Application of a Hybrid Classification Technique on the Forest Classification

Based on the good classification results to the identification of the forest and non forest using the blending method of the combination of unsupervised and supervised method, the “Iterative Guided Spectral Class Rejection” classification method was probed. First the theory of the “Iterative Guided Spectral Class Rejection” algorithm was described. Then we developed the algorithm to make a classification experiment using a multi temporal composite image in the same site. Compared with the maximum likelihood classification method, the “Iterative Guided Spectral Class Rejection”algorithm can effectively reduce the problem of training data miscellany caused by the human factors such as distinguishing inaccurateness of the spectral purity using the automatic cluster ability to the same spectral character sort of the unsupervised method, assisting with the acquirement of the training data. Thus, the IGSCR algorithm can effectively increase the classification accuracy.