欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
针对单目视觉技术的道路理解算法难以在效率和鲁棒性间达到平衡的问题,提出一种基于多特征信息融合、优化的道路理解算法;算法在初始化阶段,融合边界消失点、路面消失线和车速信息,分别提出基于最小风险函数的道路区域分割算法;在道路识别阶段,首先使用方向随动滤波器对图像进行预处理,然后融合道路方向、边界平行度、像素灰度和其他先验知识,构造车道置信度函数,采用概率Hough变换提取道路边缘;在道路跟踪阶段,提出基于粒子对滤波的道路跟踪算法,对道路进行跟踪。测试结果表明：该方法在结构化道路环境中可以快速准确提取车道线,并且对典型道路干扰具有很好的免疫作用。
英文摘要：
Authors proposed a lane understanding algorithm based on multi-feature fusion and optimization,which was expected to overcome the unbalance between efficiency and robustness with normal monocular-vision technology.In the initialization stage,authors proposed a road segmentation algorithm using smallest-risk-division algorithm,by combining vanishing point,road disappearance line and speed information.In the recognition stage,first,authors preprocessed the input images with the steerable filter.And then,combining road direction,boundary parallel,pixel gray and other prior knowledge,authors constructed confidence-lane structure function and chose probability Hough transform to extract the road edge.In the tracking stage,in order to track the recognized road-lines,a double-particle filter algorithm was presented.The results indicate that the algorithm can extract lane quickly and precisely and is immune to typical interference.
同期刊论文项目
同项目期刊论文
期刊信息
位置：