混沌PSO最小一乘空时预测的红外小目标检测  

Detection of Small Infrared Target Based on Spatial-Temporal Prediction by Chaotic PSO and Least Absolute Deviation

作　　者：吴一全[1,2] 尹丹艳[1] 

机构地区：[1]南京航空航天大学电子信息工程学院,南京210016 [2]南京大学计算机软件新技术国家重点实验室,南京210093

出　　处：《计算机辅助设计与图形学学报》2011年第5期909-914,共6页Journal of Computer-Aided Design & Computer Graphics

基　　金：国家自然科学基金(60872065);航空科学基金(20105152026);南京大学计算机软件新技术国家重点实验室开放基金(KFKT2010B17)

摘　　要：针对红外图像中背景与小目标的特点,提出一种基于混沌粒子群优化(PSO)最小一乘空时背景预测的红外小目标检测方法.首先建立最小一乘准则空时背景预测模型,根据最小一乘估计的性质,提出应用混沌PSO算法解决最小一乘估计中极值的选取问题,并用该模型预测红外图像中的背景,从原始图像中减去预测图像得到残差图像;然后提出了基于混沌PSO的二维直方图斜分模糊最大熵阈值选取方法,由此分割所得残差图像即可将小目标检测出来.将文中方法与基于最小二乘背景预测的红外小目标检测方法进行了比较实验,实验结果表明,该方法具有更高的检测概率和信噪比增益,优于基于最小二乘背景预测的红外小目标检测方法.Considering the characteristics of background and small targets in infrared images,a detection method of small infrared targets is proposed,which is based on chaotic particle swarm optimization（PSO） and spatial-temporal background prediction by least absolute deviation.Firstly,a model of spatial-temporal background prediction is built.According to the properties of least absolute deviation,extreme values in the least absolute deviation are selected by chaotic PSO.The background in the infrared image is predicted and the predicted background image is subtracted from the source image to give a residual image.Then,a two-dimensional histogram oblique segmentation method based on chaotic PSO and fuzzy maximum entropy is presented.The small target is detected by thresholding the obtained residual image.The experimental results were compared with the results of small infrared target detection method based on background predication by least squares.The experimental results show that the proposed method has higher detection probability and provide better gain of signal-to-noise ratio（GSNR）.The proposed method is superior to the method of small infrared target detection based on background predication by least squares.

关 键 词：红外小目标检测 空时背景预测 最小一乘估计 混沌粒子群优化 模糊最大熵阈值分割 二维直方图斜分 

分 类 号：TN911.73[电子电信—通信与信息系统]