To improve the low recognition rate of mobile targets in infrared images because of fuzzy outline,this paper proposed a novel recognition algorithm for infrared images of mobile targets.Optimizing the BP neural network by bacterial foraging algorithm with adaptive step length,the feature vector is constituted.In doing so,the information of the leftmost and rightmost points of the target and target boundary upper the two extreme points are used.Using the feature vector,targets are recognized by the optimized BP neural network.Experiment result indicates that the proposed recognition algorithm improves the convergence speed,the accuracy of BP neural network and the recognition rate of mobile targets.It is effective especially for the partially occluded targets.