核相关滤波跟踪算法在气体报警器自动化检定中的应用  

The Application of Kernelized Correlation Filter Tracking Algorithm in Automatic Detection System of Gas Monitors

作　　者：蒋晓慧 曹红方 赵振刚[2] 李翔 查小慧 王岩岩 JIANG Xiaohui;CAO Hongfang;ZHAO Zhengang;LI Xiang;CHA Xiaohui;WANG Yanyan

机构地区：[1]苏州市计量测试院 [2]中国科学技术大学软件学院 [3]苏州城市学院

出　　处：《计量与测试技术》2022年第4期83-85,90,共4页Metrology & Measurement Technique

摘　　要：由于被检设备外观和显示样式的非结构化等原因,现有气体报警器自动化检定系统大都是借助人工对图像待识别的区域进行选定操作。而在手动框选目标识别区域后,由于人为调整摄像头或被检报警器,往往会造成摄像头与被检报警器的相对位置发生偏移,以及已确定的检测框与实际显示位置出现偏差,导致检测结果出现问题。针对这个问题,本文利用核相关滤波(Kernelized Correlation Filter,KCF)目标跟踪方法实现候选区域的自动跟踪。实验结果表明,该方法可以自适应小幅度位置调整,具有较高的准确度和稳定性。Un-identified area selecting for the images in the automatic detection system of gas monitors was usually implemented by manual operations,because of the non-structured appearance and display style for the detection devices.Relative position shifting between the camera and detected monitor often occurred after manually selecting the target recognition area,resulting in the position deviation for the determined detection frames.Kernelized Correlation Filter(KCF)tracking algorithm was employed here to achieve automatic tracking for the candidate regions.The experimental results show that our algorithm has superior ability in self-adapting position adjustment in small amplitude,as well as high accuracy and stability.

关 键 词：气体报警器 KCF 目标跟踪 手动框选 

分 类 号：TP29[自动化与计算机技术—检测技术与自动化装置]