Detection of Lateral Borders on Unmarked Rural Roads

This work presents a method to detect roads in rural conditions, considering gravel or stabilized roads. The main features of this type of ways are based on they do not have lanes and either lateral or central delimitations. This type of road is common in several countries that, for their infrastructure road, still have a large percentage considered as belonging to red earth without paving or delineating. In order to detect the edges of the paths, this work proposes an analysis strategy based on channel selection on HIS color map representation. The selection automatically allows detecting the channel which brings better information. A convolution filter sequence combined with an adaptive Hough Transform is then applied to detect of curves in the way. Keywords—driver assistance; rural roads; image processing