基于融合策略的套牌车主动识别算法

A positive recognition algorithm for fake plate vehicles based on fusion strategy

为扩大套牌车的识别范围并提高识别精度与效率,提出了一种基于融合策略的套牌车主动识别算法。首先,基于同一辆车不可能在极短时间内出现在相同或不同地点的原理,利用车辆的拍摄时间差进行第1步识别;然后利用改进的多重标号算法计算出交通网络中任意两个监测点之间的前N条最短路径,进而得到车辆的加权平均速度,再根据实时交通信息计算车辆为套牌车的可信度,从而实现对套牌车的第2步识别。在实地采集的智能交通数据上进行的实验结果表明,所提算法能有效地实现对套牌车的主动识别。

To extend the recognition scope and improve the recognition precision and efficiency,apositive recognition algorithm for fake plate vehicles is proposed based on fusion strategy in this paper.First,the captured vehicle images are recognized by license plate recognition（LPR）algorithm.According to the recognition result of LPR,the time difference of capture is obtained for two vehicle images that are classified as the same.Based on the principle that it is unreasonable for a vehicle to appear in the same place or different places in an extremely short time,the first level of positive recognition is implemented by using the time difference information.Then an improved multi-label algorithm is adopted to calculate the Nshortest paths between any two monitor locations in the traffic network.A vehicle′s weighted average velocity is obtained by using the path distance and time difference.A reliability of fake plate is further calculated according to the real-time traffic information.Thus the second level of positive recognition is implemented by using the velocity information.Experiments are conducted on the real world data from an intelligent transportation System（ITS）.And the experimental results show that the proposed algorithm is effective to perform the positive recognition for fake plate vehicles.