Detecting difficult roads and intersections without map knowledge for robot vehicle navigation

Abstract

SCARF is a color vision system that can detect roads in difficult situations. The results of this system are used to drive a robot vehicle the Navlab on a variety of roads in many different weather cOnditions. Specifically SCARF has recognized roads that have degraded surfaces and edges with no lane markings in difficult shadow conditions. Also it can recognize intersections with or without predictions from the navigation system. This is the first system to be able to detect intersections in color images without a priori knowledge of the intersection shape and location. SCARF uses Bayesian classification a standard pattern recognition technique to determine a road surface likelihood for each pixel in a reduced color image. It then evaluates a number of road and intersection candidates by matching an ideal road surface probability image with the results from the Bayesian classification. The best matching candidate is passed to a simple path planning system which navigates the robot vehicle on the road or intersection. This paper describes the SCARF system in detail and presents some results on a variety of images and discusses the Navlab test runs using SCARF.