结合统计分布和非下采样Contourlet变换的红外小目标检测

Infrared small target detection based on nonsubsampled Contourlet transform and statistical distribution

在分析存在背景和噪声干扰下的红外小目标检测方法的基础上,提出了一种时空结合的红外小目标检测算法。首先,基于相邻帧背景图像灰度值变化的特点,提出了基于三阶中心矩统计分布确定小目标区域方法,接着对图像进行非下采样Contourlet变换(NSCT),定义子带的能量系数,并结合各个子带的能量系数得到各点的能量值,进而得到一个基于各点能量值的图像。最后,根据能量图像中小目标、背景及噪声的不同分布特点,利用阈值分割得到要检测的小目标。仿真实验结果表明,该方法能较精确地检测出红外小目标,具有较高的检测率和较小的平均虚警数;对单一背景和运动背景下的红外小目标进行检测,其检测率分别达到了98%和97%,平均虚警数仅为0.05和0.17,并且在小目标出现快速以及不规则运动时仍能进行有效检测。

The method to detect small moving targets in infrared image sequences that contain moving nuisance objects and background noises is analyzed in this paper.An infrared small target detection algorithm combined with temporal and spatial domains is put forward.On the basis of the background change slowly,the algorithm firstly confirms the object region based on the third central moments of frame difference,and decomposes the frame difference image by nonsubsampled Contourlet transform to define the energy coefficients of the sub-band images.Then the image based on the energy value of each pixel is obtained.Finally,the final detecting of the target is realised according to the different features of small targets,backgrounds and noises.The results indicate that the small infrared target detection based on nonsubsampled Contourlet transform can precisely detect the small infrared target and has better target detection performance.When the small targets with invariable and moving backgrounds are detected,the detection rate of the proposed algorithm can reach 98% and 97% and the mean of false alarm points is only 0.05 and 0.17,respectively.It concludes that the proposed algorithm can detect the small target when the target has fast or anomalistic movement.