道路标线图像分割方法研究  

Road Markings Image Segmentation Method

作　　者：刘新宇[1] 吴勇[1] 李龙[1] 

机构地区：[1]武汉理工大学水路公路交通安全控制与装备教育部工程研究中心,武汉430063

出　　处：《交通与计算机》2008年第6期56-60,共5页Computer and Communications

基　　金：国家自然科学基金项目(批准号:50578128)资助

摘　　要：采用最佳阈值法、最大类间方差法、最小误差法分别对多幅典型的破损道路标线图像进行分割,通过分析对比确定用最小误差法分割道路标线图像效果较好。但是采用最小误差法分割道路标线存在无法正确分割出细节的缺陷,为改善图像的分割效果,进一步探讨并决定采用基于最小误差的动态阈值法对道路标线图像进行分割。通过对比全局阈值法和动态阈值法的优缺点,并结合两者的优点,提出了动态阈值结合全局阈值的方法分割道路标线图像,取得较好的效果。The optimal threshold method, maximum class square error method and minimum error threshold method were used to segment the distressed road markings from some typical images in this paper. From these segmented images, it is concluded that the segmentation by using minimum error threshold method is the best, though some details cannot be segmented accurately. In order to achieve more perfect segmentation effect, the dynamic threshold method based on the minimum error has been discussed in this paper. Through comparing the global threshold method with dynamic threshold method, it is concluded that the dynamic threshold method integrated with global threshold to segment the road markings has better segmentation effect.

关 键 词：道路标线 图像分割 最小误差阈值法 动态阈值法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]