车辆捕获中的补偿算法研究

Research on Compensation Algorithm for Vehicle Capturing

应用于视频监控与流量统计的车牌识别系统,其核心模块是对图象中车辆牌照进行识别,准确地捕获车辆的到来事件是提高车牌识别率的有效方法之一。传统的基于相邻帧的差值累计方法在捕获车辆到来事件与离开事件时存在着重复捕获同一车辆的到来事件,以及抗干扰性不强等缺陷。在传统方法的基础上,提出了一种补偿算法,即对传统方法进行修正与扩充,特别是对车道背景的补偿。实验表明:该算法显著地提高了车辆捕获的准确性。

Vehicle＇s license plate recognition system has been used in video surveillance and traffic management. The essence is to recognize license plate characters. It＇s an effective method to improve recognition by capturing vehicle＇s arrival and departure accurately. Traditional way tends to multi-capture those events, which is based on difference accumulation of two consecutive images. Meanwhile, it＇s not noise-proof. In this paper, a revised algorithm is presented, including a weighted difference accumulation and compesation to the background. It has been proved by experiments that the proposed algorithm promotes the accuracy in vehicle capturing.