Remote Sensing Image Classification for Forestry Using Mrf Models and Vq Method

Conventional pixelwise classi cation techniques have two drawbacks. The rst is that they tend to occur isolated misclassi cations because they classify each site independently using spectral information only. The second is that sample data which represent each class are indispensable to estimate model parameters or to train classi cation Neural Networks. In this manuscript a new unsupervised contextual method is proposed for multispectral remote sensing image classi cation. Markov Random Field (MRF) models are used for modeling the observed and classi ed images in the proposed method. Both spectral and spatial information can be exploited by the MRF models so as to dissolve contextual inconsistency caused by the rst drawback. In addition the Vector Quantization (VQ) method is introduced to dissolve the second drawback. The VQ method classi es the observed data into several clusters without using any sample data. The image classi ed by the VQ method is so coarse that it includes misclassi cation and unknown-class sites, however it can be used as an initial classi ed image for the following MRF-based classi cation algorithm. The proposed method was applied to LANDSAT Thematic Mapper data to discriminate deciduous trees from ever-green trees. The accuracy of classi cation was 64.4% by the VQ method and it was improved up to 88.8% by the MRF-based method.