摘要
与传统的车道线检测算法不同,本文采用LDA算法对道路图像进行针对性灰度化处理。加大车道线与道路的差异,然后使用抛物线模型对车道线进行拟合,采用混沌粒子群算法对抛物线参数进行优化,以车道线的灰度特征和梯度特征作为混沌粒子群的适应度函数,经过多次的迭代得到抛物线拟合车道线的参数最优值,进而识别出车道线。实验结果表明,本文算法能在复杂环境下识别出车道线,对视频帧序列中的车道线连续追踪具有良好效果。
The method proposed in this paper is different from the traditional road lane detection algorithm. Firstly, LDA（Linear Discriminant Analysis） is used to convert road images to gray images directly, increasing the difference between the lane and the road. Secondly, the road lane is fitted by parabola model, and CPSO（Chaos Particle Swarm Optimization） is applied to optimize the parameters of parabolic model. The CPSO fitness function consists of features of grayscale and gradient. After several iterations, the optimal parameter value of parabola can be obtained, and then the road lane in the image can be detected out. Experiment results show that the proposed algorithm could detect road lane in complex environment. Also, it has a good performance on road lane tracking in the video frame sequence.
出处
《液晶与显示》
CAS
CSCD
北大核心
2017年第6期491-498,共8页
Chinese Journal of Liquid Crystals and Displays
基金
广西自动检测技术与仪器重点实验室主任基金项目(No.YQ14105)
广西省科学研究与技术开发计划资助项目(No.桂科攻:11107001-40)~~
作者简介
通信联系人，E-mail：hxg@guet．edu．cn黄新（1978-），男，湖北黄冈人，硕士，副教授，主要研究方向：可测性设计，图像处理。
