基于Gaussian模型及Kalman滤波的车辆跟踪方法  

Research on Vehicle Tracking Based on Gaussian Model and Kalman Filter

作　　者：丁晓娜[1] 

机构地区：[1]西安工业大学,陕西西安710021

出　　处：《计算机技术与发展》2016年第5期165-169,共5页Computer Technology and Development

基　　金：陕西省教育专项科研计划项目(14JK1341)

摘　　要：近年来,随着机动车增加,各大"堵城"陆续出现。各种交通问题日益增多,因此使得智能交通系统的快速发展迫在眉睫。文中在研究传统车辆跟踪方法的基础上,提出基于混合Gaussian模型和Kalman滤波的车辆跟踪算法。通过对车辆运行的外部环境和自身变换等问题的深入分析,首先采用背景减除法提取前景区域,利用混合高斯模型进行背景建模,建模过程中,依据规则不断完成背景自适应提取与更新,排除噪声及"假目标"信息的干扰。在检测出目标车辆后,为保证跟踪效果,利用目标特征参数及运动状态的一致性、连续性排除噪声干扰。通过对目标车辆建立Kalman滤波预测模型,实现对目标的稳定跟踪。实验结果表明,该方法具有较好的实时性和跟踪效果,能够满足实时监控的要求。In recent years,with the increase of motor vehicles,major＂Du City＂start to appear. The variety of traffic problems are increasing,thus making the rapid development of intelligent transport systems is imminent. Based on research of traditional tracking methods for vehicles,a tracking vehicles algorithm is proposed based on Gaussian model and Kalman filter. Through in- depth analysis of complex issues on external environment and self- conversion,the foreground is retrieved by using the background subtraction method. The mixture Gaussian model is adopted to model the adaptive background subtraction,and real- time updating is done to eliminate the interference of noise and fake target. In viewof the target properties,in order to ensure tracking effect,through the establishment of the Kalman filtering prediction model for target vehicle,the stable tracking of targets is carried out through eliminating noise disturbance by using the uniformity and continuity of characteristics of target parameters,and get the accurate traffic statistics. Experiments showthat the method has good real- time and tracking performance and meet the needs for real- time monitoring.

关 键 词：混合高斯模型 KALMAN滤波 边缘特征 车辆跟踪 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]