摘要
为提高在不同光照条件下道路识别算法的适应性,采用从局部区域提取方向、道路边界梯度和灰度等3个特征的方法,该特征的提取几乎不受不同光照导致的道路图像全局灰度变化和阴影等噪声的干扰。共选取了14种特征单元,将道路的方向特征准确融入到局部梯度和局部灰度值中。通过应用DS证据理论,将道路边界3个特征和有效统计单元所占比例进行有效信息融合,使得目标函数可以准确拟合曲线的总体质量。在大量实验样本和光强信息的基础上,设定不同光照条件下的道路边界检验阈值选取范围,提高了识别的实时性。实验结果表明,该方法可以适应各种不同光照条件下的道路边界识别,且具有良好的准确性和实时性。
In order to enhance the adaptability of lane identification algorithm to different lighting conditions,the three features of direction,gradient and grey scale of lane boundary are extracted from local areas,which are hardly interfered by the noises（global grey scale variation and shadows etc.） in road image caused by different lighting conditions.All together 14 kinds of feature units are selected,and lane direction feature is accurately fused into the values of local gradient and grey scale.The effective information fusion of three features of lane boundary and the proportion of effective statistical units is performed by applying DS evidence theory,so the overall quality of curves can be accurately fitted by objective function.Based on a great number of experiment samples and lighting information,the ranges of test thresholds for lane boundary are set for different lighting conditions to improve the realtimeness of identification.The results of experiment show that the method can adapt to different lighting conditions for lane boundary identification with good accuracy and realtimeness.
出处
《汽车工程》
EI
CSCD
北大核心
2011年第8期707-712,共6页
Automotive Engineering
基金
教育部科学技术研究重点项目(211024)
河北省科学技术研究与发展计划项目(10215652)资助
关键词
道路识别
光照条件
DS证据理论
lane identification
lighting conditions
DS evidence theory
