基于显著性与帧间差分的红外弱小目标检测

Infrared small target detection based on saliency and frame-to-frame difference

针对复杂背景下弱小目标检测问题,提出一种基于图像显著性和帧间差分相结合的红外弱小目标轨迹检测算法。该算法首先利用高效的频域残差法检测目标可能的存在区域,再利用帧间差分法消除背景孤立的奇异点和背景变化较强烈的部分,突出了目标信息。然后将差分图像进行累加初步得到目标的轨迹,经过累加后的图像目标的能量较杂波和噪声有了显著的提高。最后通过门限分割和轨迹关联进一步剔除虚警目标。仿真实验表明,算法在不同的背景杂波和目标尺寸下都有较好的检测性能,便于硬件实现和后续的识别与跟踪。

An algorithm of small and dim infrared target detection base on saliency and frame to frame difference theory is proposed complicated background.This algorithm firstly using spectral residual approach to detect the target may located region in the image with high performance.Then by use of frame to frame difference to eliminate the isolated point and change intensely part in the background,in order to prominent the small target.The next we will add the difference images to get the target trajectory preliminary,the power of small target in the added image can enhanced compare to clutter and noise.The last,adopt to threshod segmention and trajectory association to false alarm target.Analysis of the result shows that proposed algorithom background suppression performance is remarkably improved compared to traditional method.What＇s more,it very suitable for realize by hardware and contribute to recognition and tracking.