Roust Vanishing Point Detection Based on the Combination of Edge and Optical Flow

Abstract

Robust vanishing point detection is an important machine vision task and is widely used in pattern recognition for robotics, advanced driver assistance systems, and autonomous driving vehicles. At present, there are many vanishing point detection algorithms, but it is still difficult to satisfy both the accuracy and real-time requirements. Assuming that the moving direction of the vehicle is parallel to the road boundaries or lane markings, a robust vanishing point detection method based on the combination of edge and optical flow in a stream of urban scene image is proposed in this paper. The proposed scheme was applied to databases that included illumination, partial occlusion, and viewpoint changes to validate robustness. Experimental results show that the proposed technique has higher vanishing point detection accuracy than the edge-based and texture-based detection algorithm, lower complexity than the texture-based method, and can be realized in real time. The algorithm achieves a better compromise between efficiency and effectiveness in detecting vanishing point compared with some state-of-the-art methods.