Remote sensing images classification is one of important approaches for recognizing interesting ground object. It is because there are several problems for normal classification methods, for example more manual intervention influence effect of classification, neighborhood information couldn't be utilized adequately, there is bad robustness to environment and so on, that particle swarm classifying optimization algorithm is put forward in this paper. New algorithm can take full advantage of neighborhood information and is with better robustness to operation environment, and classification results are with more human. In the meanwhile, intelligence classification algorithms have better executive retractility, so they could be applied abroad to fast classification applications or those with high-quality requirement.