一种适用于多场景的红外弱小目标检测跟踪算法  

An Infrared Small Target Detection and Tracking Algorithm Applying for Multiple Scenarios

作　　者：施天俊 鲍广震 王福海 兰超飞 巩晋南[1] Shi Tianjun;Bao Guangzhen;Wang Fuhai;Lan Chaofei;Gong Jinnan(Research Center for Space Optical Engineering,Harbin Institute of Technology,Harbin 150001,China)

机构地区：[1]哈尔滨工业大学空间光学工程研究中心

出　　处：《航空兵器》2019年第6期35-42,共8页Aero Weaponry

摘　　要：红外弱小目标检测一直是军事领域研究的热门问题。本文提出了一种基于图像复杂度和显著性帧间差分的更为精确且高效的红外弱小目标检测跟踪算法。该算法整体框架包含预处理模块、检测模块、跟踪模块三部分。通过预处理对红外图像中的无效像元进行补偿;对不同红外场景进行判断,选用N-P准则直接检测或基于显著性帧间差分法检测目标;实时探测成像范围内背景的移动量、对已检出的目标运动轨迹做出连续性判断,能补偿检测所无法消除的虚警及漏警。选用四组不同场景的数据集进行测试,在目标形心真值3×3范围内的检测率为98.2%,检测失败率(检测出目标形心在其值9×9外)为0.01%,表明该算法能快速实现对各类背景下的目标跟踪检测,并有较高的适应性、准确率。Infrared small target detection has always been a hot topic in the military field. In this paper, a more accurate and efficient infrared small target detection and tracking algorithm based on image complexity and saliency frame-to-frame differences is proposed. The whole framework includes three parts: preprocessing module, detection module and tracking module. The invalid pixels in infrared images are compensated by preprocessing. After distinguishing the different scenes, N-P criterion or saliency frame-to-frame differences method can be chosen for detection. Real-time detection of the amount of background movement within the imaging range, and continuous judgment of the movement track can compensate for the false alarms and missed alarms that cannot be eliminated by detection. Four data sets of different scenes are selected for testing. The detection rate is 98.2% within the range of 3×3 area centered by the centroid truth value, and the detection failure rate is 0.01%(the range is over 9×9). The algorithm can quickly realize target tracking and detection under various backgrounds and has high adaptability and accuracy.

关 键 词：红外弱小目标 显著性 N-P准则 帧间差分 波门跟踪 

分 类 号：TJ760[兵器科学与技术—武器系统与运用工程] TN219[电子电信—物理电子学]