Study On Extraction Of Forest Vegetation Information Base On Multi-source Remote Sensing Image

Forests are important terrestrial ecosystems on earth, and the foundation of biodiversity. World countries attach great importance to the monitoring of forest resources, in a space technology to support the monitoring of forest resources, forming a three-level remote monitoring system of large-scale, medium-scale and small-scale. How to multi-source data on different measurement platform to integrate, forming multi-scale, different resolution remote sensing data system,in order to analysis and compare different remote sensing data, Has become the forefront of research issues currently.In this paper, taking ZhaoSu in Xinjiang Province as study area, selecting MODIS, RapidEye and TM data with the same time and the geographical coordinates of its registration. Using different scales vector mesh machinery swatches, Selected to meet the requirements of the sample. extraction of the three classes of each NDVI sample. Constructed linear normalization transformation models by regression analysis to MODIS NDVI data reference. Extraction forest vegetation information of RapidEye and TM images base on threshold. This paper studies the content and conclusions are as follows:(1)The correlation analysis of NDVI shows samples of different scales,750m optimal segmentation scale, this scale can reduce the error factor in image registration and sensors caused effectively.(2)By regression analysis of MODIS and RapidEye, MODIS and TM samples, linear regression model has the highest correlation and R2 were 0.71 and 0.78, linear regression models which can better reflect the correlation between NDVI samples.(3)Extraction forest vegetation information of transformed RapidEye and TM images base on threshold. The relative accuracy of information classified forest vegetation was 80% and 82.5%, absolute accuracy 81.6% and 83.1%. Description NDVI normalized conversion method based on different scales is feasible.(4) According to high score of remote sensing data, to control the accuracy of forest resources based on macro-monitoring threshold. the MODIS classification of forest vegetation in the study area absolute accuracy is improved by 11%. And in the Tekesi County and Yili region is generalized experiment, the classification accuracy is good.