SVM-based military target recognition has the characteristics of fast speed and good robustness, and accurate recognition relies on image feature selection.In field environment,the targets image is affected by complex interference factors.In view of the problem,a compound feature selection method is proposed.The visible and infrared images obtained are evaluated,in which the discriminative ones are selected and segmented based on Soft method.Then,moment-invariant features are extracted and feature fusion is made to provide reliable feature vectors for SVM.