基于离散余弦变换的车道识别算法  

An Algorithm of Lane Detection Based on Discrete Cosine Transform

作　　者：朱红[1] 吴长奇[1] 

机构地区：[1]燕山大学,河北秦皇岛066004

出　　处：《计算机工程与应用》2005年第30期94-95,98,共3页Computer Engineering and Applications

摘　　要：车道识别技术是自动驾驶领域的研究热点。通过对车载摄像机获取的公路车道线图像特征的分析,提出了一种基于图像频率域特征的车道识别算法。该算法的核心是提取车道线在离散余弦变换域的特征,再结合道路模型的先验知识利用贝叶斯决策原理,识别出路面上的车道线。实验表明,该算法在不同的路面和光照条件下均可得到良好的识别效果。Lane detection technology has attracted increasing attention in the field of autonomous driving.We analyze the characteristic of road images from CCD mounted on a car,and present an algorithm for lane detection based on frequency domain feature.The core method of our algorithm is abstracting the Discrete Cosine Transform feature of the images.Combined with a deformable template prior,using Bayes decision making principle the lane marker can be detected.Experiment has proved that the results are satisfying for lane detection under the conditions with varying lighting and environments.

关 键 词：车道识别 离散余弦变换 贝叶斯决策 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]