基于多向梯度法的红外弱小目标快速检测方法  

Infrared small dim target detection based on multi-direction gradient

作　　者：王军[1,2] 姜志[3] 柳红岩 何昕[2] 张新[2] 

机构地区：[1]苏州科技大学电子与信息工程学院,江苏苏州215000 [2]中国科学院长春光学精密机械与物理研究所,吉林长春130033 [3]中国白城兵器试验中心,吉林白城137000

出　　处：《光电子．激光》2016年第9期957-962,共6页Journal of Optoelectronics·Laser

基　　金：国家自然科学基金(11043001)资助项目

摘　　要：针对复杂背景下红外弱小目标信噪比(SNR)低、对比度小造成的红外目标检测率低和实时性差的问题,提出一种基于多向梯度法的红外弱小目标快速提取算法。目标提取前,利用多尺度拉普拉斯-高斯(LoG)算子抑制图像背景,凸显背景边缘轮廓与弱小目标;然后引入多向梯度目标搜索算法,选取最佳梯度数,利用最简算法快速搜索目标。实验结果表明,本文算法处理后的红外图像有较高的SNR与对比度,检测率为传统红外目标提取算法的1.5倍,充分保证了检测精度,且计算耗时短,实时性强。Because of the low signal-to-noise ratio,the detection rate and real-time performance of infrared small dim targets are influenced seriously.In this paper,a new small dim infrared target detection m ethod is proposed to solve the problem presented above.The background of an infrared image is process ed by multi-scale Laplacian of Gaussian （LoG） algorithm,which can highlight infrared targets and t he edge of background.The LoG operator with anisotropy is proposed,which is sens itive to multi-direction and can intensify the edges with the that have same direction as the major axis. After LoG algorithm processing,the infrared targets are searched by using multi-direction gradient algorithm,choosing optimum number of gradient and searching the infrared targets using the simplest algorithm.Finally,the morphology-based algorithm was used to fill the inner space of target for const ructing the entire area of target.The results of the experiment show that the signal-to-noise r atio and contrast ratio are enhanced after processing,and the detect ion rate of the infrared targets is 1.5times larger than that with the traditional detection method.The proposed algorithm maintains the detection accuracy and reduces time-consuming. The superiority of real-time performance validates the algorithm.

关 键 词：红外目标检测 图像背景抑制 LOG算子 多向梯度目标搜索 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]