| The snow disaster in Western Sichuan Plateau pastoral area has the characteristics of high frequency,wide range of influence,great harm and disaster relief difficulties.Therefore,continuous monitoring of snow cover in Western Sichuan Plateau is very important.Snow coverage is one of the important input parameters in global energy balance,climate,hydrology,ecological model and snow quantitative remote sensing.At present,MODIS standard snow cover products are the main snow cover monitoring products in Western Sichuan Plateau.The product data is not universal and accurate in Western Sichuan Plateau,and other ways to obtain snow cover are limited.In view of the difficulty in obtaining high-precision snow coverage in Western Sichuan Plateau and the low accuracy and universality of MODIS standard products,this paper proposes an extended linear decomposition model(nelmm)based on noise level estimation based on mixed pixel decomposition theory.Firstly,endmember scale factor is introduced to simulate different snow endmembers,it solves the endmember heterogeneity and endmember changes caused by illumination difference,and eliminates the endmember changes that cannot be handled by conventional mixed pixel decomposition algorithm.Then,the noise of adjacent bands is estimated according to the multiple regression theory in hyperspectral application,and the noise weight matrix is obtained from the estimated noise.The noise weight matrix is introduced into the extended linear mixture model to balance the noise level between different bands,so as to reduce the influence of noise level of different bands and eliminate the error caused by the difference of band noise.In order to demonstrate the superiority and robustness of nelmm,three groups of simulation experiments and three groups of real experiments are used to verify the applicability and robustness of the model.At the same time,the proposed nelmm model is compared with FCLs,plmm,elmm and GLMM,and the accuracy is evaluated by two indexes,namely,uniform root error and spectral angle.At the same time,in order to demonstrate the universality and stability of nelmm model in snow cover inversion of Western Sichuan Plateau,this paper selects two typical terrain areas of Northwest Sichuan Plateau and Western Sichuan mountain as the study area,and carries out the experimental research with MODIS09 GA data of three different periods in each study area.In order to verify the accuracy,advantages and stability of nelmm algorithm,extended linear hybrid model elmm is used for cross validation,and compared with MODIS10A1 standard snow cover products.In this paper,four auxiliary indexes(ACC,TSCR,RMSE,R and AE)are selected to evaluate the inversion accuracy,stability and adaptability of the inversion model,the accuracy of the inversion results is verified by using the snow cover produced by landsast8-oli classification image as the reference value.Quantitative analysis and experimental results show that the proposed nelmm model has high robustness and high accuracy,and demonstrates the feasibility of nelmm model in mixed pixel decomposition.The accuracy index of nelmm model in simulation experiment and real experiment is the best,and its decomposition result is also the best,which can balance the problem of different noise levels between different bands,And it can weaken the influence of endmember changes on the decomposition results,so as to improve the image decomposition accuracy of the model;The overall accuracy of nelmm model is higher than that of elmm model,which indicates that the introduction of noise weight matrix balances the noise between bands,reduces the noise level of images in different bands,and improves the accuracy of snow coverage inversion.At the same time,compared with the existing MODIS standard snow products,nlemm inversion has higher accuracy of snow coverage,and the overall accuracy of nlemm extraction is higher than 90%,The proportion of total snow area is more than 90%,and the mean square error is less than 0.18,which has good robustness,and can play a certain role in snow cover information extraction and snow monitoring in Western Sichuan Plateau. |
PDF Full Text Request