基于光流估计摄像机运动下红外小目标检测算法  

Infrared Small Target Detection Algorithms with a Moving Camera Based on Optical Flow Estimation

作　　者：单巍[1] 陈得宝[1] SHAN Wei;CHEN Debao(School of Physics and Electronic Information,Huaibei Normal University,235000,Huaibei,Anhui,China)

机构地区：[1]淮北师范大学物理与电子信息学院

出　　处：《淮北师范大学学报（自然科学版）》2019年第4期25-30,共6页Journal of Huaibei Normal University：Natural Sciences

基　　金：2018年国家级大学生创新创业训练计划项目(201810373007);安徽省高等学校自然科学研究项目(KJ2018B10);2017年安徽省大学生创新创业训练计划项目(201710373020);淮北师范大学质量工程项目(jy2016132,2017tykc159)

摘　　要：针对单一光流方法不能实现摄像机运动时红外小目标实时检测的问题,提出一种基于光流估计的摄像机运动下的小目标检测方法.该方法从动态背景静态化的角度出发,计算所提取角点处的Lucas-Kanade光流,在此基础上获取动态背景的运动补偿,进而得到当前的静态背景.利用目标与背景的运动差异性,计算相对于静态背景的Horn-Schunck光流,所计算的光流值则给出最终的检测结果.实验结果表明,该方法对摄像机运动下的地空背景红外小目标序列图像有良好的检测效果.This paper proposes a detecting algorithm for moving infrared small target with a moving camera based on optical flow estimation.Starting from the point of view of changing dynamic background into static background,we extract the Harris corners in the image and calculate the Lucas-Kanade optical flow on them,and obtain compensation of background motion on this basis to get the present static background.The Horn-Schunck optical flow relative to the static background are calculated by the diffidence between target motion and background motion.The detection results are given by the optical flow calculated.The experiments show that this algorithm can detect the small target under the background of sky and ground in infrared image sequence.

关 键 词：光流 光流柱状图 HARRIS角点 红外小目标 动态背景 目标检测 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]