摘要
在路径识别中设置搜索感兴趣区以提高弯曲道路边界的识别实时性。用抛物线模型拟合弯曲道路边界,根据道路边界位置不会发生突变的特性,在识别出的道路抛物线模型上选取有代表性的3个点,针对每个点建立预测模型,应用Kalman滤波理论准确预测道路边界的位置,并据此设置搜索感兴趣区。依照合理的目标函数,在感兴趣区范围内搜索并确定抛物线参数,从而将搜索范围较为准确地限定在较小区域,能够有效提高识别的实时性。试验验证该方法在大大提高识别实时性的同时,增强了识别的精确性和鲁棒性。
The searching region of interesting was set to cut cost time of identifying bent lane edge. The parabola mode was used to fit bent lane edge. According to the fact that the location of lane edge cannot break, three representative points in identified parabola were selected, forecast model was established aiming at each point, Kalman filtering method was used to forecast the position of the lane edge accurately and the searching region of interesting was set consequently. Parabola parameters were searched in the region of interesting based on suitable goal function, so the searched area was limited accurately in a narrow range. It is confirmed by experiments that the proposed method can shorten the real time of identification greatly and enhance the reliability and robustness of identification.
出处
《公路交通科技》
CAS
CSCD
北大核心
2009年第1期134-138,共5页
Journal of Highway and Transportation Research and Development
基金
河北省教育厅科研计划项目(2006326)
燕山大学博士基金资助项目(B164)
关键词
交通工程
感兴趣区设定
KALMAN滤波
弯曲路径识别
机器视觉
traffic engineering
setting the region of interesting
Kalman filtering
bend lane identification
machine vision
