基于改进的暗通道先验去雾辅助导航算法  

Improved Dehazing Aided Navigation Algorithm Based on Dark Channel Prior

作　　者：孙晓峰[1] 宫金良[1] 张彦斐[1] SUN Xiao-feng;GONG Jin-liang;ZHANG Yan-fei(School of Mechanical Engineering, Shandong University of Technology, Zibo 255049, China)

机构地区：[1]山东理工大学机械工程学院

出　　处：《科学技术与工程》2019年第25期168-173,共6页Science Technology and Engineering

基　　金：国家自然科学基金(61303006);山东省引进顶尖人才“一事一议”专项经费(2017ZBXC151)资助

摘　　要：经典的去雾算法无法满足车道线检测的实时性和准确性要求,因此提出一种改进暗通道与边缘检测融合的雾天车道线识别算法。首先对有雾图像进行对比度增强处理,突出边缘、颜色等有效信息,基于道路先验信息对图像进行感兴趣区域处理,利用暗通道先验算法对静态约束图像进行去雾操作,并通过双边滤波器细化透射率图,得到清晰的去雾图像;然后引入动态约束理念,提取车道线可能存在的区域,借助Sobel算子检测动态约束后的车道线区域,提取车道线边缘点;最后利用Hough变换进行准确的车道线拟合。实验表明,改进的去雾算法得到的图像清晰度与对比度更高,满足了车道线检测的准确性与实时性要求;去雾及车道线检测算法平均处理时间为297. 305 ms,满足无人驾驶时间要求。The classical defogging algorithm can not meet the real-time and accuracy requirements of lane line detection. Therefore,a lane line recognition algorithm based on modified dark channel and edge detection is proposed. Firstly,the original haze image is processed by contrast enhancement to highlight the edge、color and other effective information. Based on the road prior information,the image is processed by the region of interest. The dark channel prior algorithm is used to defog the statically constrained images. The transmission image is refined by bilateral filter to get the clearly defogging image. Secondly,the concept of dynamic constraint is introduced to extract the possible regions of lane lines. Sobel operator is used to detect the region of lane lines after dynamic constraint. And the edge points of lane line are extracted. Finally,Hough transform is used to accurately fit the lane line. The experiment shows that the defogging image obtained by the improved algorithm has higher definition and contrast. It meets the requirement of accuracy and real time of lane line detection. The average processing time of defogging and lane line detection algorithm is 297. 305 ms,meeting the requirements of unmanned driving time.

关 键 词：双边滤波器 图像去雾 车道线检测 动态约束 暗通道 

分 类 号：TH391.41[机械工程—机械制造及自动化]