Lane Detection Based on Particle Filter in Urban Environments

• Main Authors: Hsin-Chih Chen, 陳信志
• Other Authors: Shih-Shinh Huang
• Format: Others
• Language: zh-TW
• Published: 2015
• Online Access: http://ndltd.ncl.edu.tw/handle/56108714650994175114

碩士 === 國立高雄第一科技大學 === 電腦與通訊工程研究所 === 103 === The main objective of this thesis is to propose a lane detection algorithm that can sucessfully detect lanes in urban environments. The main challenges that the prposed work should tackle with are the lanes in urban environament may be faded, casted by tree shadow, etc. The proposed method to achieve both lane detection and tracking is based on the particle filter framework and uses data-driven mechanism to draw the samples from the observed image. First of all, Hough transform is applied to detect all possible lines in the image and vanishing point is the one that has the minimum sum of distances to all lines accordingly. The voting space estimated by Hough transform is aslo used for drawing partcles so that the lanes newly appeared in the image can aslo be detected. This is referred to as data-driven sampling. The measurement , one stage of the particle filering framework, integrates gradient magnitude as well as vanishing points likelihood evaluation. As in estimation stage, it detect the lanes by limiting the search space using the vanishing point so that the real-time processing can be achieved. In the experiment, four videos from the urban scene are used to validate the effectiveness of our proposed method.