结合卡尔曼滤波器噪声分析的车道线检测跟踪算法

使用卡尔曼滤波对视频序列图像中的具体信息进行跟踪的研究目前是跟踪方向的一个热点.但是在处理卡尔曼滤波跟踪过程中的过程噪声和测量噪声,大部分研究普遍采用的是初始赋值.通过不断的调整数,达到较好的跟踪效果.但是这样做不但没有遵循原始数据的规律,同时调整参数是一项耗时的工作.基于这个原因,提出了一种对卡尔曼滤波的过程噪声和测量噪声进行预估计的方法并将其应用到车道线跟踪过程中.通过对一部分离线数据进行处理,可以基本估计出系统的噪声参数.最后采用车道线跟踪算法对论文中的方法进行验证,实验证明,提出的参数估计方法在车道线的跟踪过程中达到很好的效果,同时处理每帧的时间为50ms左右,满足了实时性的要求.

Lane Detecting and Tracking Algorithm with Noise Analysis of Kalman Filter

Research of using Kalman filter to do specific information track in video image sequences is a hot topic in the tracking direction currently. But Most researches processed the noise of Kalman filter by assigning a specific value in the initial step. By constantly tuning the parameters the algorithm can achieve better tracking results. However, doing in this way not only did not follow the rules of the original data, but also adjusting parameters is a time-consuming job. For this reason, the paper proposes a Kalman filter process noise and measurement noise assignment algorithm which is used in lane detection. By processing part of the offline data, the algorithm can estimate the basic parameters of the noise. Use lane tracking to validate the algorithm in this article and experiments show that the proposed methods achieve good detection results while the time of processing each frame is about 50ms which meet the real-time requirements.