A Novel Method of Landmines Detection Based on Improved SVM

To detect landmines in real SAR images, detection process based on PCA and SVM techniques is proposed in this paper. Firstly, a referential deflection-optimal linear-quadratic(DOLQ) detector is presented. Then, the proposed centering PCA+SVM method is developed and compared with the above benchmark. A morphological processing procedure is used to remove the isolated spurious pixels and to form landmine objects in the binary images. Finally, the experimental results show that the novel method can constantly outperform the DOLQ detector in terms of POD(probility of detection) and the SVM method in terms of training and classification times in landmine detection.