Realtime Road Detection by Learning from One Example

Real-time detection and localization of a road from an aerial image is an emerging research area that can be applied to vision-based navigation of unmanned air vehicles. Existing real-time and non-real-time road detection algorithms focus on pre-defined road types, and a single algorithm cannot handle a large variety of road types such as dirt roads, local streets, and freeways. An algorithm to detecting any types of corridors is presented. First, a corridor structure is automatically learned at runtime with a single example. The corridor structure is represented as a cross-sectional 1-D signal segment. The learning procedure is to find the maximum correlation of such signals. The real-time detection consists of 1-D signal matching and robust fitting on the matching result. Real-time detection results on various road images are presented