红外图像序列中微弱点状动目标的非参数检测

Nonparametric Detection of Dim Moving Point Targets in IR Image Sequences

研究了红外图像序列的观测及目标模型,提出了一种基于多帧的微弱点状动目标非参数检测技术。采用滑动邻域加权预测抑制杂波干扰,白化噪声特性,将时空三维搜索简化为沿时间轴投影与二维空域搜索,沿可能的轨迹上将目标能量高维集成。相应算法的检测性能对噪声分布不敏感,实时性好,实现简单,具有很高的实用价值。给出了该算法性能的理论分析和实验仿真结果。

A kind of nonparametric detection of dim moving point targets in infrared (IR) image sequence data is presented. The models for observations and targets are also described. The moving weighted local average removal method is used for clutter suppression. Integrating the target energy over multiple frames along possible tracks in 2D space produced by projecting the image sequence along the time axis. This technique can be charac-terized by its insensitivity to the changes of noise's statistical distribution,suitability for real-time processing and extreme simplicity in implementation that it is very useful in many practical applications. Theoretical analysis of the algorithm's performance and experimental results are also given.