Abstract: Using projection pursuit learning network (PPLN), a new classification for remote sensing image is proposed. The PPLN algorithm integrates the advantage of artificial neural network (ANN) with nonparametric statistical technique, projection pursuit algorithm (PP), which is capable of providing less network neurons and good robustness. In this study, the structure and improved learning algorithm of PPLN is introduced in detail. Using the TM image of Suzhou, an experiment of classification is done and the classification precision is superior to that of BP neural network and conventional maximum-likelihood.