基于改进Retinex算法的雾霾天气车道线识别  

Lane detection in haze weather based on improved Retinex algorithm

作　　者：周劲草[1] 魏朗[1] 刘永涛[1] 张在吉 田顺[1] 

机构地区：[1]长安大学汽车学院,陕西西安710064 [2]中国民航大学天津市智能信号与图像处理重点实验室,天津300300

出　　处：《东北师大学报（自然科学版）》2016年第3期54-59,共6页Journal of Northeast Normal University(Natural Science Edition)

基　　金：国家自然科学基金资助项目(51278062)

摘　　要：在雾霾天气下,针对常规车道线识别方法提取车道线准确性差以及多尺度Retinex算法去雾图像质量较低的缺点,提出了一种基于改进视网膜大脑皮层理论(Retinex)的雾霾天气车道线识别算法.首先,利用改进的暗通道优先算法对雾天图像进行去雾,将去雾图像作为多尺度Retinex算法的输入图像做进一步增强;然后将多尺度Retinex算法增强的图像进行亮度修正,从而获取理想去雾图像;再利用Scharr滤波器和大津法得到包含清晰道路边缘的二值化图像;最后利用Hough变换对车道线精确提取.实验表明,该算法不但能够在雾霾天气下对车道线进行准确的识别,与常规算法相比,能够有效地提高图像质量,并且具有良好的实时性,对于提高车辆主动安全性具有重大意义.In this paper,a new algorithm based on an improved Retinex algorithm was proposed for lane detection in haze weather which couldn＇t be detected by traditional algorithm.Firstly,the lane images in haze weather was enhanced by an improved dark channel prior algorithm and using defogged images as the input image for MSR algrithm to have a further enhancement,ideal image was obtained after brightness enhancement.Binary images of road edges was obtained by Scharr filter and Ostu algorithm then.Finally,the road lane was extracted by Hough transform.Experimental results showed this new algorithm could not only detect road lane in haze weather accurately,but also could improve image quality effectively and has better real-time.Thus has great influence on the improvement of automobile active safety.

关 键 词：雾霾 视网膜大脑皮层理论 车道线识别 图像增强 车辆主动安全性 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]