CN108761456A - A kind of inversion method of leaf area index of crop - Google Patents

Abstract

The present invention discloses a kind of inversion method of leaf area index of crop, including：S1 obtains research area's polarization radar image；S2 obtains polarization data of compacting from the polarization radar image simulation in S1.Using polarization radar figure, simulation obtains polarimetric radar data of compacting；S3 carries out polarization decomposing to the polarization data of compacting in S2, obtains the polarization parameter image that compacts with different physical significances；S6 carries out non-linear heredity-partial least-square regression method modeling, obtains the predicted value of leaves of winter wheat area index.The present invention provides the extraction schemes of the compact polarization SAR parameter highly relevant with Crop leaf area index parameters.GA-PLS methods are introduced into the information analysis for polarimetric radar remote sensing of compacting by the present invention for the first time, avoid the Physical Process Analyses of complexity early period, reduce human error.The present invention provides the technical solutions of the leaf area index inverting of the crop key developmental stages based on novel polarization SAR data of compacting.

Description

A Retrieval Method of Leaf Area Index of Crop

technical field

The present invention relates to remote sensing technology, more specifically, relates to crop leaf area index inversion technology.

Background technique

Leaf Area Index (LAI) is a basic parameter closely related to vegetation canopy structure. In agricultural research, it is directly related to crop growth, growth stage and yield. It is a good indicator of crop development and health and is also used as an input variable in many crop growth and yield prediction models. It is an important parameter for evaluating the physiological and ecophysiological processes of agroecosystems, and plays an important role in monitoring crop growth and predicting yield at regional and national scales. Remote sensing technology has the characteristics of periodic observation and large area coverage in obtaining ground information. It plays an important role in agricultural resource monitoring. It can capture the distribution information of key biophysical parameters of terrestrial vegetation in space and time. Therefore, it can provide a practical method to observe LAI at the macro scale.

At present, optical remote sensing is the main means of monitoring crop growth parameters. The application of optical remote sensing data to monitor crop growth has formed a relatively mature technical method, and its accuracy has reached a high level. However, in the critical growth period of autumn harvest crops in the drylands of northern my country, cloudy and rainy weather has a great influence, and it is impossible to obtain complete and continuous optical remote sensing observation data in a timely and effective manner. Therefore, it is necessary to use radar remote sensing to monitor dryland crops. The emergence of Synthetic Aperture Radar (SAR) has made crop monitoring unaffected by clouds, fog, and rain, ensuring the independence of data acquisition from local weather, and when microwave remote sensing detects vegetation information, it can be obtained with optical sensors. Completely different information. Based on this, many scholars have conducted a large number of experimental studies to explore the sensitivity of SAR to crop LAI. Polarization is an electric field characteristic of electromagnetic waves. Radar waves scattered by different scatterers contain different polarization information and describe different physical scattering processes. The development of polarimetric radar has experienced a process from single polarization to dual polarization and then to full polarization, and full polarization data provides the most abundant polarization information. Therefore, full-polarization radar data are often used to retrieve various parameters of vegetation and land surface. At present, there are two main types of algorithms for estimating crop LAI from SAR images: methods based on empirical models and methods based on semi-empirical models. The method based on the empirical model mainly uses the method of regression analysis to estimate the LAI parameters of crops.

While full-polarization SAR performs well in LAI inversion, its known limitation is the reduced swath width resulting from using repeated pulse repetition frequencies to scan all combinations of transmit/receive polarization. To overcome such limitations, compact polarization SAR, which only transmits circular polarization and receives two orthogonal linear polarization imaging modes, has been proposed. Compact Synthetic Aperture Radar (CP SAR), also known as compact or compact polarization SAR, is a new type of imaging radar system, which emits one polarization wave and receives two orthogonal polarization waves, effectively Reducing the complexity and energy consumption of the SAR system and reducing the size of the sensor has become one of the important development trends of the new generation of earth observation SAR system. In imaging radar polarization hierarchy, compact polarization is between full polarization and dual polarization. Compared with fully polarimetric SAR, compact polarimetric SAR can not only maintain polarization information to a certain extent, but also achieve a larger swath width and incident angle range to meet some special application requirements. In addition, compact polarization SAR also has the advantages of self-calibration and cross-validation. In April 2012, the first earth observation radar satellite RISAT-1 (Radar Imaging Satellite 1) with compact polarization measurement capability was successfully launched. The Japanese ALOS-2 (Advanced Land Observation Satellite 2) satellite launched in 2014 also uses compact polarization as an experimental data model. In the next few years, Canada's RCM (Radar Constellation Mission), Argentina's SAOCOM (Satelite Argentino de Observacion Con Microondas), and the United States' DESDynI (Deformation, Ecosystem Structure and Dynamics of Ice) will also launch SAR satellites with compact polarization observation modes. With the increasing abundance of compact polarimetric SAR systems for earth observation, it is particularly urgent to carry out research on key technologies for applications based on compact polarimetric SAR data. At the same time, using the new compact polarimetric SAR data to analyze the response characteristics of typical ground objects and develop a highly robust information extraction algorithm is of great significance for promoting the development of my country's future SAR sensors and the application of radar remote sensing technology in related fields. important meaning. The technical premise of inverting crop parameters based on compact polarimetric radar data is to find the polarization parameters most relevant to the biophysical and chemical parameters of crops, so that an accurate inversion model can be established. Therefore, in the face of a large amount of polarization information brought by the compact polarization data, it is very important to screen the compact polarization information.

Most of the existing radar-based crop parameter inversion techniques use full-polarization data, and most of them use a single polarization parameter inversion method. The specific steps are as follows:

① According to the existing cognition of the research area, artificially judge which scattering mechanism the research target is.

② Field measurement, taking several samples in the research area, and using field testing or laboratory sample measurement methods to obtain the characteristic parameter data of the research target.

③Based on the premise that the physical scattering mechanism of some polarization parameters is known, find a polarization parameter that is most consistent with the scattering mechanism of the research target, and use polarization decomposition and other techniques to obtain the polarization parameter data.

④ Establish a simple linear correlation between the polarization parameter data and the characteristic parameter data of the research target through data training.

⑤ According to the correlation, input the polarization parameter image, and invert the characteristic parameter image of the study area.

There are following defects and deficiencies in the prior art:

① Existing technologies are mainly aimed at full-polarization radar data. The coverage of full-polarization data is limited, and there will be many restrictions in the promotion and application of large-scale applications.

② In the application of radar polarization, the existing technology only uses the full polarization parameters of the known physical scattering mechanism, and does not make full use of the new polarization information brought by the compact polarization data. In terms of the selection of multi-dimensional polarization characteristic parameters sensitive to crop parameters, the current main technical method is to use artificial judgment of the scattering mechanism of the research target, and its accuracy depends on the researchers' understanding of the research target and the interference of human errors. larger.

③Many existing technologies only use one polarization parameter to simulate the physical scattering process approaching the research target. However, the scattering mechanism of natural ground objects is very complex, and in most cases it is difficult to express it with a significant physical process, but a comprehensive performance of multiple physical processes.

Contents of the invention

Although the dimensionality of the observation space of compact polarimetric SAR is reduced compared to that of full polarimetric SAR, it has been shown to have great potential in rice mapping with performance similar to that of full polarimetric SAR. There are few studies on the inversion of winter wheat leaf area index based on compact polarimetric radar data. Few people use nonlinear genetic partial least squares (GA-PLS) algorithm to select SAR characteristic parameters and reduce dimension. The purpose of the invention is to obtain the leaf area index of crops by inversion of compact polarization radar data, and expand the application potential of compact polarization radar data in inversion of crop parameters.

For this reason, the present invention proposes a kind of leaf area index inversion method of crops, comprising:

S1, to obtain the full polarization radar image of the research area;

S2, Compact polarization data simulated from the fully polarimetric radar images in S1. Using fully polarized radar graphics, simulate and obtain compact polarized radar data;

S3, performing polarization decomposition on the compact polarization data in S2 to obtain compact polarization parameter images with different physical meanings;

S5, measure and obtain the parameter value of the leaf area index of winter wheat in the selected research area;

S6, performing nonlinear genetic-partial least squares regression method modeling to obtain the predicted value of winter wheat leaf area index.

The beneficial effects of the present invention include:

1) The present invention provides an extraction scheme for compact polarimetric SAR parameters highly correlated with crop leaf area index parameters.

2) The present invention introduces the GA-PLS method into the information analysis of compact polarization radar remote sensing for the first time, avoids the complex physical process analysis in the early stage, reduces human errors, does not make any quantitative assumptions, and screens polarization information from the data level Channel: In terms of dimensionality reduction of multi-dimensional polarization characteristic parameters, the present invention proposes a highly efficient multi-parameter dimensionality reduction algorithm for compact polarization radar data.

3) The present invention provides a complete set of technical solutions for leaf area index inversion of key growth stages of crops based on novel compact polarimetric SAR data.

Description of drawings

Fig. 1 is a technical roadmap of an embodiment of the method of the present invention.

Figure 2 is a flowchart of one embodiment of the method of the present invention.

Fig. 3 is the composite graph obtained by Renny's compact polarization decomposition into three components.

Fig. 4 is a schematic diagram of the importance selection of compact polarization parameters.

Figure 5 is a scatter plot of measured and inverted LAI values.

Detailed ways

Embodiments of the present invention are described below with reference to the drawings, in which like parts are denoted by like reference numerals. In the case of no conflict, the following embodiments and the technical features in the embodiments can be combined with each other.

The following takes winter wheat as an example to describe the technical scheme of the method of the present invention, so that those skilled in the art can expand to other types of crops without creative labor.

In the present invention, by studying the tight polarization SAR response law and mechanism of the key phenological phases of winter wheat, the application potential of the tight polarization SAR data in the winter wheat LAI inversion is deeply explored, and the winter wheat LAI inversion based on the tight polarization SAR data is established. To improve the application level of compact polarization SAR in dryland crop monitoring, and provide basic information for winter wheat growth monitoring and yield estimation, field management, resource allocation and decision-making.

Referring to Fig. 1, the technical principle of the present invention is, firstly, CP SAR data is simulated by using RADARSAT-2 full-polarization SAR data. On this basis, the CP SAR parameters with certain physical meaning are obtained based on the extraction of compact polarization scattering matrix and compact polarization decomposition method. Then, the most sensitive CP SAR parameters were selected by feature selection algorithm, and the inversion model of winter wheat LAI was constructed based on mathematical regression modeling algorithm. Finally, the inversion accuracy is comprehensively and systematically verified using the LAI data measured during the time period in which the field and the satellite transit time are synchronized.

More specifically, referring to Fig. 2, the method of the present invention includes:

S1, to obtain the full polarization radar image of the study area.

Fully polarimetric radar images contain the most comprehensive polarization information and can more fully describe the scattering characteristics of scatterers.

In one embodiment, RADARSAT-2 data may be used. Preprocessing the raw data of RADARSAT-2 full-polarization radar through radiation calibration, filtering, geometric calibration, etc., to obtain the full-polarization radar image, the radiation calibration, filtering, geometric calibration processing can adopt the general method in this field .

S2, Compact polarization data simulated from the fully polarimetric radar images in S1. Using the full polarization radar graph, the simulation obtains the compact polarization radar data used in the subsequent processing.

More specifically, using RADARSAT-2 full polarization data (S2 matrix), a covariance matrix C3 is generated. Then, based on the conversion relationship between linear polarization and right-hand circular polarization, the relationship between the Stokes vector of compact polarization SAR data and the covariance matrix C3 matrix of full polarization SAR data is established to realize the simulation of compact polarization SAR data . Simulation data are stored as Stokes vectors.

S3, performing polarization decomposition on the compact polarization data in S2 to obtain compact polarization parameters with different physical meanings, and then obtain a compact polarization parameter image. Polarization decomposition technology is the main means of polarization analysis. Different polarization parameters can be obtained through polarization decomposition technology, and the scattering characteristics of scatterers can be described from different physical aspects. By extracting and decomposing the compact polarization data (CP SAR matrix), 19 related compact polarization parameters (CP parameters) are obtained, namely Raney_Rnd, Raney_Dbl, RV, RR, RL, RH, Raney_Odd, p2 ,p1,l2,l1,Contrast,LPR,H,DoLP,DoCP,CPR,A and Raney_m, these compact polarization parameters can be combined as input variables for GA-PLS modeling.

Figure 3 is a composite image of the three components extracted from the compact polarization data simulated from the full polarization data. Figure 3 shows the composite map obtained by Renny compact polarization decomposition into three components (including: Raney_dihedral scattering component; Raney_volume scattering component; Raney_surface scattering component).

S4, preprocessing the compact polarization parameter image obtained in S3. The preprocessing of the compact polarization parameter image is divided into two parts:

S41, perform geometric correction on all compact polarization parameter images.

S42. Based on the compact polarization parameter image obtained in S41, read the compact polarization parameter value of each sampling point on the image. In one embodiment, due to the presence of speckle noise in the radar image and the interference effect of electromagnetic waves, a small area is defined near the sampling point, and the average value of the compact polarization parameters in this area is taken as the compact polarization parameter value of the sampling point .

S5, measure and obtain the parameter value of leaf area index of winter wheat in the selected research area. Measuring instruments can be used to obtain parameter values of leaf area index of winter wheat in the selected study area. For example, the LAI-2200 vegetation canopy analyzer produced by LI.COR Company is used. LAI-2200 does not need to touch the crops, and directly observes the diffuse scattering changes above and below the wheat canopy to obtain the effective LAI of winter wheat indirectly. The effective LAI is equivalent to the actual LAI in practical applications, and they all have the same light interception ability. In one embodiment, several test areas can be evenly selected in the research area, and the value of the leaf area index of the samples can be measured.

S6, performing GA-PLS modeling. The modeling process is divided into two parts:

S61, use the genetic algorithm (GA) to screen out several compact polarization parameters most relevant to the characteristic parameters of the study area.

The overall parameters before screening are Raney_Rnd, Raney_Dbl, RV, RR, RL, RH, Raney_Odd, p2, p1, l2, l1, Contrast, LPR, H, DoLP, DoCP, CPR, A and Raney_m. The most relevant judgment standard is to judge the importance of compact polarization parameters through genetic algorithm, that is, the degree of correlation with the parameters to be inverted.

The genetic algorithm is to encode all the independent variables and dependent variables into binary codes, and based on the principles of genetics, select the independent variable most related to the dependent variable. Through the screening of compact polarization parameters, the final inversion accuracy can be effectively improved.

S62, based on the polarization parameter data selected in S61, use the partial least squares (PLS) regression method to establish an inversion model, and obtain the predicted value of the winter wheat leaf area index. GA-PLS is a relatively mature mathematical algorithm, which is widely used in various multiple regression analysis, but it has not been applied in the field of inversion of compact polarimetric radar remote sensing parameters. The present invention introduces it into crop parameter inversion technology based on compact polarization parameters, which is not available in the prior art.

The principle and steps of obtaining the predicted value of winter wheat leaf area index are as follows:

The principle of the partial least squares regression method is as follows:

Suppose the sample size is N, the data matrix of the independent variable is X, and the data matrix of the dependent variable is y.

First, principal component analysis is performed on X and y:

X=TP'+E

y=UQ'+F

Among them, T and U are score matrices (i.e., principal component matrices), which represent the main information of X and y, P and Q are loading matrices, and E and F are residual matrixes.

Second, establish the regression equation of principal components T and U

U=BT

Finally, calculate the final inversion model

y=Xb+c

where b is the regression coefficient vector and c is the residual vector.

Based on the above-mentioned GA-PLS principle, write the code, organize the polarization parameter data and the measured characteristic parameter data of all samples into a data table, read it into the compiled program and run it, and obtain the final linear regression equation, from the equation The filtered polarization parameters and corresponding weights are obtained. The linear regression equation is as follows:

y=a×X ₁ +b×X ₂ +c×X ₃ +…

Among them, Xi ( _i =1, 2, 3...) is the selected polarization parameter, the coefficients a, b, and c are the corresponding weights, which represent the magnitude of the correlation, and y is the characteristic parameter.

S7. Accuracy evaluation is performed based on the measured value of winter wheat leaf area index obtained in S5 and the predicted value obtained in S6, and the accuracy evaluation result of the inversion result of winter wheat leaf area index is obtained.

Calculate the root mean square error (RMSE) of all samples to evaluate the prediction accuracy of the model. The accuracy of the algorithm results is tested and judged by the accuracy evaluation.

Among them, y _i ' and y _i are the predicted value and the measured value of sample i respectively.

S8, generating an inversion diagram according to the inversion model established in S6. Based on the calculated inversion formula, input the compact polarization parameter image screened in S4, and use remote sensing software or write code to generate the final remote sensing inversion map of winter wheat leaf area index in the study area.

Overall, through CP SAR simulation and CP decomposition method, CP parameters with different physical meanings can be generated from the original SAR image. These CP parameters were all correlated with winter wheat leaf area index. The introduction of feature selection algorithm can select the parameters that can best characterize winter wheat leaf area index. On this basis, the precise mathematical relationship between the selected CP parameters and the leaf area index variables can be established by using the algorithm of regression modeling. That is, the scattering matrix S and the complex covariance matrix C of the compact polarization data are simulated first by using the full polarization data. Then, many compact polarization parameters can be extracted from these matrices. The CP polarization decomposition parameters are calculated by Cloude decomposition and Renny decomposition. These compact polarization parameters are modeled as input variables. According to the position of SAR image and leaf area index sampling sample, some sample points are selected for model correction, and the rest samples are used for verification. The CP parameters calculated based on the CP synthetic aperture radar theory and the CP decomposition method are used as the independent variable matrix x, and the leaf area index values measured at each sampling point form the dependent variable vector. Through mathematical modeling, the final inversion model can be expressed as the following linear regression equation:

y＝a ₁ x ₁ +a ₂ x ₂ +a ₃ x ₃ +... (1)

where x _i is the polarization parameter screened out by feature selection and a _i is the corresponding regression coefficient (i=1,2,3...). A schematic diagram of the process of parameter importance selection is shown in Fig. 4 .

When genetic algorithm is applied to feature selection, its function is to select independent variables that are closely related to independent variables. The independent variables closely related to the relevant variables are defined, and a specific mathematical multiple regression model is established by partial least squares method. The scatter plot of the modeling results is shown in Figure 4. Dots represent calibration samples, while triangles represent validation samples (Figure 4). The solid blue line is the 1:1 standard line. In this study, the coefficient of determination (R2, unitless) and root mean square error (RMSE, m²/m²), were used to evaluate the relationship between measured and estimated LAI, as well as the estimation accuracy of the method. The value of R2 in this study reaches 0.70, and the root mean square error deviation is 0.40 m2/m2, which shows that the inversion results of winter wheat LAI with higher accuracy can be obtained based on CP SAR parameters. From the distribution of scatter points shown in Figure 4, there is an approximate linear correlation between the selected CP parameters and LAI of winter wheat.

The above-described embodiments are only preferred specific implementations of the present invention, and ordinary changes and replacements performed by those skilled in the art within the scope of the technical solution of the present invention should be included in the protection scope of the present invention.

Claims (10)

1. A leaf area index inversion method of crops, characterized in that, comprising: S1, to obtain the full polarization radar image of the research area; S2, Compact polarization data simulated from the fully polarimetric radar images in S1. Using fully polarized radar graphics, simulate and obtain compact polarized radar data; S3, performing polarization decomposition on the compact polarization data in S2 to obtain compact polarization parameter images with different physical meanings; S6, performing nonlinear genetic-partial least squares regression method modeling to obtain the predicted value of winter wheat leaf area index. 2. the leaf area index inversion method of crop according to claim 1, is characterized in that, in step S3, described compact polarization parameter comprises: Raney_Rnd, Raney_Dbl, RV, RR, RL, RH, Raney_Odd, p2, p1, l2, l1, Contrast, LPR, H, DoLP, DoCP, CPR, A and Raney_m. 3. the leaf area index inversion method of crop according to claim 1, is characterized in that, also comprises between step S3 and S5: S4, preprocessing the compact polarization parameter image obtained in S3, the preprocessing of the compact polarization parameter image includes: S41, perform geometric correction on all compact polarization parameter images; and S42. Based on the compact polarization parameter image obtained in S41, read the compact polarization parameter value of each sampling point on the image. 4. the leaf area index inversion method of crop according to claim 3, is characterized in that, In S42, a small area is defined near the sampling point, and the average value of the compact polarization parameters in this area is taken as the compact polarization parameter value of the sampling point. 5. the leaf area index inversion method of crop according to claim 1, is characterized in that, step S6 comprises: S61, using a nonlinear genetic algorithm to screen out several compact polarization parameters most relevant to the characteristic parameters of the study area; S62, based on the polarization parameter data selected in S61, use the partial least squares regression method to establish an inversion model, and obtain the predicted value of the winter wheat leaf area index. 6. the leaf area index inversion method of crop according to claim 5, is characterized in that, In S61, the overall parameters before screening are Raney_Rnd, Raney_Dbl, RV, RR, RL, RH, Raney_Odd, p2, p1, l2, l1, Contrast, LPR, H, DoLP, DoCP, CPR, A and Raney_m, The most relevant judgment standard is to judge the importance of compact polarization parameters through genetic algorithm, that is, the degree of correlation with the parameters to be inverted. 7. the leaf area index inversion method of crop according to claim 6 is characterized in that, in step S62, the step of obtaining the predicted value of winter wheat leaf area index comprises: 1) Based on the principle of nonlinear genetic-partial least squares regression method, the polarization parameter data and measured characteristic parameter data of all samples are organized into a data table, and the final linear regression equation is obtained; 2) Obtain the filtered polarization parameters and corresponding weights from the equation; Eggplant River, the linear regression equation is as follows: y=a×X ₁ +b×X ₂ +c×X ₃ +… Among them, Xi ( _i =1, 2, 3...) is the selected polarization parameter, the coefficients a, b, and c are the corresponding weights, which represent the magnitude of the correlation, and y is the characteristic parameter. 8. The leaf area index inversion method of the crop according to claim 1, is characterized in that, comprising: S5, measure and obtain the parameter value of the leaf area index of winter wheat in the selected research area; In S7, accuracy evaluation is performed based on the measured value of winter wheat leaf area index obtained in S5 and the predicted value obtained in S6, and the accuracy evaluation result of the inversion result of winter wheat leaf area index is obtained. 9. the leaf area index inversion method of crop according to claim 8, is characterized in that, The prediction accuracy of the model is evaluated by calculating the root mean square error of all samples. 10. the leaf area index inversion method of crop according to claim 8, is characterized in that, comprises after S7: S8, according to the inversion model established in S6, generate an inversion map, based on the calculated inversion formula, input the compact polarization parameter image selected in S4, and generate the final remote sensing inversion of winter wheat leaf area index in the research area picture.