ハフ変換とパーティクルフィルタを用いた左右レーンマーク検知

概要

論文の詳細を見る
This paper proposes an algorithm to detect lane marks on a road very fast by means of the Hough transform and the particle filter methods. The proposed method is composed of two parts, the voting part based on the Hough transform as a bottom-up process and the parameter estimation part using the particle filter as a top-down process. In the bottom-up process, the method extracts candidate lines by voting procedure for each lane marks features. In the top-down process, the method verifies the lane model hypotheses using the voting result of lines extraction. The bottom-up process can be executed fast because the parameter space in extracting lines is low-dimensional, and the top-down process becomes more efficient by using the detection results of primitive shapes. Experiment results show that the proposed method can detect lane marks robustly, and can achieve an average speed 0.06ms/frame (Intel Core i7 3.2GHz) with an average position error 9.42cm.