Multi-scale target detection in SAR image based on visual attention model

This paper proposes a novel method for synthetic aperture radar (SAR) target detection by using multi-scale SAR images based on visual attention model, which can automatically find the vehicle targets from the complicated background with clutters such as trees and buildings. In our method, firstly, a saliency map is obtained from a Gaussian pyramid of the original SAR image, where the image scales are selected based on the prior size information of the targets to be detected in the image. Secondly, we use the method based on shifts of the focus of attention (FOA) in the saliency map to get a binary image. Finally, the clustering algorithm based on the prior length of targets is employed to extract the target candidate chips in the binary image. In the experiment based on the real SAR image, we compare the proposed method with the classical constant false alarm rate (CFAR) target detection method, which indicates that our method can detect vehicle targets in the image more quickly and with fewer false alarms.