红外搜索跟踪系统的虚警率降低方法  

False Alarm Rate Reduction in Infrared Search and Track Systems

作　　者：宋治杭 张晋[1] 朱亮 林丹丹[1] 杜欣悦 林宇[1] SONG Zhihang;ZHANG Jin;ZHU Liang;LIN Dandan;DU Xinyue;LIN Yu(Kunming Institute of Physics,Kunming 650223,China)

机构地区：[1]昆明物理研究所,云南昆明650223

出　　处：《红外技术》2022年第4期371-376,共6页Infrared Technology

摘　　要：红外全景搜索跟踪系统能对360°大视场提供实时多目标搜索与定位功能,与此同时大视场的引入也带来虚假目标增多,以及同一运动目标在前后生成的两幅全景图像中位置差异过大的困扰。本文针对以上问题,引入图像金字塔变换降低全景图像分辨率,再利用Lucas-Kanade(LK)光流法计算所有疑似运动区域的光流,最后分析所有光流特征结合二分类思想将真实运动目标和虚假运动目标区分开,针对全景图像能够有效的将虚假目标检测出来,剔除该类目标对后续处理结果的干扰。Infrared panoramic search and track systems can provide real-time multi-target search and positioning functions for a 360°field of view.However,the involvement of the large view field introduces problems such as increased number of false targets and a large displacement of moving targets between panoramic frames.To address these problems,this study uses the image pyramid transform to reduce the panoramic image resolution and the Lucas-Kanade optical-flow method to calculate the optical flow of all predicted movement areas.Finally,the optical-flow features are analyzed to separate the real target from the false target area.In this way,false targets can be effectively detected in the panorama images,and the interference of such targets in the subsequent processing can be eliminated.

关 键 词：全景搜索跟踪 虚警率 Lucas-Kanade光流 二分类 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]