序列帧间双重匹配的红外点目标跟踪算法  

Algorithm of dual matching between image sequences used for infrared point target tracking

作　　者：王乐东[1] 王江安[1] 吴荣华[1] 

机构地区：[1]海军工程大学光电对抗与光通信研究所,湖北武汉430033

出　　处：《光电子．激光》2010年第3期465-469,共5页Journal of Optoelectronics·Laser

基　　金：总装"十一五"预研项目资助(4010605020402)

摘　　要：针对海空复杂背景下红外点目标的检测与跟踪,提出了基于图像序列帧间双重匹配的边跟踪边检测算法,并建立了数学模型。它采用标记序列帧M以帧对帧的方式记录输入序列的帧间匹配结果,标记帧T以点对点的方式记录标记序列帧M的帧间匹配结果,统计帧S记录T中各像素的匹配成功次数,输入单帧图像同步输出矩阵T和S分别显示目标运动轨迹和迎头目标检测结果。算法匹配过程不随目标数目或运动状态而改变,且无需提前判断疑似目标位置,有效解决了目标在图像序列中突然丢失或出现被干扰情况下的跟踪,尤其可以对跟踪结果实时地进行目标分离,解决了迎头目标跟踪的难题。仿真和实际工程图像实验结果表明,算法具有较高的可靠性和实时性。In order to realize the infrared point targets detection and tracking under complicated sea-sky background,an algorithm that can detect and track targets synchronously is proposed based on a double matching mathematical mode.The model adopts a mark image sequences M to memorize the input image sequences match result,a mark image T to memorize the mark image sequences match result,a statistical image S to memorize the successful match times of each pixel in the image T,and matrixes T and Sto show moving target track and head-on target detection results in real time.There is no need to suppose the pixel position of possible targets before match process in this algorithm which solves the track problem when targets suddenly vanish or appear in some image sequences.The whole track process doesn′t change with the background complex degree and the number of point targets.The simulation and experiment show that this method is feasibility and validity.

关 键 词：红外图像序列 目标检测 目标跟踪 图像匹配 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]