Remote Sensing Target Recognition Based on Contourlet and Kernel Fisher Discriminant

An efficient feature extraction method for remote sensing target recognition was proposed in this paper, which was based on contourlet and kernel Fisher discriminant (KFD). After the contourlet decomposition, the contourlet features are fused (the weight for fusion is chosen by cross validation), and then KFD was used for further feature extraction, finally k-nearest-neighbor (KNN) was used for classification. Experimental results show that the proposed feature extraction method reach a higher correct rate than KFD and the method which use KFD on the lowpass filtered images. Moreover, when dealing with large scale images our method achieves a lower computation complexity than KFD