基于特征融合的视觉导航智能车辆的车道识别  

Lane Identification of Vision-Guided Intelligent Vehicle Based on Feature Fusion

作　　者：李进[1] 陈杰平[1] 易克传[1] 徐朝胜[1] 

机构地区：[1]安徽科技学院机械工程学院,滁州233100

出　　处：《汽车工程》2015年第5期587-592,共6页Automotive Engineering

基　　金：国家自然科学基金(51075112);安徽高校省级自然科学研究项目(KJ2013B074);安徽科技学院自然科学研究项目(ZRC2014409)资助

摘　　要：为兼顾车道识别的实时性和鲁棒性,有序融合了图像中车道线边缘、边缘方向、车道宽度和边缘灰度等特征信息。同时,在识别第1帧车道图像时对整帧图像进行处理;在识别后续各帧车道图像时,根据上一帧图像的识别结果确定动态约束区,并在约束区域内完成识别。实车道路试验结果表明,通过有序融合各种特征信息和合理设置动态约束区能有效识别路面车道线,且车道跟踪具有较好的精确性和可靠性。For achieving both the robustness and real-time performance of lane idenfication, the feature information on the edge and its direction and gray scale of lane markings and lane width in the image are sequentially fused. In identifying the first frame of lane image, the whole image frame are processed, while in identifying the fol- lowing frames, dynamically constrained areas are determined according to the identification results of previous frames, within which identification completes. The results of real vehicle road tests show that by orderly fusing vari- ous feature information and properly setting dynamically constrained areas, lane marking identification can be effec- tively achieved with better accuracy and reliability in vehicle lane tracking.

关 键 词：智能车辆 车道识别 特征融合 动态约束 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]