无人船红外图像单目视觉检测与跟踪研究  

Research on monocular vision detection and tracking in infrared images of unmanned ships

作　　者：熊守丽[1] XIONG Shou-li(School of Information and Electromechanical Engineering,Yangtze University College of Arts and Sciences,Jingzhou 434020,China)

机构地区：[1]长江大学文理学院信息与机电工程学院,湖北荆州434020

出　　处：《舰船科学技术》2024年第7期159-162,共4页Ship Science and Technology

基　　金：湖北省教育厅科学研究计划指导性项目(B2018417)。

摘　　要：无人船可用于海洋探测、军事侦查、军事打击等领域,结合GPS、AIS、无线通信技术以及图像视觉技术使得无人船实现智能化控制。无人船红外图像的单目视觉检测对实现目标检测以及跟踪具有至关重要的作用。本文提出使用YOLOV7算法对红外图像目标进行识别,确定红外图像训练集,对图像进行预处理后完成模型的训练及更新,最终实现对红外图像目标的识别。将KCF算法应用于红外图像目标跟踪,研究KCF算法对目标的跟踪流程,使用KCF算法和DCF算法进行仿真分析发现,同等情况下,KCF算法的跟踪准确率为78%,优于DCF算法。Unmanned ships can be used in ocean exploration,military reconnaissance,military strikes and other fields,combined with GPS,AIS,wireless communication technology and image vision technology to achieve intelligent control of unmanned ships.The monocular vision detection of infrared images of unmanned ships plays an important role in target detection and tracking.This paper proposes to use YOLOV7 algorithm to recognize infrared image targets,determine the training set of infrared images,and complete the training and update of the model after image preprocessing,and finally realize the recognition of infrared image targets.KCF algorithm is applied to infrared image target tracking,and the tracking process of KCF algorithm is studied.Simulation analysis using KCF algorithm and DCF algorithm shows that the tracking accuracy of KCF algorithm is 78%,which is better than DCF algorithm in the same case.

关 键 词：无人船 单目视觉检测 YOLOV7 KCF算法 目标跟踪 

分 类 号：U667.65[交通运输工程—船舶及航道工程]