A simulation method, device, equipment and medium for leaf spectral reflectance

Abstract

The application discloses a kind of simulation methods, devices, equipment and medium of leaf spectral reflectance, it is related to vegetation remote sensing technical field, including considering the optical properties of leaf surface, biochemical parameter related diffuse reflection component and with surface structure related specular reflection component, obtain leaf surface geometric feature structure, and leaf surface roughness, introduce light source incident zenith angle;Using Ctenophore colony algorithm fitting obtains leaf surface roughness;Two layer radiation transfer model PIOSL model is superimposed, and PIOSL-surf model is constructed;Leaf surface roughness fitted out, light source incident zenith angle, structure parameter and biochemical content are as the input of PIOSL-surf model, obtain leaf total reflectivity and total transmittance, and generate simulation spectrum chart.The present application is based on PIOSL-surf model, introduces the optical properties of leaf surface, can be according to input leaf biochemical parameter, absorption coefficient and light source incident zenith angle, carry out the simulation of spectral reflectivity, accuracy greatly improves.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of vegetation remote sensing, and in particular to a leaf spectral reflectance simulation method, device, equipment and medium. BACKGROUND

[0002] The leaf optical model uses mathematical, physical and chemical principles to quantitatively analyze factors affecting the optical properties of leaves; these factors can include the shape, structure, pigment content, water content, cell arrangement of the leaf, etc., so as to simulate its optical properties or inversely retrieve the leaf parameters; at present, the leaf radiation transfer models based on different mathematical physical optical transfer theories mainly include: LEAFMOD model, SLOP (optical property random model) model and flat plate model; the PROSPECT series model is all based on the PLATE model theory, and uses the correlation between the transfer of Stokes' optical quantum to quantitatively describe the relationship between each parameter of N leaf element layers based on the influence parameters such as leaf water absorption coefficient, leaf pigment absorption coefficient, dry matter absorption coefficient and leaf average refractive index. Yu Fenghua et al. based on the flat plate model assumption that the leaf interior is a non-uniform distribution, the leaf internal structure is regarded as the superposition of each layer with different optical properties, and proposed a PIOSL (PROSPECT consider the internal optical structure of the leaves) model, which realizes the simulation of the leaf reflectance radiation transfer process by assuming that the leaf structure is composed of two layers of flat plates with different biochemical parameters and structures, and simulating the complex action of light in the two layers after light incidence, greatly improving the simulation of crop spectral information capability, and the accuracy of crop physiological and biochemical content retrieval is obviously improved.

[0003] Based on the PIOSL model considering internal layer and physiological and biochemical content retrieval, the physiological and ecological functions and states of the leaf can be reflected, but few studies are directed to the influence of the existence of the leaf surface geometric feature structure on the spectrum, thereby causing the simulation spectrum accuracy and physiological and biochemical content retrieval accuracy based on the radiation transfer model; therefore, due to the lack of quantitative research on the PIOSL radiation transfer model in studying the leaf surface geometric features, the light source incidence angle of the model is only one angle perpendicular to the leaf irradiation surface, which limits the application of the model in uneven leaves, and even may affect the extraction of the vegetation canopy pigment information under different light source incidence angles, so that the simulation accuracy of the leaf spectral reflectance is low. SUMMARY

[0004] The embodiment of the present application provides a leaf spectral reflectance simulation method, device, equipment and medium, which can solve the problem of low simulation accuracy of leaf spectral reflectance in the prior art.

[0005] The embodiment of the present application provides a simulation method for leaf spectral reflectance, comprising the following steps:

[0006] According to the optical characteristics of the leaf surface, the diffuse reflection component and the specular reflection component, the geometric feature structure of the leaf surface is acquired, and the leaf light source zenith angle is acquired;

[0007] According to the geometric feature structure of the leaf surface and by using the Ctenophore colony algorithm, the roughness of the leaf surface is acquired;

[0008] According to the diffuse reflection component and the specular reflection component of the leaf surface, light radiation transmission is performed through a first layer of radiation transmission PIOSL model, and then is reflected to the first layer of radiation transmission PIOSL model after being transmitted through a second layer of radiation transmission PIOSL model, the two layers of radiation transmission PIOSL models are fused, and a PIOSL-surf model is constructed;

[0009] The geometric feature structure of the leaf surface, the roughness of the leaf surface and the light source zenith angle are input into the PIOSL-surf model, the total reflectance and the total transmittance of the leaf are acquired, and the leaf simulation spectrum is generated according to the total reflectance and the total transmittance of the leaf.

[0010] Preferably, the geometric feature structure of the leaf surface is acquired, comprising the following steps:

[0011] The leaf is divided into a first leaf unit layer and other leaf unit layers, the optical characteristics of the first leaf unit layer and the optical characteristics of the other leaf unit layers are acquired respectively;

[0012] The reflected radiation energy in the leaf is divided into specular reflection and diffuse reflection, and is specifically divided into a biochemical parameter related diffuse reflection component and a leaf surface structure related specular reflection component;

[0013] The geometric feature structure of the leaf surface is acquired through the optical characteristics of the first leaf unit layer, the biochemical parameter related diffuse reflection component and the leaf surface structure related specular reflection component included in the reflected radiation energy of the leaf.

[0014] Preferably, the leaf light source zenith angle is acquired, comprising the following steps:

[0015] According to the geometric feature structure of the leaf surface, the included angle between the light source incident direction and the zenith direction is acquired, that is, the leaf light source zenith angle is acquired.

[0016] Preferably, the roughness of the leaf surface is acquired, comprising the following steps:

[0017] According to the shape of the leaf, the average roughness of the leaf is determined, and the average roughness is defined as a target function;

[0018] The average roughness is the sum of the absolute values ​​of the height differences between each point on a surface and its average line divided by the measured length;

[0019] The objective function is introduced into the salp swarm algorithm, the parameters of the salp swarm algorithm are set and the algorithm iteration is performed to fit the blade surface roughness, and the result is set as the blade surface roughness.

[0020] Preferably, obtaining the total reflectivity and total transmittance of the blade comprises the following steps:

[0021] Obtain the reflectivity BRF of the interaction between air and the first unit layer of the blade surface;

[0022] Obtain the bidirectional reflectance distribution function (BRDF) representing the reflectance characteristics of the blade surface; the blade surface BRDF is composed of diffuse reflection and specular reflection, and the acquisition method is as follows:

[0023]

[0024] Where: BRDF spec (n, σ, θ i ,θv ， θ ψ ) represents the specular reflection of the blade surface; BRDF diffuse (n, σ, θ i ,θ v ,θ ψ ) represents the diffuse reflection of the blade surface; n is the average refractive index; σ is the surface roughness of the blade;

[0025] The BRF acquisition formula is:

[0026] BRF=π*BRDF spec (n, σ, θ i ,θ v ,θ ψ )+k(1-DHRF spec (n, σ, θ i ))

[0027] Where: k is the coefficient for calculating diffuse reflection on the blade surface;

[0028] The reflectivity BRF of the interaction between the air and the first unit layer of the blade surface and the bidirectional reflectance distribution function BRDF of the blade surface reflection characteristics are input into the PIOSL-surf model to obtain the total reflectivity and total transmittance of the blade.

[0029] Preferably, the BRDF spec (n, σ, θ i ,θ v ,θ ψ ) is obtained as follows:

[0030]

[0031] wherein: F(n, θ α ) represents a reflection factor of the leaf surface; G(θ i , θ v , θ ψ ) represents an attenuation factor of the leaf surface geometry; and D(α, Φ, σ) represents a leaf geometry distribution function;

[0032] The BRDF spec (n, σ, θ i ) is obtained by the following steps:

[0033]

[0034] The embodiment of the present application also provides a simulation device for leaf spectral reflectance, comprising:

[0035] A geometry module is configured to obtain a leaf surface geometry structure according to the leaf surface optical properties, the diffuse reflection component and the specular reflection component, and obtain a leaf light source incident zenith angle;

[0036] A roughness obtaining module is configured to obtain a leaf surface roughness according to the leaf surface geometry structure and by using a sea cucumber colony algorithm;

[0037] A model constructing module is configured to reflect the light radiation transmission through a first layer of radiation transmission PIOSL model, then through a second layer of radiation transmission PIOSL model, and then to the first layer of radiation transmission PIOSL model according to the leaf surface diffuse reflection component and the specular reflection component, fuse the two layers of radiation transmission PIOSL model, and construct a PIOSL-surf model;

[0038] A spectral simulation module is configured to input the leaf surface geometry structure, the leaf surface roughness and the light source incident zenith angle into the PIOSL-surf model, obtain total reflectance and total transmittance of the leaf, and generate a leaf simulation spectrum according to the total reflectance and the total transmittance of the leaf.

[0039] The embodiment of the present application also provides an electronic device, comprising a memory and a processor;

[0040] The memory is configured to store a computer program;

[0041] The processor is configured to execute the computer program stored in the memory, and realize the steps of the simulation method for leaf spectral reflectance.

[0042] The embodiment of the present application also provides a computer readable storage medium for storing a computer program, wherein the computer program is executed by a processor to implement the steps of the simulation method for leaf spectral reflectance.

[0043] The embodiment of the present application provides a simulation method, device, equipment and medium for leaf spectral reflectance, and has the following beneficial effects compared with the prior art.

[0044] The embodiment of the present application is based on the optical characteristics of the leaf surface, and considers the diffuse reflection component related to biochemical parameters and the specular reflection component related to the surface structure, so as to accurately obtain the geometric feature structure of the leaf surface, the leaf surface roughness and the zenith angle of the light source incidence, and simulate the spectral reflectance of the leaf based on the PIOSL-surf model, so as to greatly improve the simulation accuracy of the spectral reflectance of the leaf by increasing the optical characteristics of the specular reflection and the diffuse reflection of the leaf surface, and introducing the leaf surface roughness and the zenith angle of the light source incidence. BRIEF DESCRIPTION OF DRAWINGS

[0045] Figure 1 It is a whole flowchart of the embodiment of the present application;

[0046] Figure 2 It is a schematic diagram of the transmission process of light in the PIOSL model of the embodiment of the present application;

[0047] Figure 3 It is the leaf surface roughness parameter value obtained by fitting based on the Ctenopharynx colony algorithm in the data set 1, the data set 2 and the data set 3 of the embodiment of the present application;

[0048] Fig. 4(a) is a diagram of the 400-2500 waveband simulation spectrum and the measured spectrum of a certain sample in the data set 1 of the embodiment of the present application based on the PROSPECT, PIOSL and PIOSL-surf models;

[0049] Fig. 4(b) is a diagram of the 400-2500 waveband simulation spectrum and the measured spectrum of another sample in the data set 1 of the embodiment of the present application based on the PROSPECT, PIOSL and PIOSL-surf models;

[0050] Fig. 4(c) is a diagram of the 400-2500 waveband simulation spectrum and the measured spectrum of a certain sample in the data set 2 of the embodiment of the present application based on the PROSPECT, PIOSL and PIOSL-surf models;

[0051] Fig. 4(d) is a diagram of the 400-2500 waveband simulation spectrum and the measured spectrum of another sample in the data set 2 of the embodiment of the present application based on the PROSPECT, PIOSL and PIOSL-surf models;

[0052] Fig. 4(e) is a diagram of the simulated and measured spectra of a sample of dataset 3 based on the PROSPECT, PIOSL and PIOSL-surf models in the 400-2500 nm band according to an embodiment of the present application;

[0053] Fig. 4(f) is a diagram of the simulated and measured spectra of another sample of dataset 3 based on the PROSPECT, PIOSL and PIOSL-surf models in the 400-2500 nm band according to an embodiment of the present application;

[0054] Fig. 5(a) is a diagram of the root mean square error of PIOSL-surf compared with PIOSL and PROSPECT in the 400-2500 nm band of dataset 1 according to an embodiment of the present application;

[0055] Fig. 5(b) is a diagram of the root mean square error of PIOSL-surf compared with PIOSL and PROSPECT in the 400-2500 nm band of dataset 2 according to an embodiment of the present application;

[0056] Fig. 5(c) is a diagram of the root mean square error of PIOSL-surf compared with PIOSL and PROSPECT in the 400-2500 nm band of dataset 3 according to an embodiment of the present application;

[0057] Fig. 6(a) is a diagram of the root mean square error of PIOSL-surf compared with PIOSL and PROSPECT in the 400-1000 nm band of dataset 1 according to an embodiment of the present application;

[0058] Fig. 6(b) is a diagram of the root mean square error of PIOSL-surf compared with PIOSL and PROSPECT in the 400-1000 nm band of dataset 2 according to an embodiment of the present application;

[0059] Fig. 6(c) is a diagram of the root mean square error of PIOSL-surf compared with PIOSL and PROSPECT in the 400-1000 nm band of dataset 3 according to an embodiment of the present application. DETAILED DESCRIPTION

[0060] In order to make the above objectives, features and advantages of the present application more clear and easily understood, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings. In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present application. However, the present application can be practiced in a variety of ways beyond the specific embodiments described herein without departing from the scope of the present application, and it is understood that similar improvements can be made by those skilled in the art without departing from the spirit of the present application, and therefore the present application is not limited to the following specific embodiments disclosed.

[0061] Reference Figure 1 The embodiment of the present application provides a simulation method of leaf spectral reflectance, comprising the following steps:

[0062] Step 1: PIOSL (PROSPECT consider the internal optical structure of the leaves) model, such as Figure 2 As shown in the figure, it assumes that the leaf structure is composed of two overlapping layers of flat plates with different biochemical parameters and structures, and simulates the complex effects of light within the two layers after the incident light to simulate the radiation transmission process of the leaf reflectivity, which greatly improves the ability to simulate crop spectral information, and at the same time significantly improves the accuracy of inverting crop physiological and biochemical content; considering the internal stratification and inverting physiological and biochemical content based on the PIOSL model can reflect the physiological and ecological functions and status of the leaves, but there are few studies on the influence of the optical properties of the leaf surface on the spectrum, which leads to the simulation spectrum accuracy and physiological and biochemical content inversion accuracy based on the radiation transmission model; due to the PIOSL The radiation transfer model is insufficient in studying the optical properties of leaf surfaces, which results in the model's light source incident angle having only one angle perpendicular to the leaf radiation surface. This limitation affects the model's application in uneven leaves and may even affect the extraction of vegetation canopy pigment information under different light source incident angles. In order to solve this problem, the present invention considers the diffuse reflection component related to biochemical parameters and the specular reflection component related to the surface structure under the leaf optical radiation transfer framework of the PIOSL radiation transfer model, introduces the leaf surface geometric roughness and the light source incident zenith angle, and more accurately describes the relationship between the leaf surface geometric characteristics and the pigment information band spectrum.

[0063] Step 2: Use the salp swarm algorithm to fit and obtain the blade surface roughness (σ) parameters.

[0064] To fit the blade surface roughness using the PIOSL-surf model, we employed the salp swarm algorithm. First, we defined the mean roughness (the sum of the absolute height differences between each point on a surface and its mean, divided by the measured length) as an objective function. This function calculates the roughness based on the acceptable range of the blade surface and serves as the fitting target for the salp swarm algorithm. Key algorithm parameters were set, and finally, the algorithm was iterated to fit the blade surface roughness.

[0065] The salp swarm algorithm first divides the population into two groups: leaders and followers; the leaders are the salps at the front of the chain, while the rest of the salps are considered followers; as the name of these salps implies, the leaders guide the swarm, and the followers follow each other (as well as the leaders directly or indirectly); similar to other swarm-based techniques, the position of the salps is defined in an n-dimensional search space, where n is the number of variables of the given problem; thus, the positions of all salps are stored in a two-dimensional matrix called x, it is also assumed that there is a food source in the search space called F as the objective of the swarm; the position of the leader is updated, proposing the following equation:

[0066]

[0067] where: represents the position of the first salp (pioneer) in the jth dimension; F j is the position of the food source in the jth dimension; ub j represents the upper bound of the jth dimension; lb j represents the lower bound of the ith dimension; c1, c2, and c3 are random numbers. The director only updates its position with respect to the food source, the c1 coefficient is the most important parameter in the SSA because it balances exploration and exploitation, defined as follows:

[0068]

[0069] where: l is the current iteration, L is the maximum number of iterations. The parameters c2 and c3 are random numbers generated uniformly in the interval [0, 1]; in fact, they determine whether the next position in the jth dimension is towards positive infinity or negative infinity, as well as the step size, to update the position of the driven, the following equation (Newton's law of motion) is used:

[0070]

[0071] when i > 2, represents the position of the ith follower salp in the jth dimension; t is time; v0is the initial velocity; a = v final / v0, where v = (x - x0) / t.

[0072] Since the time in the optimization is the iteration, the difference between iterations is equal to 1, and considering v0= 0, this equation can be expressed as follows:

[0073]

[0074] where: i > 2; represents the position of the ith follower salp in the jth dimension, simulating a salp chain.

[0075] The root mean square error (RMSE) was calculated to assess the difference between the measured reflectance of leaf elements and the simulated reflectance from the PISL-2 model:

[0076]

[0077] where: X meas,j and X mod,j is the measured and simulated reflectance of a leaf j; n is the number of samples.

[0078] Step three: Superimpose the PIOSL model to build the PIOSL-surf model.

[0079] PIOSL model construction principle: The PIOSL model is based on the PROSPECE model as the theoretical basis. This model assumes that the leaf is composed of multiple layers of flat plates with different structures, element content, and absorption coefficients. The leaf is considered as a multi-layered optical property homogenous layer, and the structure, element content, and absorption coefficient of each layer are different. The leaf is divided into two layers, and the applicability of the model is improved by using a numerical continuous iterative method to solve the optical quantitative problem. At the same time, the geometric feature structure of the leaf surface is considered. The geometric feature structure is considered in the first unit layer of the PIOSL model, and the first layer of leaf unit layer and other leaf unit layers are calculated respectively. Finally, the total reflectance and total transmittance of the leaf are simulated. A beam of light is incident on the leaf surface through the air medium and enters the interior of the leaf after a series of reflection processes. After a series of transmission and absorption processes in the interior of the leaf, it finally enters the air, and the reflectance r10 and transmittance t20 are obtained.

[0080] In the PIOSL radiation transfer model, the geometric properties of the leaf surface change the incident angle characteristics of the incident light within the solid angle of the incident light source; specifically, the "V" shape of the leaf surface affects the average transmission of the interface between the air medium within the unit solid angle of the incident light. The range of this change in angle is from 0 to 40° (Feret et al., 2008); based on this change in incident angle, the model obtains the optical properties (average transmittance) of the first layer of leaf unit layer; for other unit layers inside the leaf, the interface incident light of these layers is derived from the light entering the leaf; these light rays are scattered multiple times by various cells inside the leaf; here, you think that this type of light radiation is scattered light, and its characteristics are not affected by the optical properties of the leaf surface. The optical properties of the first layer of leaf unit layer are determined by the geometric features of the leaf surface, while the optical properties of other unit layers of the leaf are not affected by this factor.

[0081] The PIOSL radiation transfer model considers the leaf surface geometric structure, and the first unit layer of the leaf surface is greatly affected by the leaf surface geometric structure, and the optical properties of other unit layers of the leaf are obviously different, so the first unit layer needs to be calculated separately; in the PIOSL model, the leaf surface geometric structure is considered, in order to reduce the error, the optical radiation transfer characteristics of this part need to be calculated separately; Jacquemoud et al. (1990) considered that the incident light is vertically irradiated into the leaf, so the reflectivity and transmissivity of the first layer unit layer of the leaf surface are represented as:

[0082]

[0083] R 1st,l→out (90, n, τ) = R (90, n, τ)

[0084] T 1st，l→out (90, n, τ) = T (90, n, τ) st indicates the first unit layer of the leaf surface; out→1 indicates the total direction of the radiation from the leaf outside to the inside of the leaf; 1→2 indicates the radiation transmission from the air to the leaf; indicates the average reflectivity of the first layer of the leaf unit layer and the air interface when the radiation transmission direction is from the air to the leaf, and the transmissivity can be obtained: In the formula, β is the incident angle of the light source within a unit solid angle, n is the average refractive index, and τ is the average transmissivity; the β incident angle is defined as 0-90°, and β cannot reach 90° due to the influence of the leaf surface geometric structure, therefore, β is defined as 59° in the PIOSL model.

[0085] Let the transmissivity of the second layer be τ2, and the transmitted light after passing through the first layer is used as the incident light of the second layer, which will be absorbed, reflected and transmitted by the upper surface of the second layer. In this study, the first layer and the second layer are assumed to be closely attached, and the energy loss of light between the two layers is ignored, so that T 1st，out→l (β, n, τ) will directly enter the second layer and be absorbed, reflected and transmitted by the second layer; therefore, the transmissivity of the second layer after the action of T1 is:

[0086] T2 = T 1st，out→l (β, n, τ) · τ2

[0087] The reflectivity generated by the action of T1 on the second layer is:

[0088] R2 = T 1st，out→l (β, n, τ) · τ2r 02 τ2

[0089] r 02It is the reflectivity of the lower surface of the leaf when the light reaches the lower surface of the leaf. Since the absorption of light by the lower surface of the leaf is not considered, the relationship between transmittance and reflectivity is:

[0090]

[0091] It can be seen from this that changes in the geometric characteristic structure of the blade surface have a huge impact on radiation transfer models such as PIOSL. The reflectivity and transmittance of the first layer of unit layers affect the total reflectivity and transmittance of the blade. This study considers that the geometric characteristic structure of the blade surface will greatly affect the change of spectral reflectivity. Therefore, PIOSL can improve the accuracy of the model simulation of blade reflectivity and transmittance by introducing the geometric characteristic structure of the blade surface (blade surface roughness).

[0092] PIOSL-surf model construction principle: The PIOSL model considers the influence of the geometric structure of the leaf surface on the optical properties of the leaf, but the model uses fixed values ​​for substitution, which greatly limits the model simulation accuracy and scope of application. In this study, within the framework of the PIOSL model, quantitative leaf surface geometric structure parameters are introduced to improve the spectral accuracy of the model simulation. In the PIOSL model, when light strikes a leaf, it first interacts with the first unit layer of the leaf surface. The geometric structure characteristics of the leaf first unit layer affect the leaf reflectance and transmittance values. However, the absorption characteristics of the first leaf unit layer are usually ignored. Therefore, the average transmittance of the leaf surface can be expressed as the following formula:

[0093]

[0094] The BRF represents the reflectivity of the interaction between the air and the first unit layer of the blade surface. It takes into account the reflectivity of the blade under different illumination angles and observation angles. It provides the ability to comprehensively compare and analyze the reflectivity characteristics of the blade in different directions and is usually used as the basic quantity to define the reflectivity characteristics of the blade (Schaepman-Strub et al., 2006). Defined as:

[0095]

[0096] BRDF (bidirectional reflectance distribution function) can describe the reflection characteristics of the blade surface. The value of BRF is obtained by associating DHRF with BRF through BRDF.

[0097] BRF=π*BRDF

[0098] The BRDF of a Lambertian surface is 1 / π (Schaepman-Strub et al., 2006).

[0099] On the plant leaf surface, the geometric "sharp" light radiation features usually only appear at the leaf edge. When collecting leaf spectral data, the leaf edge area is often ignored; therefore, the reflected radiation energy in the leaf is divided into two types: specular reflection and diffuse reflection; this study will consider the diffuse reflection component related to biochemical parameters and the specular reflection component related to surface structure to construct the improved version of the PIOSL model, the PIOSL-surf radiation transfer model; the PIOSL-surf model considers the diffuse reflection component related to biochemical parameters and the specular reflection component related to surface structure to express the average reflectance of the first layer of leaf elements from air to leaf in the optical radiation transfer of the PIOSL model and construct the optical radiation transfer model considering the leaf surface structure; the leaf surface BRDF is composed of diffuse reflection and specular reflection.

[0100]

[0101] where BRDF spec (n, s, q i , q v , q ψ ) represents the specular reflection of the leaf surface; BRDF diffuse (n, s, q i , q v , q ψ ) represents the diffuse reflection of the leaf surface and is related to the absorption and scattering properties inside the leaf.

[0102] The semi-spherical specular reflection factor DHRFspec is used to express the average reflectance of the leaf surface interface sheet through the equivalent relationship of the interface optical properties to quantify the leaf surface geometric feature structure, and the leaf surface geometric feature structure parameters and the incident light source angle are introduced; DHRF spec (n, s, q i ) can be represented as follows:

[0103]

[0104] According to Bousquet et al. (2005) and Bousquet et al. (2007), the bidirectional reflection distribution function of the dielectric medium and the leaf interface can be expressed as:

[0105]

[0106] where F(n, q α) is the reflection factor of the leaf surface; Bousquet et al. (2005) reported that Fresnel reflection factor can quantify the interface reflection of incident light on the leaf surface, but this quantification is based on the assumption that the leaf surface has very weak absorption ability, which can be ignored.

[0107]

[0108] g 2 = n 2 cos θ α 2 -1

[0109] where G(θ i , θ v , θ ψ ) is the attenuation factor of the leaf surface geometry; this function factor was first introduced into the quantification of specular reflection radiation by Torrance et al., (1967), thus forming the T-S specular reflection model; according to Bousquet et. al 2005 the upper surface of the leaf at the microscale is composed of multiple rough sides with the same area following a normal distribution function, so the cross section of the upper surface geometry of the leaf can satisfy the attenuation function factor in the T-S specular reflection model; for this reason, the attenuation function factor of the dielectric surface geometry to the incident radiation in the radiation transfer model can be expressed as:

[0110]

[0111] where D(α, Φ, σ) is the distribution function of the leaf surface geometry; due to the isotropic characteristics of the leaf, the distribution function of the leaf surface geometry follows a Gaussian normal distribution function, which depends on two parameters, the roughness of the leaf surface (σ) and the angle between the normal of the rough side of the leaf and the normal of the leaf (α).

[0112]

[0113] The relationship between θ α , α, θ s , θ v , θ ψ can be obtained through the trigonometric function relationship of the spatial spherical coordinates as follows:

[0114] cos2θ α = cos θ s cos θ v - cos θ s sin θ v cos θ ψ

[0115]

[0116] The quantification angle of the radiative transfer ray in the equation is θ α and α through θ s , θ v , θ ψ It is shown that θ v , θ ψ can be eliminated in the integration process, and θ s can be used as a quantitative parameter of the incident light source angle, which can be used as an input variable of the radiative transfer model; in the calculation of the diffuse reflection of the leaf surface, the coefficient k can be defined as:

[0117]

[0118] Therefore, we have:

[0119]

[0120] BRDF spec (n, σ, θ i , θ v , θ ψ ) and BRDF diffuse (n, σ, θ i ) can be expressed as:

[0121] BRF = π * BRDF spec (n, σ, θ i , θ v , θ ψ ) + k (1 - DHRF spec (n, σ, θ i ))

[0122] The multi-parameter of the leaf is input into the PIOSL-surf model, the total reflectivity and total transmissivity of the leaf are calculated, and the simulated spectrum diagram is generated.

[0123] In order to more intuitively see the effect of the simulated spectrum of PIOSL-surf, the simulated spectrum of PIOSL-surf, the simulated spectrum of PIOSL model, and the measured spectrum are compared, and the LOPEX93 dataset, ANGERS dataset, and CABO dataset are used as the input of PIOSL, PROSPECT, and PIOSL-surf model, and the leaf roughness fitted to the corresponding dataset sample is used as the input, as shown in Figure 3The three models were then used to forward simulate the reflectance of the sample spectra, and for each model, the RMSE was calculated for each wavelength band to compare the measured and simulated spectral reflectance of the dataset sample for model accuracy evaluation.

[0124] The simulated spectral graphs of the two samples randomly selected from the LOPEX93 dataset, the ANGERS dataset, and the CABO dataset, respectively, as shown in Figure 4, are simulated in the spectral region of 400-2500 nm; in Figure 4(a) 、 4(c) In Figures 4(a), 4(b), and 4(d), the three models are very similar to the measured spectral graphs of the dataset, and the RMSE values are not much different. In Figure 4(c), the RMSE values of the PIOSL-surf model and the PIOSL model are the same, which is 0.0148, and the RMSE value of the PROSPECT model is 0.0203, which is 0.0055 different. In Figure 4(e), the difference is more obvious. The spectral graph of the PIOSL-surf model is more similar to the true spectral graph than the PIOSL model, and the RMSE values are 0.0153, 0.0167, and 0.0350, respectively. In Figures 4(b) and 4(f), the PIOSL-surf model is closest to the actual spectral graph compared with the simulated spectral graphs of the other two models, and the RMSE values are 0.0155 and 0.0247, respectively. The PIOSL-surf model has a lower RMSE in the spectral region of 400-2500 nm, which is better than the PROSPECT model and the PIOSL model, and is more similar to the measured spectral graph of the dataset. However, as can be seen from Figure 4, in the region of 400-1000 nm, the simulated spectral graph of the PIOSL-surf model is almost close to the measured spectral graph, while in the near-infrared region above 1000 nm, the PIOSL-surf model is slightly worse than the other two models, especially in Figure 4(f). The reason may be that the leaf surface is composed of multiple substances (such as chlorophyll, carotenoids, water, etc.), which have different absorption and scattering characteristics in the visible and near-infrared bands. Although the radiation transfer model can simulate the absorption and scattering of these substances on the spectrum, in the long-wave near-infrared band, the absorption characteristics of some substances may become less obvious, resulting in no improvement or slightly lower accuracy of the model simulation.

[0125] The root mean square error (RMSE) was used to evaluate the spectral accuracy of the PIOSL-surf model. A more detailed comparison of the PIOSL model, the PROSPECT model and the PIOSL-surf model (Figures 5 and 6) showed that, in the spectral domain of 400-2500 nm, the average RMSE of the PIOSL-surf model was similar to that of the other two models, and was lower than that of the other two models in the 400-1000 nm region. Figure 5 is the root mean square error of the model simulated spectrum in the 400-2500 spectral band. The RMSE of the PIOSL-surf model based on the LOPEX93 dataset and the ANGERS dataset is lower than that of the PIOSL model and the PROSPECT model, but in the CABO dataset, the RMSE of the PIOSL-surf model is lower than that of the PROSPECT model and higher than that of the PIOSL model. The RMSE of the PIOSL-surf model simulated spectrum ranges from 0.007 to 0.03. The RMSE values of the simulated spectra of the three models are similar, ranging from 0.007 to 0.08. The simulation spectrum capability of the PIOSL-surf model cannot be shown, which is consistent with the results of the 400-2500 nm simulated spectrum figure. Figure 6 is the root mean square error of the model in the 400-1000 spectral band. Overall, the RMSE of the PIOSL-surf model is lower than that of the other two radiation transfer models on the three datasets, and the range is concentrated in 0.0025 to 0.02. The simulation spectrum capability is obviously optimal, and the simulation spectrum capability of the PIOSL-surf model is better than that of the other two models. Among them, the PIOSL-surf model has the minimum RMSE value of 0.0025 and the maximum RMSE value of 0.1357 in the spectral region of 400-1000 nm, which is much smaller than the minimum and maximum RMSE values of the PIOSL model and the PRPOSPECT model.

[0126] From the above data, it can be seen that the RMSE of the PIOSL-surf model in the 400-1000 nm spectral region is greatly reduced compared to the 400-2500 nm spectral region, and the RMSE is extremely low compared to the other two radiation transfer models. Combined with the simulation spectrum results in the spectral region of 400-2500 nm, it is verified that the simulation spectrum accuracy of the PIOSL-surf model in the 400-1000 nm spectral region is higher than that of the PIOSL model and the PROSPECT model, and it is concluded that the PIOSL-surf model can better represent the optical properties in the 400-1000 nm spectral domain.

[0127] The traditional PROSPECT radiative transfer model considers a leaf as a structure of N layers of closely homogeneous leaf element layers and N-1 layers of air uniformly spaced, and solves the optical quantitative problem in homogeneous and heterogeneous media by using a numerical continuous iterative method based on the assumption that the leaf is composed of optically isotropic plates, while considering the geometric characteristics of the leaf surface at the microscopic scale. The PIOSL radiative transfer model is constructed based on the actual leaf cross-section scanning electron microscope results and the theoretical construction of the internal structure of the leaf. The model assumes that the leaf is composed of two structures with different optical properties. The structure parameters and material content distribution of these two layers are different, resulting in differences in their optical properties. N1 and N2 are used to represent the two structure parameters, although these parameters are not an accurate description of the actual physical layering, but they do help to clarify how the differences in the internal structure of the leaf lead to the layering phenomenon of optical properties. When the direction of illumination changes, the reflection and transmission behavior of the leaf will be affected to varying degrees, which involves more complex physical and optical mechanisms. Non-zenith direction light source will cause the optical behavior of the leaf at different angles to change, thereby affecting the transmission of light in the leaf. The existing PROSPECT and PIOSL models are mainly based on vertical incident light, and the incident angle and direction of the non-zenith light source. The PROSPECT and PIOSL radiative transfer models are mainly used for quantitative calculation of leaf optical radiation transfer. They are based on the zenith direction light source, considering the geometric characteristics of the leaf structure and the surface optical properties to simulate the transmission of light in the leaf. However, these models are mainly suitable for zenith direction light source, and the applicability of the model may be limited for light from other non-zenith light source directions. Among them, the surface structure parameters of the leaf, such as refractive index and roughness, will be affected by a variety of factors. The morphology and arrangement of wax on the leaf cuticle, the type of plant, and environmental conditions can all affect the surface structure parameters of the leaf. These surface structure parameters further determine the distribution of the leaf's directional reflection, and how this distribution changes with changes in the illumination and observation geometry; this study will consider the leaf surface optical properties to construct a new radiative transfer model PIOSL-surf model.

[0128] The PIOSL-surf radiation transfer model established by the application is based on the leaf surface geometric structure, the PIOSL-surf radiation transfer model is constructed based on the PIOSL radiation transfer model, the influence of the leaf surface geometric structure on the optical properties is considered under the theory of dividing the leaf into two layers, and the difference of the leaf surface optical properties is described. Compared with the PIOSL radiation transfer model, the PIOSL-surf radiation transfer is realized by increasing the leaf surface specular reflection and diffuse reflection optical properties and introducing the leaf surface geometric structure parameters (roughness). Therefore, the PIOSL-surf radiation transfer model can provide a large number of observed reflectance simulation of the leaf, and it is proved that the PIOSL-surf radiation transfer model is superior to the PROSPECT radiation transfer model and the PIOSL radiation transfer model, and the accuracy is higher, especially in the visible and near-infrared spectral domain.

[0129] The above-mentioned embodiments only express several embodiments of the application, the description is more specific and detailed, but it cannot be understood as the limitation of the patent scope of the application. It should be pointed out that for ordinary skilled in the art, without departing from the concept of the application, several modifications and improvements can be made, which belong to the protection scope of the application. Therefore, the protection scope of the patent of the application should be subject to the appended claims.

Claims

1. A method for simulating leaf spectral reflectance, characterized in that: The following steps are involved: According to the optical characteristics, diffuse reflection component and specular reflection component of the blade surface, the geometric characteristic structure of the blade surface and the incident zenith angle of the blade light source are obtained; The surface roughness of the blade is obtained based on the geometric characteristic structure of the blade surface and using the salp swarm algorithm; According to the diffuse reflection component and specular reflection component of the blade surface, the light radiation is transmitted through the first layer of radiation transmission PIOSL model, then through the second layer of radiation transmission PIOSL model and reflected back to the first layer of radiation transmission PIOSL model. The two layers of radiation transmission PIOSL models are fused to construct the PIOSL-surf model. The blade surface geometric structure, blade surface roughness and light source incident zenith angle are input into the PIOSL-surf model to obtain the total reflectivity and total transmittance of the blade. And generate a leaf simulation spectrum according to the leaf total reflectivity and total transmittance; The PIOSL model assumes that the leaf structure is composed of two overlapping flat plates with different biochemical parameters and structures, and simulates the complex effects of light incident on the two layers to achieve the simulation of the radiation transmission process of leaf reflectivity.

2. The method for simulating blade spectral reflectance according to claim 1, characterized in that: The method of obtaining the geometric characteristic structure of the blade surface comprises the following steps: The blade is divided into a first blade unit layer and other blade unit layers, and the optical properties of the first blade unit layer and other blade unit layers are respectively obtained; The reflected radiation energy in the leaves is divided into specular reflection and diffuse reflection; specifically, it is divided into diffuse reflection components related to biochemical parameters and specular reflection components related to the leaf surface structure; The geometric characteristic structure of the blade surface is obtained through the optical properties of the first blade unit layer of the blade, as well as the diffuse reflection component related to the biochemical parameters included in the blade reflected radiation energy and the specular reflection component related to the blade surface structure.

3. The method for simulating blade spectral reflectance according to claim 1, characterized in that: The obtaining of the incident zenith angle of the blade light source comprises: According to the geometric characteristic structure of the blade surface, the angle between the incident direction of the light source and the zenith direction is obtained, that is, the incident zenith angle of the blade light source is obtained.

4. The method for simulating blade spectral reflectance according to claim 1, characterized in that: The method of obtaining the surface roughness of the blade comprises the following steps: According to the blade shape, the average roughness of the blade is determined and the average roughness is defined as the objective function; The average roughness is the sum of the absolute values ​​of the height differences between each point on a surface and its average line divided by the measured length; The objective function is introduced into the salp swarm algorithm, the parameters of the salp swarm algorithm are set and the algorithm iteration is performed to fit the blade surface roughness, and the result is set as the blade surface roughness.

5. The method for simulating blade spectral reflectance according to claim 1, characterized in that: The method of obtaining the total reflectivity and total transmittance of the blade comprises the following steps: Obtain the reflectivity BRF of the interaction between air and the first unit layer of the blade surface; Obtain the bidirectional reflectance distribution function (BRDF) representing the reflectance characteristics of the blade surface; the blade surface BRDF is composed of diffuse reflection and specular reflection, and the acquisition method is as follows: Where: represents the specular reflection of the blade surface; represents the diffuse reflection of the blade surface; n is the average refractive index; σ is the surface roughness of the blade; The BRF acquisition formula is: Where: k is the coefficient for calculating diffuse reflection on the blade surface; The reflectivity BRF of the interaction between the air and the first unit layer of the blade surface and the bidirectional reflectance distribution function BRDF of the blade surface reflection characteristics are input into the PIOSL-surf model to obtain the total reflectivity and total transmittance of the blade.

6. The method for simulating blade spectral reflectance according to claim 5, characterized in that: described The formula for obtaining is: in: represents the reflectivity factor of the blade surface; Attenuation factor representing the geometric characteristics of the blade surface; represents the distribution function of blade geometric characteristics; described The steps to obtain are: 。 7. A device for simulating leaf spectral reflectance, characterized in that: include: A geometric structure module is used to obtain the geometric characteristic structure of the blade surface and the incident zenith angle of the blade light source according to the optical characteristics, diffuse reflection component and specular reflection component of the blade surface; A roughness acquisition module is used to obtain the surface roughness of the blade according to the geometric characteristic structure of the blade surface and using the salp swarm algorithm; A model construction module is used to construct a PIOSL-surf model by fusing the two layers of radiation transmission PIOSL models based on the diffuse reflection component and the specular reflection component of the blade surface. The light radiation is transmitted through the first layer of radiation transmission PIOSL model, then reflected back to the first layer of radiation transmission PIOSL model after passing through the second layer of radiation transmission PIOSL model. The spectral simulation module is used to input the blade surface geometric structure, blade surface roughness and light source incident zenith angle into the PIOSL-surf model to obtain the total reflectivity and total transmittance of the blade; And generate a leaf simulation spectrum according to the leaf total reflectivity and total transmittance; The PIOSL model assumes that the leaf structure is composed of two overlapping flat plates with different biochemical parameters and structures, and simulates the complex effects of light incident on the two layers to achieve the simulation of the radiation transmission process of leaf reflectivity.

8. An electronic device, characterized in that: include: memory and processor; The memory is used to store computer programs; The processor is configured to implement the steps of a method for simulating blade spectral reflectance as described in any one of claims 1 to 6 when executing the computer program stored in the memory.

9. A computer-readable storage medium, characterized in that Used to store a computer program, which, when executed by a processor, implements the steps of a method for simulating the spectral reflectance of a blade as described in any one of claims 1 to 6.

Images