A combined approach to single-camera-based lane detection in driverless navigation

Abstract

A combined approach is presented with real-world examples to illustrate how to detect lane lines under several road conditions in driverless navigation. This approach is designed for single-camera perception and is achieved using camera calibration, un-distortion, color thresholding, perspective transformation, pixel identification and ployfitting. Limiting factors such as illumination and radius of curvature of the testing road is considered. The results show that the method is effective in normal conditions but needs further improvement in extremely challenging environments.