Semi-Automatic Road Extraction Algorithm for High Resolution Images Using Path Following Approach

An approach for the semi-automatic extraction of roads from high-resolution satellite imagery is proposed. Scale space, Edge-detection techniques are used as pre-processing for segmentation and estimation of road width. The detection of road is based on the cost minimization technique. The cost is estimated by taking various factors into consideration like variance, direction, length and width of the road. The process starts with the selection of seed points provided by the user. The approach is called as ‘path following’ as it follows the path having minimum cost repetitively. Thus the path having minimum cost will be considered as a part of the road. The algorithm puts weights on the directions for estimation of cost of the path. The algorithm works successfully for roads having junctions and minimum obstacles and detects central line of the roads in any orientation with moderate curvature accurately. Current limitations are that the algorithm may not work on the road cast by shadow and the user must select valid road seeds since algorithm cannot judge the validity of input seeds. The algorithm requires input roads of considerable width. The roads that will not give edges in edge detector will not be extracted . The high-resolution images are preferable due to use of width and variance information for road extraction. The contribution of this paper is that it showed cost minimization and path following approach with pre-processing by scale-space and canny edge detector works well for road extraction. It also handles tracing of junctions of the road. Seed points need not be selected on road central lines and each road segment need not be given separate seed points.