摘要
对于智能车辆在一些常见复杂环境下的道路识别,传统的预处理过程要对图像进行二值化,造成大量有用信息的丢失。文中直接利用灰度图像上道路边界的灰度及其梯度信息,构建基于小块统计的目标函数,用于评价道路边界的拟合质量,起到滤波的作用,有效地剔除复杂环境下的各种不规则纹理噪声。试验结果表明,本方法能有效地消除光照条件和树木阴影等因素的不良影响,准确地识别直线或弯曲道路边界。同时由于无须对图像进行预处理,大大提高了识别的实时性。
In road identification under complicated conditions for intelligent vehicle, the procedure of traditional image preprocessing includes binarization, which causes the loss of a large number of useful information. In this paper, the grayscale and grayscale gradient of road boundary in monochrome image is directly used to construct objective function based on grid counting, which is used to evaluate the fitting quality of lane boundary, and playing a role of wave filtration, effectively reject all kinds of irregular texture noises in complex conditions. The test results show that the method can eliminate effectively the negative effects of illumination conditions or tree shadows etc and accurately identify linear or curved lane boundary. In addition, due to the omission of image preprocessing, the real-timeness of identification is enhanced greatly.
出处
《汽车工程》
EI
CSCD
北大核心
2010年第4期351-355,共5页
Automotive Engineering
基金
河北省教育厅科研项目计划(Z2008472)和(2006326)资助
关键词
智能车辆
灰度图像
道路边界线识别
intelligent vehicle
monochrome image
road boundary recognition