| Leaf area index(LAI)is a particularly important vegetation parameter that can reflect the process of forest growth and living environment,and it can also directly quantify the growth status and canopy structure of forest land.The forest ecological process is closely related to LAI,and the biological and physical processes in forest ecosystem are controlled by LAI,in addition,the change of LAI can reflect the different situation of forest growth and development,it can also give common understanding analysis and estimation on various groups of forest ecosystem.Therefore,the study of leaf area index(LAI)has become a hot topic for domestic and foreign scholars.Although the LAI can be measured accurately by traditional method,but such measurement is time-consuming and laborious,and cannot be carried out in large-scale research.Now the remote sensing inversion is the only method to estimate LAI in large-scale time-spacdTexture and spectral features are important sources of image analysis.At present,remote sensing is mainly using texture or spectral features.If the texture and spectral characteristics can be combined,the effect is better than using a single information.Broad-leaved forest is widely distributed in the southwestern mountains of Sichuan Province,and has great significance to the local soil and water conservation and the stability of the ecological environment.However,the remote sensing inversion of LAI is rarely reported.Therefore,it is necessary to carry out the research on leaf area index inversion of broad-leaved forest in southwest Sichuan based on image spectral and texture features.This research wascarried out in the Shangli town and it is based on the Pleiades image data,the GPS image is used to divide the sample,we measured the effective leaf area index(LAIe)and locating the sample plots by using GPS.Furthermore,using spss20.0,the spectral estimation of the spectral information,vegetation index,texture parameter and LAIe of each plot was established.The optimal estimation model of LAIe was established.And the texture parameters extracted from different texture windows are analyzed,came to a conclusion:(1)There was a significant correlation between NIR and LAIe in the spectral information of the four bands.In addition,there was a significant correlation between the five vegetation indices and LAIe.Furthermore,the correlation between NDVI and LAIe was the best,and the effect of the multiple regression model was better than that of the univariate regression model when the spectral information and vegetation index were compared with LAIe In the regression analysis,and the regression model is:LAIe=0.5109NDVI+0.3406SAVI+0.0680RVI+0.00011DVI+0.00018PVI+0.00005NIR+1.8468(R2=0.8486,RE=0.0972,RMSE=0.1226)(2)The eight texture parameters extracted from four different windows hadsignificant correlation with LAIe.When the texture parameter values of four windows were analyzed by unity regression,the best window of different texture parameter fitting model was different:the best window of ASM and VAR is 3 x 3;the best window of COR is 5 x 5;ENT,the best window of MEA is 7 x 7;the best window of CON,DIS,HOM is 9 x 9.(3)When the texture parameters of four windows and LAIe were analyzed by multiple regression analysis,the effect was better than that of univariate regression analysis.And the larger the window was,the better the regression model was.the regression effect was the best whenthe window is 9 x 9,the model was expressed as:LAIe=3.746-0.230CON+0.295ENT-2.062HOM(R2=0.7102,RMSE=0.1967,RE=0.1184)(4)In the case of LAIe,the effect of the regression model was better than that of the texture parameter,and only the texture parameters were used to invert LAIe to show the same rule.The larger the window was,the better the regression model was,9 window to join the vegetation index regression model for the optimal model,the model expressed as:LAIe=0.425+2.91NDVI+0.206CON+0.187ENT-0.01OMEA(R2=0.8616,RMSE=0.1352,RE=0.1250)... |
PDF Full Text Request