MODIS EVI AS AN ANCILLARY DATA FOR AN OBJECT-BASED IMAGE ANALYSIS WITH MULTI-SPECTRAL MODIS DATA

摘要

This paper investigates the contribution of Enhanced Vegetation Index (EVI) data to the improvement of object-based image analysis using multi-spectral Moderate Resolution Imaging Spectral-radiometer (MODIS) imagery. Object-based image analysis classifies objects instead of single pixels. The idea to classify objects stems from the fact that most often the important information to process an image is not presented in single pixels but in groups of pixels (objects) (Blaschke et al. 2001). Based on image segmentation, object-based image analysis uses not only spectral related information, but spatial, textural and contextual information as well. However, which type of information to use depends on the image data and the application, among many other factors. EVI data are from the MODIS sensor aboard Terra spacecraft. EVI improves upon the quality of Normalized Difference Vegetation Index (NDVI) product. It corrects for some distortions in the reflected light caused by the particles in the air as well as the ground cover below the vegetation. The EVI data product also does not become saturated as easily as NDVI when viewing rainforests and other area of the Earth with large amounts of chlorophyll. In this research, 69 EVI data (scenes) collected during the period of three years (from January of 2001 to December of 2003) in a mountainous vegetated area were used to study the correlation between EVI and the typical green vegetation growth stages. These data sets can also be used to study the phenology of the land cover types. Different land cover types show distinct fluctuations over time in EVI values and this information might be used to improve land cover classification of this area. Object-based image analysis was used to perform the land cover classification: one was only with MODIS multispectral data (seven bands), and the other one included also the 69 EVI images. Eight land cover types were distinguished and they are temperate forest, tropical dry forest, grassland, irrigated agriculture, rain-fed agriculture, orchards, lava flows and human settlement. The two classifications were evaluated with independent (from the training data) verification data, and the results showed that with EVI data, the classification accuracy was significantly improved, at 0.01percent level, evaluated by McNemar's test.