Study On Vegetation Index Of Remote Sensing And Its Aplications

Vegetation index（VI）plays an important role in reflecting the vegetations, which containsmore than90%information of the remote sensing images. While enhancing the vegetationinformation, it can weaken the non-vegetation information at the same time. By now, more than100kinds of VIs are proposed. On the one hand, its develepment promotes the research ofremote sensing vegetation, but on the other hand, its diversity causes some blindness whenselecting the suitable VI in application. This paper took Yongan City of Fujian province as thestudy area and studied the characteristics of VI and its application in forest extraction,classification and leaf area index（LAI）estimation, based on the images of BJ₁,IRS_P6andMODIS. The main contents are described as follows:（1） Analysis of the characteristics of VI based on classification: according to the groundsamples, this paper compared and analysed the VIs of BJ₁and IRS_P6by variance andmultiple comparisons. The VIs features for classification were studied with BJ₁data aftergeometric correction, atmospheric correction and topographic correction. And this paperanalysed the differences of VIs in classification with different images by using BJ₁and IRS_P6data. The results indicated that RGNDI can eliminate shadows effectively and EVI2candistinguish forest and farm well, which showed the accuracy of forest extraction can meet therequirements of macro-survey. In addition, RVI、EVI2、NDVI were better in classifying bambooand broad-leaved forest, while EVI2、GRNDVI、RVI、GEMI were better in extracting bamboo.However, the classification result of tree species is not satisfactory.（2） Estimation of forest LAI based on VIs: investigate the relationship between VIs and LAIin the masson pine, fir, bamboo and broad-leaved forest based on IRS_P6（LISS-III） images.Specific results were listed as follows: First, both exponential curve model and power curvemodel of TNDVI, NDVI and RVI estimated the LAI of masson pine well. Second, for estimatingthe LAI of fir, exponential curve model and power curve model of MSAVI, RVI and EVI2werebetter. Third, exponential curve model and power curve model of MSAVI, EVI2and RDVI werebetter in estimating the LAI of bamboo.Last, the quadratic curve model of NDVI, TNDVI andRVI showed well in estimating the LAI of broad-leaved forest.（3） VI normalization and its application in LAI inversion: taking NDVI as an example, thispaper analysed VIs normalization in different sensors of BJ₁CCD, IRS_P6LISS3and TerraMODIS. On the basis of the total radiance ratio, the data of BJ₁’CCD sensor and IRS_P6’LISS3sensor were normarlized as MODIS sensor data, and then the NDVI before andafter normalization were calculated. The result indicated that the NDVI relationship of differentsensors after normalization was much closer to the ideal relationship of y=x. The LAI ofmasson pine was estimated by using NDVI, TNDVI and RVI of normalized IRS_P6data, ofwhich the accuracy was0.9%lower than that before normalization, almost the same. Moreover,the VIs of different sensors were nearly equal after nomalization. So the VIs of different sensorsafter nomalization can be applied to estimate LAI generally, as well as to multiple sensors.