Directional feature: a novel feature for group target detection in high resolution SAR images

In this letter, we propose a new feature for group target detection in high resolution Synthetic Aperture Radar (SAR) images. This study aims to reduce the false alarm rate by adding a novel directional feature to typical SAR target detection algorithms. Unlike other shape- or contrast-based features, the directional feature contains the orientation and angle information of targets. Based on this feature, we can distinguish group targets from false alarms by analysing the correlation of their directional features. First, the proposed feature extraction approach generates an enhanced image to overcome speckle noises and a low signal to noise ratio (SNR). Next, a contour extraction algorithm is used to generate a line-drawing edge map of the target. Finally, the directional feature is obtained based on the major principal axes of the edge map. The experimental results using real high resolution SAR images verify the validity and effectiveness of the SAR detection algorithm.