Land-Cover Hierarchical Classification Method Study of TM Image in Loess Hilly Ravine Area

This article puts forward a land-cover hierarchical classification method of TM image in loess hilly ravine area. This method selects different typological samples of ground object, analysises the statistics feature of the samples, confirms the TM image's hierarchical classification trees. The hierarchical classification trees exist more difference in the classification Node, and have higher divisibility. Based on the ground object's spectrum feature, this method put forwards band selection and feature extraction scheme according to different ground objects. This job makes the ground objects having higher divisibility in the selective bands. This article does precision evaluation with the classification results by means of confusion matrix, and compares the classification results with the results of maximum likelihood supervised classification. The land-cover classification accuracy has been increased.