一种改进的红外弱小目标快速检测方法  

An improved detection method for infrared dim and small target

作　　者：孙慧婷[1] 姜志[2] 王军[1,3] 张新[3] 何昕[3] 

机构地区：[1]苏州科技大学虚拟现实智能交互及应用技术重点实验室,江苏苏州215009 [2]中国白城兵器试验中心,吉林白城137000 [3]中国科学院长春光学精密机械与物理研究所,吉林长春130033

出　　处：《激光与红外》2017年第10期1310-1315,共6页Laser & Infrared

基　　金：江苏省建设系统科技项目(No.2016ZD87);苏州市科技发展计划(重点实验室)项目(No.SZS201609);江苏省普通高校研究生科研创新计划项目(No.KYLX15_1311);苏州科技大学研究生创新项目(No.SKCX15_050)资助

摘　　要：针对复杂背景下红外弱小目标检测率低、目标跟踪困难的问题,提出一种改进的红外弱小目标快速检测方法。该方法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估计得到整个图像序列像素得分,图像中像素信噪比高的被标记为目标像素,再对标记过的图像进行分块分析,最终准确提取出连续图像序列中的目标像素。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。As infrared dim and small target is difficult to detect and track under complicated background,an improved fast detection method for dim and small infrared targets is proposed. Firstly,the background noise is suppressed by the improved morphological filtering,and the target pixel is highlighted preliminarily by the variance estimation of the processed multi-frame image,then the SNR is estimated to get the whole image sequence pixel score. The pixels with high scores are marked as the target pixels,and the marked image is divided and analyzed. Finally,the target pixels in the continuous image sequence is extracted accurately. The target pixels are treated as the input of the target tracking algorithm of the Hough transform,and the double thresholds are set to achieve the effective tracking of the target. The experimental results show that the target detection based on the noise variance estimation has a high detection probability and a low false alarm rate in the infrared dim and small target extraction under complex background,and the target pixels obtained as the input of the Hough transform can effectively track the target and simplify the complexity of the algorithm to achieve rapid extraction and tracking of targets,and it has a high application value.

关 键 词：方差估计 形态学滤波 阈值 目标检测 

分 类 号：TP394.1[自动化与计算机技术—计算机应用技术]