基于噪声方差估计的红外弱小目标检测与跟踪方法

针对红外弱小多目标图像背景杂波干扰严重、弱 小目标检测率低和目标跟踪困难的问题,提出一种 基于噪声方差估计的红外弱小目标快速检测与目标跟踪算法。首先采用改进的形态学 滤波抑制背景噪声, 对处理后的多帧图像进行方差估计初步突出目标像素；然后对其进行信噪比(SNR)估计得到整个图像序列像素得 分,图像中像素SNR高的被标记为目标像素；再对标记过的图像进行分块分析 ,准确提取出连续图 像序列中的目标像素；将检测出的目标像素作为Hough变换的目标跟踪算法的输入,设置双 阈值实现目标 的有效跟踪。实验结果表明,在复杂背景下的红外弱小目标提取中,基于噪声方差估计的目 标检测拥有较 高的检测概率和较低的虚警概率,将其获得的目标像素作为Hough变换的输入,不仅可以有 效跟踪目标, 而且简化了算法的复杂度,实现目标的快速提取和跟踪,具有很高的应用价值。

Detection and tracking of weak infrared targets based on noise variance estimat ion

An algorithm on fast detection and tracking of weak infrared targets from comple x background is proposed based on noise variance estimation.Firstly,the backgro und noise is suppressed by the improved morphological filtering,and the target pixel is hig hlighted preliminarily by the variance estimation of the processed multi-frame image,th en the signal-to-noise natio (SNR) is estimated to get the whole image sequence pixel score.The pixels with high scor es are marked as the target pixels,and the marked image is divided and analyzed.Finally,th e target pixels in the continuous image sequence are extracted accurately.The target pixels are de tected as the input of the target tracking algorithm of the Hough transform,and the double-th reshold is set to achieve the effective tracking of the target.The experimental results show t hat the target detection based on the noise variance estimation has a high detection probabilit y and a low false alarm probability in the infrared small target extraction under complex backgrou nd,and the target pixels obtained as the input of the Hough transform can not only effectiv ely track the target,but also simplify the complexity of the algorithm to achieve rapid extra ction and tracking of targets.