Semi-automatic Road Extraction from SAR Images Using an Improved Profile Matching and EKF

As one of typical linear targets, road extraction is the hot spot of remote sensing image interpretation. Accurately and efficiently extracting roads from Synthetic Aperture Radar (SAR) images is of great significance. SAR which can not only acquire data regardless of weather and time, but also provide valuable information on geophysical parameters widely used in road extraction. The traditional method of road extraction based on automatic and semi-automatic method. Because of the complexity of the roads, such as disturbing of the trees along or cover the road, building to cover, cars getting on the road et al, automatic method have some missed or erroneous and need abundant post-processing. The semi-automatic methods that interact with a human operator are considered to be a good compromise between the calculation speed of a computer algorithm and the interpretation skills of the operator. In this paper, a novel semi-automatic road extraction method based on improved profile matching and extended Kalman filtering (EKF) using SAR imagery is introduced. In our method, a road extraction model is built firstly. And then accurate observations are obtained through improved profile matching. Finally, EKF is adopted to update the observations to get the optimal estimates of the road. The effectiveness and steadiness of the proposed method is demonstrated through two experiments using data of Howland, Maine by UAVSAR in L-band and data of Lingshui, Hainan by airborne SAR in X-band. The results of road extraction show that the proposed method is effective, less human intervention, and accurate in the some special scene road of SAR imagery.