Particle filter based road detection in SAR image

A novel method is proposed for the detection of roads in synthetic aperture radar (SAR) images. We first claim that roads in SAR images can be extracted in a Bayesian tracking framework. Then the particle filter is employed for the tracking implementation, where local linear feature and global contextual knowledge are incorporated into the filtering procedure. The effectiveness of the proposed method is demonstrated by the experimental detection results of single road as well as road network.