Study of target detection of SAR image using directional roughness feature

Based on the roughness feature indicating texture information, the method of the directional roughness feature is studied deeply for the target detection of SAR images. The directional roughness feature of each pixel can be extracted from the exponential wavelet filtered image and the energy function in one direction. The optimal parameters of the method are determined for SAR images of MSTAR and ADTS database respectively. For above two different band SAR images, the target detections are implemented by using the directional roughness feature and CFAR. The comparison of the detection results showed the directional roughness feature could not only detect all size-fixed targets but also have lower false alarm rate, higher spatial resolution and more accurate locations of the detected targets.