基于形态滤波和均值漂移红外小目标检测方法

A Method of small target detection in IR image based on morphological filter and Mean Shift algorithm

低对比度复杂背景下的小目标检测一直是研究的热点和难点,检测的困难主要在于背景噪声的复杂和目标的微弱。分析和研究了形态膨胀算法均值漂移(Mean Shift)算法:形态膨胀算法对目标进行有效增强,而均值漂移算法改善目标与背景对比度,有利于有效分割目标。最后实现了基于该方法的两种不同情景下的小目标的检测,实验表明该算法具有较好的有效性和鲁棒性。而且,该方法在最终目标选取采用了自适应阈值方法。实验分析表明:算法基本上是定点运算,效率较高,易于实时硬件实现。

Small target detection in IR images with low contrast and clutter is always the hot and difficult topic. The difficulty of detect is complexity of background noise and weakness of object.. In this paper, we analysis and study the dilation in morphological filter and the Mean Shift algorithm. The method of dilation in morphological filter enhances the object and algorithm of mean shift improves in contrast of object and background while it is benefit to segment object from background. And implement the detection two small targets in the different scene with low contrast. The experiments show its robustness and validity. The method takes the adapted threshold to segment the small target finally. The experiment analysis shows it is basically integer calculation and has more effect and easy to implement in real time with hardware.