A universal inversion model and method for leaf chlorophyll content in different types of vegetation

Abstract

The present application relates to a general inversion model and method for leaf chlorophyll content of different types of vegetation. The general inversion model for leaf chlorophyll content is expressed as the relationship between the chlorophyll content of plant leaves and the reflectance spectrum and chlorophyll absorption coefficient of vegetation leaves in the optimal spectral range. The simulation data set is constructed by the PROSPECT radiation transmission model, and the SAC similarity index is calculated by the SAC similarity index calculation formula. The spectral band resolution of the reflectance spectrum and chlorophyll absorption coefficient of vegetation leaves is 1nm. Through research and analysis, the present application finds the wavelength region where the change of chlorophyll content has the greatest impact on the change of reflectivity. The SAC similarity distance measurement method is used in this wavelength region to calculate the similarity chlorophyll index, thereby constructing a CSSI chlorophyll general inversion model, achieving high-precision inversion of chlorophyll in vegetation leaves, improving the inversion and prediction accuracy, and expanding the scope of application.

Description

Technical Field

[0001] The present application relates to a method for utilizing vegetation leaf reflectance spectral data to calculate a spectral similarity measurement index between a leaf reflectance spectrum and a chlorophyll absorption coefficient spectrum, replacing measured data with simulated data generated based on a radiation transfer model, and correcting and constructing a general inversion model for chlorophyll content in vegetation leaves based on the spectral similarity measurement index and verifying and evaluating the inversion model. The inversion model can be used to quantitatively invert the chlorophyll content of multiple types of vegetation leaves with high precision. Background Art

[0002] Leaf chlorophyll is the core pigment of plant photosynthesis. It is directly involved in the capture and conversion of light energy and is the basis of plant growth and metabolism. Therefore, the quantitative analysis of leaf chlorophyll not only helps to understand the photosynthesis efficiency of plants, but also provides a scientific basis for crop management, ecosystem monitoring and environmental change assessment. Traditionally, chlorophyll content is mainly measured through field sampling and laboratory chemical analysis, which is a time-consuming and complicated method, and is also limited by time and space scales. Studies have shown that it is feasible to invert vegetation biochemical parameters through reflectance spectra obtained by remote sensing measurements. At the same time, this method has the characteristics of real-time, non-destructive, and wide detection range, which greatly makes up for the limitations of traditional methods.

[0003] In the field of remote sensing, most studies on the inversion of chlorophyll content in vegetation leaves rely mainly on vegetation indices calculated through reflectance. This is because the increase or decrease of chlorophyll content in leaves will affect the absorption of blue and red light bands, thereby changing the reflectance of vegetation leaves. The reflectance spectrum used in the vegetation index obtained in the study is also concentrated in the blue and red light bands. For example, through literature research, the modified normalized difference index mND 705 , modified simple quantitative index mSR 705 , the new three-band index NTBI, the DATT index, the MDRI index, etc. These indices all show a significant linear relationship with the chlorophyll content, and can be used to build a more accurate chlorophyll inversion model based on the measured data. However, these models also have limitations, such as the decrease in accuracy when inverting the chlorophyll content of leaves of multiple species, and the saturation that easily occurs when inverting the chlorophyll content of leaves, resulting in deviations in the inversion results. The reasons for these problems are, first, that these indices are too concentrated in specific spectral bands and fail to fully capture the overall changes in chlorophyll; second, because these index models are constructed based on measured data collected by the group, these measured data sets are limited, resulting in the model may only be applicable to the current sample, and the effect is not good when applied to samples of different vegetation types.

[0004] In order to solve the problem of limited accuracy caused by the low universality of the above model, some studies use simulated data generated by the radiation transfer model (PROSPECT radiation transfer model) instead of measured data for model correction, which improves the universality of the model. However, the study found that the inversion prediction effect of the chlorophyll index on the simulated data set is very good, but when the model fitted by the simulated data set is applied to the measured data set, a large error will occur. This is because the difference between the simulated data set and the measured data set is ignored, or the chlorophyll index cannot express the spectral information of the reflected chlorophyll well, which leads to a decrease in the accuracy of the inversion results of the model. There is still no ideal solution to this problem. In summary, a universal remote sensing extraction method for leaf chlorophyll height suitable for different vegetation types is still an urgent need. Summary of the invention

[0005] The technical problem solved by the present application is to overcome the above-mentioned deficiencies in the prior art and to provide a universal inversion model and method for leaf chlorophyll content of different types of vegetation, so as to quantitatively invert the chlorophyll content of leaves of multiple types of vegetation with high precision, and provide key information for carbon cycle and biomass estimation of terrestrial ecosystems.

[0006] The main principles of this application are as follows:

[0007] The absorption and reflection of light by leaves is the result of the combined effect of the content and structure of leaf components. It can be regarded as a comprehensive expression of the overall absorption coefficient of the leaf, which is obtained by weighted average of the absorption coefficients of different components in the leaf. It should be noted that the absorption coefficient k of each component is closely related to the wavelength, but has nothing to do with the type of vegetation. In addition, the absorption coefficient of each component usually shows sensitive absorption characteristics within a specific wavelength range. For example, water effectively absorbs radiation in the 900-2500 nm region, but the absorption in other regions is not obvious; chlorophyll mainly absorbs radiation effectively in the 400nm-800nm ​​region. It can be deduced that in certain specific wavelength regions, if the absorption coefficient of a certain component changes strongly relative to the absorption coefficient of other components, then in this region, there must be a very strong similarity relationship between the change in reflectivity and the absorption coefficient of the component. This similarity relationship can also be used as an evaluation indicator for evaluating this component. Through research and analysis, this application finds the wavelength region where the change in chlorophyll content has the greatest impact on the change in reflectance, and then uses the SAC similarity distance measurement method to calculate the similarity chlorophyll index in this wavelength region, thereby constructing the CSSI chlorophyll universal inversion model to achieve high-precision inversion results of vegetation leaf chlorophyll.

[0008] The general inversion model for leaf chlorophyll content of different types of vegetation in this application meets the following requirements:

[0009] LCC CSSI= a·SAC+ b = a·SAC( r (λ) w (λ) [OSI ] + b(2)

[0010] LCC refers to the chlorophyll content of plant leaves. r (λ), w (λ) represents the reflectance spectrum of vegetation leaves with consistent spectral resolution and the chlorophyll absorption coefficient of vegetation leaves, respectively. OSI is the optimal spectral interval. The coefficients a and b are calculated from the simulated data set, which is constructed by the PROSPECT radiation transfer model. The SAC similarity index calculation formula is as follows;

[0011]

[0012] Where: SAC similarity index is the spectral angle cosine distance, where r and w refer to the reflectance spectrum and chlorophyll absorption coefficient of the leaves, respectively. L is a spectral interval. In the CSSI general inversion model, this interval is the optimal spectral interval OSI. l It is a band in the optimal spectral range, and the spectral band resolution of the leaf reflectance spectrum and chlorophyll absorption coefficient is 1nm.

[0013] The PROSPECT radiation transfer model adopts the PROSPECT-D radiation transfer model.

[0014] The SAC similarity index values ​​in the simulated data set only retain data corresponding to the SAC similarity index values ​​in the measured data set.

[0015] The method for constructing the universal inversion model of leaf chlorophyll content for different types of vegetation in this application is as follows:

[0016] S1: Construct a sample data set (the sample data set includes a simulated data set generated by the PROSPECT radiation transfer model and a measured data set when necessary). The spectral band resolution of the reflectance spectrum and chlorophyll absorption coefficient in the sample data set is 1 nm.

[0017] S11: Generate simulation data sets using the PROSPECT radiative transfer model;

[0018] Steps S12 and S13 of the present application are optional steps and are only used when it is necessary to use the measured data set to correct the simulated data set or to verify and evaluate the inversion effect of the CSSI chlorophyll universal inversion model.

[0019] S12: The vegetation leaf samples collected in the field are measured and analyzed in the laboratory to construct a measured data set, which includes: reflectance spectra of fresh leaves and chlorophyll content of fresh leaves;

[0020] S13: Obtain the leaf chlorophyll absorption coefficient of the measured data set. This part can be measured by yourself or obtained from literature.

[0021] S2: Calculate the optimal spectral interval;

[0022] S21: According to formula (1), the leaf similarity chlorophyll index and the SAC similarity index in the range with high correlation between chlorophyll (400nm-800nm) are calculated;

[0023] (1)

[0024] In formula (1), the value range of SAC similarity index (spectral angle cosine distance) is between 0 and π, where r and w refer to the reflectance spectrum and chlorophyll absorption coefficient of the leaf, respectively. L represents a spectral interval, l It is a band in this spectral range;

[0025] S22: Correlation analysis was performed on the interval with high correlation between leaf SAC similarity index and chlorophyll content;

[0026] S23: The spectral interval with the best correlation between SAC similarity index and chlorophyll content is defined as the optimal spectral interval (OSI);

[0027] S3: Construct the CSSI (Chlorophyll Similarity Spectral Index) chlorophyll universal inversion model;

[0028] S31: Obtain the SAC similarity index value of the simulated data set in the optimal spectral range, analyze the correlation expression between the SAC similarity index and its corresponding leaf chlorophyll content, and determine the construction expression of the CSSI chlorophyll general inversion model;

[0029] S33: Construction of a CSSI chlorophyll general inversion model using simulated datasets.

[0030] In the present application, step S32 may be set between step S31 and step S33, wherein S32 is to screen the SAC similarity index values ​​of the simulated data set, and only retain the SAC similarity index values ​​of the simulated data set corresponding to the SAC similarity index values ​​in the measured data set; on the one hand, this step greatly reduces the amount of data and calculation, and on the other hand, it also makes the simulated data set closer to the measured data set, which helps to improve the inversion accuracy of plant varieties related to the measured data set.

[0031] The verification and evaluation method of the universal inversion model of leaf chlorophyll content for different types of vegetation in this application is as follows:

[0032] S4: Use the CSSI chlorophyll universal inversion model to predict the inverted chlorophyll content;

[0033] S5: Evaluating the performance of the CSSI chlorophyll general inversion model.

[0034] The verification and evaluation method of the universal inversion model of leaf chlorophyll content for different types of vegetation in this application may also include the following (comparison) steps:

[0035] S6: Compare the CSSI chlorophyll universal inversion model with other chlorophyll inversion models.

[0036] The S6 step includes:

[0037] S61: Setting the chlorophyll spectral index comparison model;

[0038] S62: Use formula (2) to construct a chlorophyll inversion comparison model corresponding to the chlorophyll spectral index comparison model;

[0039] S63: Use the chlorophyll inversion comparison model to obtain the predicted inversion results on the measured data set;

[0040] S64: Discuss and analyze the inversion verification results and compare them with the inversion results of the CSSI chlorophyll general inversion model.

[0041] The CSSI chlorophyll universal inversion model described in this application covers as many vegetation samples as possible in reality, and the optimal spectral range is basically determined, so it is highly adaptable and can be updated very conveniently. In general, it is only necessary to add the newly collected vegetation sample parameters to the original vegetation sample parameters and regenerate the simulated data set using the PROSPECT radiation transfer model.

[0042] This application uses the coefficient of determination (R 2 ) and normalized root mean square error (nRMSE%) were used to evaluate the performance of the chlorophyll inversion comparison model and compared with the CSSI chlorophyll general inversion model.

[0043] The PROSPECT radiation transfer model described in this application simulates the reflectance spectra of different vegetation leaves by inputting leaf trait data, and these traits include leaf chlorophyll content, leaf dry matter content, leaf equivalent water thickness, leaf carotenoid content, leaf brown pigment content, leaf structural parameters, and anthocyanin content. In order to make the simulated data set more representative, this application obtains the size range of normal vegetation traits in nature by consulting existing public literature, traverses these size ranges, and generates a representative vegetation simulation data set. The generated simulation data set sample includes: leaf reflectance spectrum and corresponding leaf chlorophyll content.

[0044] The step of performing correlation analysis on the interval with high correlation between leaf similarity index and chlorophyll content uses Pearson coefficient for correlation analysis, and the SAC similarity index of each band is put into a matrix, the horizontal axis of the matrix represents the starting band, the vertical axis represents the ending band, and the point in the matrix represents the Pearson coefficient value of the SAC similarity index and chlorophyll content obtained in the spectral interval from the starting band to the ending band of the point.

[0045] The chlorophyll content of fresh leaves in the present application is measured by reagent extraction.

[0046] Compared with the prior art, the present invention has the following advantages and effects: 1. The calculation of the SAC index used in the CSSI chlorophyll universal inversion model is not complicated; 2. The SAC index can be obtained by only obtaining the reflectance spectrum of the vegetation to construct the CSSI chlorophyll universal inversion model, and the operation is simple; 3. The inversion accuracy of the CSSI chlorophyll universal inversion model is very high. In the evaluation results of the measured data set of the CSSI chlorophyll universal inversion model, R 2 is 0.8022 (range 0-1, the closer to 1, the more correlation exists), and nRMSE (%) is 13.56% (the smaller the percentage, the higher the inversion accuracy). 4. Compared with the results of other vegetation models on measured data sets and simulated data sets, the inversion effect of the CSSI chlorophyll universal inversion model is also better, which is mainly reflected in the following: The CSSI chlorophyll universal inversion model and other chlorophyll indices were inverted and predicted on the same five measured data sets, regardless of R 2 5. Combining the above results, it can be considered that the CSSI chlorophyll universal inversion model, as a new model, not only has higher inversion accuracy, but also can achieve good prediction accuracy in multiple vegetation types, has stronger universality, and is superior to the existing chlorophyll inversion models. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1It is a spectral interval importance contour map. Among them, A, B, C, D, and E are the spectral interval importance contour maps of five field-measured vegetation sample data sets (referred to as measured data sets) from different teams, namely ANGERS, BM, JTL, NX, and XS. F is the spectral interval importance contour map of the PROSPECT vegetation simulation sample data set (referred to as simulated data set) generated by the PROSPECT-D model (the PROSPECT model can simulate vegetation spectra and is representative). G is a combination of Figure 1 The importance of the spectral intervals of the six data sets AF in the above figure is used to obtain the contour map of the total spectral importance. It can be seen intuitively that the SAC similarity index calculated in the spectral interval of 727-798nm is most related to vegetation chlorophyll, so it is confirmed as the optimal spectral interval (OSI) of SAC.

[0048] Figure 2 It is a linear expression diagram obtained by polynomial fitting of the SAC similarity index value calculated by constructing the reflectance spectrum of the sample data set in the optimal spectral interval (OSI) and the corresponding sample leaf chlorophyll content (LCC). Figure 1 A, B, C, D, and E are the linear expression graphs of the five measured data sets, ANGERS, BM, JTL, NX, and XS, respectively. (F) is the linear expression graph of the PROSPECT-D simulated data set. Figure 2 The six linear equations in the figure are intuitive expressions of the SAC similarity index and vegetation chlorophyll content of the corresponding sample data set, and the coefficients of the linear equations (i.e. the value before x in the equation and the constant after x) are used as model parameters to construct the CSSI chlorophyll general inversion model for the sample data set.

[0049] Figure 3The results of the inversion of the CSSI chlorophyll general model of the simulated data set constructed by the SAC similarity index calculated for the reflectance spectrum of the PROSPECT-D simulated data in the best spectral interval on five measured data sets. Among them, A, B, C, D, and E correspond to the inversion results of the CSSI chlorophyll general model on the five measured data sets of ANGERS, BM, JTL, NX, and XS. Due to the differences between the simulated data set and the measured data obtained by manual measurement, mainly due to the instrument and noise, there are inevitable errors in the manual measurement. At the same time, due to the limitations of the model, the simulated data set cannot perfectly simulate the reflectance spectrum of vegetation and can only be approximately simulated. Therefore, the results of the inversion of the CSSI chlorophyll general inversion model fitted by PROSPECT on the measured data set have a certain offset. However, it can be seen that although the CSSI chlorophyll general inversion model is constructed using PROSPECT-D, the nRMSE (%) of the model on the measured data set shows that the inversion error is not large. It shows that using the SAC similarity index, the universal inversion of different vegetation types can be achieved through the PROSPECT simulated data set.

[0050] Figure 4 This is a schematic diagram of the comprehensive inversion results, which expresses the comprehensive inversion results of the CSSI chlorophyll general inversion model constructed using only the PROSPECT-D simulation dataset on the validation sets of five measured datasets.

[0051] Figure 5 The results of the inversion of six chlorophyll inversion comparison models on the PROSPECT-D simulation data set are compared. H, I, J, K, L, and M in the figure correspond to the inversion results of the models constructed by six chlorophyll indices (referred to as comparison models). These six vegetation indices are currently the most accurate vegetation indices in the field of chlorophyll research. The ROSPECT-D simulation data set was used when constructing the six comparison models to better reflect the comparison. DETAILED DESCRIPTION

[0052] The meaning of the English name or abbreviation of this application is as follows:

[0053] SAC - Spectral Angle Cosine Distance

[0054] ANGERS

[0055] BM-Baima (place name, located in Nanjing)

[0056] JTL-Jingtian Road (place name, located in Nanjing)

[0057] NX-Nanxin (short for the place name, experimental base of Nanjing University of Information Science and Technology)

[0058] XS-Xiashu (place name, located in Jiangsu)

[0059] The above five measured place names serve as identifiers for the corresponding measured data sets.

[0060] The present application is further described in detail below in conjunction with the accompanying drawings. The main steps of the construction and verification evaluation method of the CSSI chlorophyll universal inversion model of the present application are as follows:

[0061] S1: Construction of sample dataset.

[0062] S11: Generate a simulated data set using the PROSPECT-D radiation transfer model. This step is a prior art, which mainly simulates the plant leaf reflectance spectrum by inputting the normal values ​​of the biochemical parameters of natural plant leaves as wide as possible. At the same time, when generating, since the radiation absorption of the equivalent water thickness biochemical parameter is not within the chlorophyll radiation absorption range, it is set to a fixed value to reduce the generation time of the simulated data set. The final generated PROSPECT-D simulation data set contains the following data: the simulated vegetation sample reflectance spectrum and the corresponding chlorophyll content during the simulation, which correspond to the reflectance spectrum of the field-measured leaf sample data set and the chlorophyll in the sample. The reason for using the PROSPECT-D simulation data set is that it can basically reflect the physiological characteristics of all measured data set vegetation, and can be considered as a special field measurement data set. At the same time, in order to make the PROSPECT-D simulation data set more representative, this application uses the PROSPECT-D radiation transfer model, and according to the literature survey, determines the range of normal values ​​of natural leaf biochemical parameters (Table 2), and uses it as input to make the PROSPECT-D simulation data set cover the vegetation samples in reality as much as possible. After inputting these parameters into the PROSPECT-D radiative transfer model, a total of 47,142 simulated data sets were generated;

[0063] S12: Collect five leaf sample datasets measured in the field. These leaf sample datasets measured in the field contain more than one type of vegetation samples, many of which are widely used in chlorophyll vegetation research. These five datasets all collect reflectance data of several vegetations through an spectral acquisition device (ASD) with a resolution of 1nm, which is consistent with the absorption coefficient resolution used in the application. At the same time, all five datasets contain the chlorophyll content value of each sample, which is specifically recorded in Table 1.

[0064] S13: Obtain the chlorophyll absorption coefficient from the literature. The absorption coefficient spectrum resolution in the literature is 1 nm.

[0065] S2: Calculate the optimal spectral interval of all sample data sets;

[0066] S21: From step S1, the reflectance spectrum of the field-measured sample and the chlorophyll absorption coefficient can be obtained. According to the following formula (1), the SAC similarity index can be calculated. This step needs to be performed on all bands in the range (400-800nm) where the correlation between the leaf similarity chlorophyll index and chlorophyll is relatively high in the five measured data sets and the PROSPECT-D simulation data set, ensuring that the starting band is smaller than the ending band. Among them, the five measured data sets are added to obtain the best spectral interval with relatively comprehensive representativeness, and to prove that this spectral interval has a correlation with the leaf chlorophyll content (LCC) in both the measured data set and the PROSPECT-D simulation data set.

[0067] (1)

[0068] Formula (1): The range of the spectral angle cosine distance is between 0 and π. The smaller the value, the higher the similarity between the two vectors and the smaller the angle. When the two vectors are exactly the same, the spectral angle cosine distance is 0; when the angle between the two vectors is 180 degrees, the spectral angle cosine distance is π. Where r and w refer to the reflectance spectrum and chlorophyll absorption coefficient, respectively. L Indicates the number of bands contained in a spectral interval. l It is a band in this spectral range. For example, L represents all spectral bands in the range of 600-700nm. l is 605nm, then r l w l It can be expressed as the product of the reflectance spectrum value in the 605nm band and the chlorophyll absorption coefficient.

[0069] S22: Then the SAC similarity index of each band is correlated with the chlorophyll content, that is, the Pearson coefficient of the two is calculated. According to the interval length, the results are placed in a 400*400 plane matrix. The results of the five measured data sets and the PROSPECT-D simulation data set are shown in Figure 2. Figure 1 As shown in AF, the horizontal axis represents the starting band, the vertical axis represents the ending band, and a point in the matrix represents the Pearson coefficient value of SAC and chlorophyll from the starting band to the ending band (the obtained Pearson coefficient ranges from -1 to 1, and the closer to 1, the larger the value).

[0070] S23: Then, the importance of the Pearson coefficients obtained after traversing the intervals with high correlation is distinguished, and then the best spectral interval is selected according to the importance. The specific operation is to sort all the Pearson coefficients obtained, filter out the corresponding bands of the top 15%, 10%, 5%, and 1% according to the value (-1 to 1), and assign colors of different importance to distinguish them. Therefore, it can be obtained from Figure 1 It is intuitive to see which bands have correlations with the SAC similarity index and chlorophyll in different data sets. Finally, based on the comprehensive performance of five measured data sets from different teams and containing different vegetation types, we selected the SAC similarity index obtained in the 727-798nm sub-range as the most correlated with chlorophyll. Figure 1 Derived from this, Figure 1 From the importance level, we can see that the five measured data sets and the PROSPECT-D simulation data set are the most relevant areas in this interval (R 2 Therefore, 727-798nm is confirmed as the optimal spectral interval (OSI).

[0071] S3: Based on the OSI optimal spectral range, a CSSI (Chlorophyll Similarity Spectral Index) chlorophyll general inversion model is constructed.

[0072] S31: From step S2, we can see that in the OSI interval, the similarity measure between the calculated vegetation reflectance spectrum and the chlorophyll absorption coefficient has the greatest correlation with the chlorophyll content. In order to better build the model, it is necessary to find the correlation expression between the SAC similarity index value and the corresponding LCC in the optimal spectral interval. Therefore, we analyze the correlation expression of the measured data sets and simulated data sets of five different teams, and the results are shown in Figure 2 The coefficient of determination (R 2 ), SAC has a strong negative correlation with leaf chlorophyll, and the trends of the five measured data sets and the PROSPECT-D simulation data set are the same. It is deduced that for the SAC similarity index value, it is suitable to use polynomial fitting to construct a universal chlorophyll inversion model, namely the CSSI chlorophyll universal inversion model, to achieve the inversion prediction of chlorophyll in measured samples. Among them, CSSI is the chlorophyll similarity spectral index, which refers to the SAC similarity index value in the optimal spectral interval.

[0073] S32: Due to the large differences between various samples (such as differences caused by different biochemical parameter testing methods and different reflectance spectroscopy testing instruments), the inversion model constructed based on only a few samples will have the problem of non-transferability or weak transferability (i.e., universality). In order to make the CSSI chlorophyll universal inversion model universal, the data sample set used to build this model needs to cover multiple vegetation and multiple types of vegetation information, so the PROSPECT simulation data set is used to solve this problem. At the same time, before building the model, the characteristics of the PROSPECT-D simulation data set must also be considered. It has a large amount of data and covers a wide range of biochemical parameters, which basically represents the physiological characteristics of all vegetation. In order to make the physiological characteristics of the simulation data set more concentrated on the corresponding field measurement data set and reduce the impact of irrelevant information. Therefore, before measuring the data set, the PROSPECT-D simulation data set can be screened according to the measured data set as a preferred step. For example, in this application, the SAC similarity index values ​​of the PROSEPCT-D simulation data set are screened according to the ranges of the SAC similarity index values ​​corresponding to the five measured data sets, and only the SAC similarity index values ​​of the simulation data set corresponding to the SAC similarity index values ​​of the measured data set are retained, so that the SAC similarity index of PROSPECT-D is closer to the SAC similarity indexes of the five measured data sets.

[0074] S33: Based on the PROSPECT-D simulation data set screened in step S32, the polynomial fitting method is used to find the best linear expression result of the SAC similarity index and construct the CSSI chlorophyll universal inversion model. The model is expressed as follows:

[0075] LCC CSSI = a·SAC+ b = a·SAC PROSPECT-D ( r (λ) w (λ) [OSI ] + b(2)

[0076] in r (λ), w (λ) represents the vegetation reflectance spectrum and vegetation chlorophyll absorption coefficient with consistent spectral resolution, respectively. OSI is the optimal spectral interval selected according to the operation in step S2 above. The coefficients a and b are obtained by the SAC similarity index calculation formula of the screened PROSPECT-D simulation data set.

[0077] S4: Use the CSSI chlorophyll general inversion model to predict and invert leaf chlorophyll content.

[0078] The SAC similarity index values ​​of the five measured data sets within the OSI interval were put into the CSSI chlorophyll general inversion model, and the predicted chlorophyll content results of the corresponding samples were obtained.

[0079] S5: Evaluating the performance of the CSSI chlorophyll general inversion model.

[0080] The coefficient of determination (R 2 ) and normalized root mean square error (nRMSE%) were used to analyze the inversion results, such as Figure 4 The results show that the CSSI general vegetation index inversion model constructed using the PROSPECT-D simulated dataset has good inversion accuracy. The R 2 They all reflect a good correlation. When analyzing nRMSE, although there is a slight offset, the error is not large, indicating that it has good potential.

[0081] S6: Compare the CSSI chlorophyll universal inversion model with other chlorophyll inversion models to demonstrate the superiority of the CSSI universal inversion model.

[0082] S61: Through literature research, a chlorophyll spectral index model with excellent performance in chlorophyll inversion is found as a comparison model. Generally, it is a chlorophyll spectral index model proposed by other researchers, and its input parameter is also the leaf reflectance spectrum. After literature research, this application has obtained a total of six chlorophyll spectral index comparison models, which are shown in Table 4 respectively. The subscript numbers in the table represent the corresponding reflectance spectrum wavelengths.

[0083] S62: Use the polynomial fitting steps in formula (2) to construct chlorophyll inversion comparison models corresponding to the six chlorophyll spectral index comparison models, reduce differences, and thus achieve better comparison.

[0084] S63: Use each chlorophyll inversion comparison model to obtain its predicted inversion results in all measured data sets, see Figure 5 HM.

[0085] S64: The inversion results of the CSSI chlorophyll universal inversion model and the chlorophyll inversion comparison model are compared and verified, and the inversion performance of the CSSI chlorophyll universal inversion model and different chlorophyll inversion comparison models on the same data set is further studied. The verification results of the CSSI chlorophyll universal inversion model and the chlorophyll inversion comparison model are shown in Figure 4 , Figure 5 The results show that the performance of the CSSI general vegetation index inversion model built using the PROSPECT-D simulation data set is better (R 2This shows that the CSSI universal chlorophyll inversion model can better reflect the correlation with chlorophyll content than other chlorophyll inversion comparison models. In addition, since there are five measured data sets from different teams, these data sets contain a lot of vegetation types, but high-precision inversion can be achieved on these data sets, which shows that the CSSI chlorophyll universal inversion model is indeed very versatile.

[0086] Compared with other chlorophyll indices, this application can better capture the chlorophyll information carried in the vegetation reflectance spectrum. Therefore, the CSSI chlorophyll universal inversion model constructed using the PROSPECT-D simulation data set has higher inversion accuracy and better universality. Although the PROSPECT-D simulation data set can basically represent the physiological status of the measured data set, there are still subtle differences between it and the measured data set. Therefore, when the measured data set is sufficient and reliable, and only the vegetation type corresponding to the measured data set is predicted, the processing can be simplified, and the measured data set can be directly used (the content and steps related to the simulation data set in S1~S3 can be omitted) to construct the CSSI chlorophyll universal inversion model to invert the chlorophyll content of vegetation.

[0087] The CSSI chlorophyll universal inversion model of the present application can be actually applied to remote sensing inversion or drone detection inversion of chlorophyll content in leaves of multiple types of vegetation with high precision and quantitative inversion. The inversion method is the existing technology, that is, the vegetation spectral data of a certain area is collected by remote sensing satellites or drones, and the chlorophyll content of the vegetation leaves in the area is calculated by the CSSI chlorophyll universal inversion model of the present application, so as to obtain the vegetation status of the area in real time, conveniently and relatively accurately, and provide basic vegetation data for social economy and disaster prevention and mitigation.

[0088] Table 1 Main characteristics of five measured data sets

[0089]

[0090] Table 2 Parameter ranges and step sizes used to input the PROSPECT-D model

[0091]

[0092] Table 3 Examples of input parameters applicable to similarity index inversion prediction

[0093]

[0094] Table 3 introduces the input parameter information required for the CSSI chlorophyll universal inversion model to invert and predict leaf chlorophyll and the predicted results of leaf chlorophyll content. Among them, the first to third columns of Table 3 are input parameters, the first column of reflectance spectra 1-5 refers to the reflectance spectra of five leaves measured using a spectral measurement instrument, the second column refers to the spectral measurement range of the corresponding leaves, the third column refers to the resolution of the spectral measurement, and the fourth column refers to the corresponding spectrum inverted using the CSSI chlorophyll universal inversion model of this application to obtain the predicted value of the chlorophyll content of the corresponding leaf.

[0095] Table 4 Spectral index used to predict leaf chlorophyll content

[0096]

[0097] Note: D in Table 4 730 It refers to the derivative of the 730 band of the spectrum, and R is the reflectance spectrum of the corresponding band.

Claims

1. A method for constructing a universal inversion model for leaf chlorophyll content of different types of vegetation, wherein the universal inversion model for leaf chlorophyll content satisfies LCC CSSI = a·SAC+ b = a·SAC( r (λ) w (λ) [OSI ] + b(2) LCC refers to the chlorophyll content of plant leaves. r (λ), w (λ) represents the reflectance spectrum of vegetation leaves and the chlorophyll absorption coefficient of vegetation leaves with consistent spectral resolution, OSI is the optimal spectral interval, coefficients a and b are calculated from the simulated data set, and the spectral band resolution of the reflectance spectrum and chlorophyll absorption coefficient of the leaves is 1nm. Its characteristics are The following steps are involved: S1: Construct a sample data set, in which the spectral band resolution of the reflectance spectrum and chlorophyll absorption coefficient is the same; S11: Generate simulation data sets using the PROSPECT radiative transfer model; S2: Calculate the optimal spectral interval; S21: According to formula (1), the leaf similarity chlorophyll index and the SAC similarity index in the high correlation interval between chlorophyll are calculated: （1） In formula (1), r and w refer to the reflectance spectrum and chlorophyll absorption coefficient of the leaves, respectively. L represents a spectral interval, l It is a band in this spectral range; S22: Correlation analysis was performed on the interval with high correlation between leaf SAC similarity index and chlorophyll content; S23: The spectral interval with the best correlation between SAC similarity index and chlorophyll content was defined as the optimal spectral interval; S3: Construct a universal chlorophyll inversion model for CSSI; S31: Obtain the SAC similarity index value of the simulated data set in the optimal spectral range, analyze the correlation expression between the SAC similarity index and its corresponding leaf chlorophyll content, and determine the construction expression of the CSSI chlorophyll general inversion model; S33: Construction of a CSSI chlorophyll general inversion model using simulated datasets.

2. The method for constructing a universal inversion model for leaf chlorophyll content of different types of vegetation according to claim 1 is characterized by: The PROSPECT radiation transfer model adopts the PROSPECT-D radiation transfer model, and the simulation data set includes chlorophyll content, dry matter content, equivalent water thickness, carotenoid content, brown pigment content, structural parameters, and anthocyanin content of plant leaves.

3. The method for constructing a universal inversion model for leaf chlorophyll content of different types of vegetation according to claim 1 is characterized by: The SAC similarity index values ​​in the simulated data set only retain data corresponding to the SAC similarity index values ​​in the measured data set.

4. The method for constructing a universal inversion model for leaf chlorophyll content of different types of vegetation according to claim 1, characterized in that Step S12 and step S13 are also provided: S12: The vegetation leaf samples collected in the field are measured and analyzed in the laboratory to construct a measured data set, which includes: reflectance spectra of fresh leaves and chlorophyll content of fresh leaves; S13: Obtain the leaf chlorophyll absorption coefficient of the measured data set.

5. The method for constructing a universal inversion model for leaf chlorophyll content of different types of vegetation according to claim 4 is characterized by: There is also a step S32 between step S31 and step S33. S32: Screening the SAC similarity index values ​​of the simulated data set, and only retaining the SAC similarity index values ​​of the simulated data set corresponding to the SAC similarity index values ​​in the measured data set.

6. The method for constructing a universal inversion model for leaf chlorophyll content of different types of vegetation according to claim 4 is characterized by: The step S22 uses the Pearson coefficient to perform correlation analysis, and the SAC similarity index of each band is placed in a matrix, the horizontal axis of the matrix represents the starting band, the vertical axis represents the ending band, and the point in the matrix represents the Pearson coefficient value of the SAC similarity index and chlorophyll content obtained from the starting band to the ending band of the point.

7. A verification and evaluation method for the universal inversion model of leaf chlorophyll content for different types of vegetation as claimed in any one of claims 1 to 6, characterized in that The following steps are involved: S4: Use the CSSI chlorophyll universal inversion model to predict the inverted chlorophyll content; S5: Evaluating the performance of the CSSI chlorophyll general inversion model.

8. The verification and evaluation method of the universal inversion model for leaf chlorophyll content of different types of vegetation according to claim 7, characterized in that After step S5, step S6 is also provided: S6: Compare the CSSI chlorophyll universal inversion model with other chlorophyll inversion models.

9. The verification and evaluation method of the universal inversion model for leaf chlorophyll content of different types of vegetation according to claim 7 or 8, characterized in that: The performance of the inversion model was evaluated using the coefficient of determination and normalized root mean square error.

Images