Forest cover-type classification using SPOT4 and SPOT5 images

This study focuses on the forest cover-type classification using remote sensing techniques. The SPOT4 and SPOT5 images, which cover the Chilanshan mountainous area located in the northeastern part of Taiwan, are applied to classify forest cover types into three classes using supervised classification approach. The classified three classes are natural Chamaecyparis formosensis forests, coniferous plantations, and nature broadleaf forests, respectively. To improve the classification accuracy of forest cover types, SPOT5 image (spatial resolution is 10m) and SPOT4 image (spatial resolution is 20m) were integrated together by the mosaic technique. In addition, the ground permanent samples established for continuous forest inventory were also used to assess the classification accuracy of forest cover types. The result indicates that the classification accuracy in SPOT5 is better than in SPOT4 because of the finer resolution. However, among these three classes, the natural broadleaf forests have low accuracy due to the complicated stand structure.