements is not exactly the same. Figure 7 The variations of five structural parameters on different samples are plotted.
[0139] from Figure 7 It can be seen that the changing trend of the same parameter, that is, the symbols of the same color, have a certain correlation, such as Figure 7 As shown in (a), N1 of dataset 3 is mostly distributed in the range of 0.5 to 1, while N1 of dataset 4 is mainly in the range of 1 to 1.5; Figure 7 As shown in (b), the structural parameters N2 of the second layer are concentrated near 0, indicating that the structure of the second layer is thinner than that of the first layer, but there are also some individual parameters N2 located at a higher position, which may be caused by measurement errors or optimization thresholds, but does not affect the overall change trend of the structural parameters N2; Figure 7 As shown in (c), the chlorophyll content ratio coefficients of the four data sets are close to 1, indicating that the chlorophyll in the first layer accounts for a large proportion, and only a small amount of chlorophyll is in the second layer; Figure 7 (d) and (e) show that the equivalent water thickness and dry matter content of the fourth data set are mainly located in the first layer, which is the opposite of the other three data sets.
[0140] In order to more vividly see the simulated spectrum effect of PISL-2, the simulated spectrum of PISL-2 and PROSPECT are compared with the measured spectrum. The comparison results are shown in Figure 8 .
[0141] The present invention uses the root mean square error (RMSE) to measure the error between the simulated spectrum of the PISL-2 model and the measured spectrum.
[0142] Figure 8 is the simulated spectra of two samples randomly selected from the many samples in each data set, from Figure 8 It can be seen that compared with the traditional PROSPECT model, the simulated spectra based on the PISL-2 model are more consistent with the measured spectra in the four data sets. Among them, the results of a and b are the best, and the simulated spectra are almost consistent with the measured spectra. The root mean square error of PISL-2 is 0.0046 and 0.0094, and the RMSE of the PROSPECT model is 0.1089 and 0.1301, which is 0.1220 and 0.1207 lower than the RMSE of the PROSPECT model. From Figures c and d, it can be seen that PISL-2 performs better than the PROSPECT model in the visible light band and worse than the PROSPECT model in the near-infrared band, but the overall root mean square error is 0.0313 and 0.0251 lower than the PROSPECT model. The simulation effect of PISL in Figures c, d, e and f is much better than that of the PROSPECT model. It can be clearly seen from the figures that there is a large gap between the simulated spectrum of the traditional PROSPECT model and the measured spectrum, especially in the band after visible light, where the consistency is poor. In this regard, the PISL model shows great advantages, and the root mean square error is reduced by 0.0504, 0.0474, 0.1201 and 0.1155 respectively compared with the PROSPECT model.
[0143] Fig. 9 The root mean square error between the simulated spectra and the measured spectra based on the PISL-2 model or the PROSPECT model on different data sets varies with the sample. In general, the RMSE based on the PISL-2 model is lower than that of the PROSPECT model, but the RMSE of PISL-2 is slightly larger than that of PROSPECT on the second data set. The statistical results of RMSE are given in Table 3. The RMSE of PISL-2 on these four data sets is between 0.001842 and 0.177052, with an average value of 0.01372 to 0.110924. The errors of the first three data sets are relatively smaller than those of the fourth data set; while the RMSE of PROSPECT is between 0.007632 and 0.200478, with an average value of 0.018841 to 0.154588, and the error is slightly larger than that of PISL-2.
[0144] Table 3 RMSE statistical results of simulated spectra based on PISL-2 and PROSPECT and measured spectra of four data sets
[0145]
[0146]
[0147] The present invention superimposes the traditional PROSPECT model and proposes the PISL-2 model. The assumption of this model is that the blade is regarded as composed of two different layers of structure. The structural parameters and material content distribution of the two layers are different. The structural parameters of the two layers are N1 and N2. Figure 7 a and Figure 7 As can be seen from b, the sizes of N1 and N2 are not the same in different datasets. N1 in dataset 3 is mostly distributed between 1-1.5, and in dataset 4 is mostly distributed between 0.5-1. This shows that the first layer of leaf mesophyll in dataset 3 is more complex, contains more intracellular air space, and has a richer ventilation system between cells. The leaf mesophyll cells in dataset 4 are simple and compact, with small intercellular gaps, while Figure 7 In b, N2 of dataset 4 is mostly distributed in the range of 0-0.05, indicating that the first layer of the leaf in dataset 4 is thicker than the second layer, the intracellular gap is larger, but the overall leaf cells are very compact. N1 of datasets 1 and 2 is mainly distributed in the range of 1-2, and N2 is close to 0, indicating that the second layer of leaf cells in datasets 1 and 2 is very compact, the gaps between cells are very small, and most of the gaps are located in the first layer. Each dataset contains a variety of plant leaves, and the distribution of the internal material content of different plant leaves is not exactly the same. It is also necessary to determine the optimal number of layers suitable for different plant types. In future studies, the number of layers can be expanded to 3 or even more to further improve the applicability of the model.
[0148] The present invention proposes the PISL-2 model, which is a new version of the PROSPECT radiation transfer model. Based on the assumption that leaves can be divided into two isotropic mean layers, the material content ratio of each layer can be optimized according to the input leaf biochemical parameters and absorption coefficients, and the spectral reflectance can be simulated. The PISL-2 model is used to simulate the biochemical parameters of plant leaves from four data sets from different study areas. The root mean square error between the measured values and the simulated values of the leaf optical properties of each data set is reduced to a certain extent compared with PROSPECT.
[0149] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.
[0150] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.
Claims
1. A method for calculating the spectral reflectance of leaves, It is characterized in that The following steps are involved: Multiple PROSPECT models were superimposed to construct the PISL-2 model; The leaves were divided into two different layers to obtain the structural parameters, biochemical content and spectral absorption coefficient of the leaves; The vulture optimization algorithm was used to optimize the structural parameters and biochemical content ratio of the leaves; The spectral absorption coefficient of the leaves, as well as the optimized structural parameters and biochemical content ratios were input into the PISL-2 model, and the total reflectance of the leaves was calculated using the PISL-2 model. The structural parameters of the leaf include the structural parameters N1 of the first layer and the structural parameters N2 of the second layer, and the biochemical content includes the chlorophyll content Cab of the first layer. 1 and the chlorophyll content of the second layer Cab 2 , the equivalent water thickness of the first layer Cw 1 and the equivalent water thickness ratio Cw of the second layer 2 And the first layer dry matter content Cm 1 The dry matter content of the second layer Cm 2 ; The total reflectivity is calculated by the following formula: Where R is the total reflectivity of a beam of light after it is incident on the blade surface and then reflected back to the upper surface of the blade multiple times, r 10 is the reflectivity of the first layer, t 10 is the transmittance of the first layer, τ 1 is the transmission coefficient of the first layer, τ 2 is the transmission coefficient of the first layer, t 01 is the transmittance of light on the upper surface of the leaf, r 21 is the reflectivity of light at the interface between the first layer and the second layer, t 12 is the transmittance of light at the interface between the first layer and the second layer, r 02 is the reflectivity of the light on the lower surface of the blade when it passes through the lower surface of the blade from the second layer, t 21 is the transmittance of light at the interface where the second layer passes through the first layer, r 01 Yes 21 The reflectivity produced on the upper surface of the blade; The total transmittance of the blade is calculated by the PISL-2 model, and the total transmittance is calculated by the following formula: In the formula, T represents the total transmittance of a beam of light after passing through the transmittance of the first layer of medium, entering the second layer of medium, and then undergoing multiple reflections on the lower surface of the blade. 20 It is the transmittance of light when it passes through the lower surface of the blade from the second layer.
2. A method for calculating leaf spectral reflectance as claimed in claim 1, It is characterized in that The spectral absorption coefficient of the blade adopts the spectral absorption coefficient in the PROSPECT model, and the wavelength range is 400nm to 2500nm, with an interval of 5nm.
3. A method for calculating leaf spectral reflectance as claimed in claim 1, It is characterized in that The calculated total reflectivity was evaluated using the root mean square error; The root mean square error is calculated by the following formula: Where, X meas,j and X mod,j is the measured reflectivity and simulated reflectivity of a leaf j, and n is the number of samples.
4. A method for calculating leaf spectral reflectance as claimed in claim 1, It is characterized in that The processing process of the vulture optimization algorithm includes three parts: selecting a search space, searching in the selected search space, and diving for prey.
5. A method for calculating leaf spectral reflectance as claimed in claim 4, It is characterized in that The vulture optimization algorithm expresses the selection search space by the following formula: P i,new =P best +α·r(P mean -P i ) Where P I,new is the current position of the vulture, α is the parameter that controls the position change, and its value range is [1.5, 2], r is any value between [0, 1], and P best is the optimal position identified by the vulture in the last search, P mean is the location data of the vulture at each previous time point; P i is the position of the i-th vulture.
6. A method for calculating leaf spectral reflectance as claimed in claim 5, It is characterized in that The vulture optimization algorithm expresses the search in the selected search space by the following formula: P i,new =P i +y(i)×(P i -P i+1 )P best +x(i)×(P i -P mean ) Where P i+1 is the position of the i+1th vulture, and the values of x(i) and y(i) are the positions of the vulture in the coordinates of the search phase.
7. A method for calculating leaf spectral reflectance as claimed in claim 6, It is characterized in that The vulture optimization algorithm expresses the dive hunting as follows: P i,new =rand*P best +x 1 (i)×(P i -c 1 *P mean )+y 1 (i)×(P i -c 2 *P best ) In the formula, c 1 、c 2 is the moving speed of the vulture when it reaches the optimal position, c 1 、c 2 ∈[1,2], rand is any value in [0,1], x 1 (i) and y 1 (i) is the position of the vulture in coordinates during the dive phase.
