It is an important research direction to classify military images.The classification accuracy based on traditional visual features is not high due to the high similarity of different military targets in complex background.A method of military image classification is proposed by combining Local Binary Pattern (LBP) with Pyramid Histogram of Oriented Gradients (PHOG) through Local Discriminant Canonical Correlation Analysis (LDCCA).Firstly, the LBP and PHOG features of military images are extracted, and then the extracted features are fused by LDCCA.Finally, K-nearest neighbor classifier is used to classify the military images.The advantage of this method is that the fused LBP and PHOG features are robust and able to classify images.The results on a data set of military targets show that this method is effective and feasible, which provides a technical reference for the military target identification system.