Retrieval Of Leaf Area Index Based On HJ-1A/B Satellite Data

Leaf Area Index(LAI)is a very important parameter of canopy structure of terrestrial surface ecosystem.In recent years,remote sensing satellite images provide a reliable data source for rapid extraction of leaf area index of regional and global.In terrain complex areas,the influence of terrain effect,the leaf area index of remote sensing inversion has the problems of low inversion precision and high time consuming.Based on the multi-spectral image of the environment-1 satellite,the influence of the terrain on the vegetation index is evaluated.Based on the LAI data measured by the ground instrument,the statistical model and the BP neural network model are used to reconstruct the leaf area index of the study area by eliminating the terrain-influenced vegetation index.Spatial distribution of LAI in study area by geographic environment data.The research contents are as follows:(1)Analysis of the impact of terrain on vegetation index.The common vegetation indices such as TAVI,NDVI,RVI,DVI and SAVI were selected,and the correlation between vegetation index and cosi was calculated by combining DEM data.And the terrain correction of HJ-1-B CCD image is carried out.The effect of image terrain correction is compared and analyzed.It is concluded that the non-proportional vegetation index DVI is very strongly affected by the terrain,and the terrain is more serious as the slope increases.The ratio of vegetation index NDVI and RVI can weaken the influence of terrain to a certain extent.However,when the slope is more than 18°,the terrain will have a great influence on NDVI and RVI.The same is the ratio of vegetation index SAVI,although it can eliminate the impact of soil background,but it is more sensitive to the terrain,the impact of terrain is more serious.TAVI has a strong effect on the impact of terrain,terrain on its basic has no effect.C correction has a significant terrain correction effect on the environment satellite image of the study area,which can greatly weaken the influence of the terrain on the vegetation index,and the effect of the area is larger.(2)Prediction of vegetation index before and after terrain correction.The results show that the TAVI,NDVI and RVI vegetation indices have a good fitting effect on the LAI of the study area,and the SAVI vegetation index is not suitable for the LAI in the single vegetation index statistical model.Among the four adopted models,the quadratic curve model has poor fitting effect in each vegetation index,and the power function curve model and the exponential curve model fit well.The TAVI,NDVI and RVI vegetation indices in the multiple linear regression model have a good fitting effect on the LAI of the study area.In complex terrain areas,TAVI can eliminate the terrain effect without the need of terrain correction,and the accuracy of LAI have a greater improvement.(3)BP neural network model retrieval LAI.Compared with the single vegetation index statistical model and the multiple linear regression statistical model,the BP neural network model has higher overall accuracy,and the BP neural network model with TAVI,NDVI and RVI vegetation index as the input data is the highest in different vegetation index combinations.The RMSR is 0.326 and the RMSE is 0.324.The results of the inversion are in good agreement with the measured LAI values.Finally,the LAI of the LAI is retrieved by using the BP neural network model with the highest precision.