Forest Information Extraction Based on GIS and ETM+ Image Features

It is especially significant for improving classification accuracy to make use of effective multiple features of remote sensing data and a suitable classification method. According to the feature knowledge, a method combined supervised classification with knowledge classification is presented to extract the forest information in Zayu County in this paper. In order to eliminate the influence of the shadow, the texture feature is used to obtain it, which is received from the image fused the panchromatic band with multispectral band from ETM+ image. And then, the gray values below the shadow are resumed by means of histogram matching. On the process of information extraction, the optimal band combination which is generated by difference or ratio approaches based on spectral feature, GIS data which includes DEM, slope and aspect, and the habitat difference of species are considered as classification knowledge to establish the corresponding expert rule of each object. Appling this, the total accuracy of classification and the kappa coefficient reach 84.22%, 81.83% respectively. KeywordsGIS; ETM+ image feature; information extraction; supervised classification; knowledge-based classification; accuracy assessment