Computationally efficient vanishing point detection algorithm based road segmentation in road images

Road segmentation is one of the most challenging task in a vision based autonomous road navigation system. Adaptability to detect various types of roads and reduction in computational complexity are the two main challenges in existing vision based road navigation algorithms. This paper modifies existing vanishing point based road segmentation algorithm and presents computationally efficient reduced voter scheme for general road segmentation. The goal of this paper is to present an accurate and precise road segmentation algorithm that can work on structured, semi-structured and unstructured roads in presence of different surrounding environments and various illumination variations in daytime. The major contribution of this algorithm is reduction of total number of voters used for voting process of vanishing point detection and analysis of general road segmentation results for various types of roads. The algorithm is implemented on MATLAB. Various qualitative and quantitative results are presents.