基于双核判决的红外小目标检测方法

An Infrared Small Target Detecting Method Based on Two-Core Judging

复杂背景下红外小目标检测与跟踪是红外目标探测系统的核心技术之一。为了准确检测与跟踪空中多个小目标,针对复杂背景下单帧检测虚警率偏高,导致帧间检测计算量大的问题,提出了一种基于双核判决的小目标检测算法。首先根据目标、背景及噪声的不同灰度特性,利用Robinson滤波器结构,设定背景区域和小目标区域,更为真实的表现背景与小目标的区域分布;然后计算外核目标与背景在四个方向上的对比度及内核目标区域内的相似度,若双核度量同时达到设定阈值条件,则判定为真实目标点,这样消除背景边缘及噪声的影响,减少目标检测虚警。实测数据表明,与传统的高通滤波法、Top-hat算法相比,该算法能在单帧检测上更有效检测出小目标,抑制背景边缘和噪声点,提高抗噪能力,降低虚警率,增强帧间目标的检测识别率,能大大减少序列图像间目标检测与跟踪的计算量。

It＇s one of the core technologies to detect and track small target under complex background in infrared target detection system. In order to track multi-targets accurately and solve the high false alarm rate and high computation problem, a method for detecting infrared small targets is proposed based on Two-core judging. Firstly, according to the difference among background, noise and targets, we design background and target area based on Robinson filter＇s structure. Then compute the comparison in four directions and the comparability. If both of them reach the condition, it＇s a real target. The experimental result shows that comparing with traditional high-pass filter and Top-hat filter, this method can detect small targets more efficiently, improve the ability of resisting noise and decrease false alarm rate, which lay a firm foundation for the small targets detection in image frames.