基于多关联模板匹配的火箭目标跟踪算法  

Rocket Target Tracking Algorithm Based on Multiple Related Template Matching

作　　者：尹宏鹏[1] 柴毅[1] 屈剑锋[1] 翟茹玲[1] 

机构地区：[1]重庆大学自动化学院,重庆400044

出　　处：《计算机工程》2009年第18期31-33,36,共4页Computer Engineering

基　　金：国家"863"计划基金资助项目"GIS支持下的卫星发射场及火箭发射飞行安全控制应急保障系统"(2003AA132050)

摘　　要：针对火箭目标在复杂场景的跟踪问题,提出一种基于多关联模板匹配的模板匹配策略,通过仿射变换从上帧最优模板中产生多关联模板以自适应火箭目标的大小和姿态改变,提高算法的匹配精度。采用Kalman滤波器对火箭飞行进行轨迹预测。多个火箭飞行录像跟踪实验表明,该算法具有很好的匹配精度与实时性,对火箭目标的状态、大小变化与云层遮挡现象具有较好的鲁棒性。Aiming at the tracking problem of the rocket target in complex scene, this paper proposes a template matching strategy based on multiple related template matching. This strategy can improve the track accuracy by getting the best template from the previous frame using the mapping transform, which can fit rocket＇s size and state automatically. The Kalman filter is used to predict the track of the rocket. Several tracking experiments show the approach is accurate and fast, and it has a better robust pertormance during the state changing, the size changing and the shelter instance.

关 键 词：火箭目标跟踪 多关联模板 模板匹配 KALMAN滤波器 轨迹预测 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]