一种基于动态规划的点目标轨迹关联算法

Track-Association Algorithm of Point Target Based on Dynamic Programming

根据低信噪比条件下图像序列中运动小目标检测涉及信息量大、目标轨迹复合复杂、实时处理困难等特点,在对现有的检测算法研究分析的基础上,详细阐述和研究了一种基于动态规划能量累加的二级检测算法,此算法根据点目标运动的连续性和相关性,轨迹关联时利用最小二乘的方法预测目标位置,在预测位置对搜索窗口采取方向性限制,实验效果指出此算法可有效的检测出低信噪比条件下图像序列中的慢速运动点目标。该算法也可检测交叉运动的多个运动点目标。

According to these characters such as large information, complicate resuming of target trajectories and difficult real-time disposal which occur in detecting moving targets in IR image sequences, a two-steps detection algorithm is presented with a thoroughly analysis, which is track associating after dynamic programming energy accumulating. According to continuity and relativity of movement, target position is predicted by Least-Square algorithm, and meanwhile, searching window will be directionally limited during track-association. Based on the performance of simulation test,the algorithm can detect low SNR moving point target in image sequences effectively.Meanwhile,it can be used to detect multiple point targets which move cross each other, too.