Potential of Road Stereo Mapping with RADARSAT Images

Two stereo pairs generated with standard mode images (S1-S7 and S4-S7) and one with fine mode images (F1-F5) are used to evaluate the potential of RADARSAT-SAR for extracting planimetric features, such as roads on a PC-based stereo workstation. First, monoscopic and stereoscopic plotting for the GCPs are performed to evaluate the impact on the accuracy. It is prerequisite, especially for smaller intersection angle stereo pairs, to acquire GCPs in stereoscopy since monoscopic collection mode degrades the relative and absolute orientations of the stereo model with a ratio of two to four depending of the stereo geometry. The roads are then interactively stereo extracted by an operator and compared with the roads of the digital topographic maps. Statistical results over a large sample (more than 900 km) show accuracy of about 15-24 m for fine mode and 25-50 m for standard mode stereo-pairs with 90% confidence levels, independently of the stereo configuration. The impact of image sampling on the road positioning accuracy is also addressed. Finally, comparisons with the ortho-rectification process show that the stereoscopic method to extract planimetric features is slightly more accurate.