SAR target recognition based on texture feature and contour feature fusion

This paper proposes a SAR image recognition method that combines contour features and texture features. The moment feature is selected as the contour feature, and the HOG feature is selected as the texture feature. Considering that the single feature recognition method has limitations and the accuracy of multiple classifications is poor, the feature fusion method is used for optimization. This paper proposes a new feature-level fusion idea, try to use Fisher scoring method to select features, at the same time use FS as the fusion weight, and reduce the dimensionality of the moment features by PCA according to the weight value to obtain the fusion feature matrix. The recognition performance is better than ordinary weighted fusion algorithm. At the same time, the idea of heterogeneous feature fusion is given, that is, decision-level fusion.