一种针对道路潜在安全隐患的跟踪学习方法

Tracking Learning Method on Traffic Hiding Dangers

提出了一种对存有安全隐患车辆的跟踪学习方法,对传统的在线学习方法进行改进,通过改善学习性能,使算法在不稳定的安全隐患车辆跟踪中发挥作用,实验结果表明,在同等条件下,该算法能够获取更多有效样本,使得分类器更快收敛。

This thesis proposes a tracking method on danger-hiding vehicles, which betters traditional online learning method. Through promoting learning efficiency, this method puts into work in the unstable tracking of danger-hiding vehicles. The experiment data tells that this method, under the same conditions, can get more effective samples and make classifiers having much faster convergence.