摘要
在无人水下航行器(UUV)自动接驳技术中,近距离接驳阶段操作的准确性和稳定性是有待攻克的技术难题。文中以叉柱式UUV接驳方式为背景,提出了一种基于单目视觉导引灯阵定位方法。首先,针对UUV接驳的机构特性及视觉对接要求,设计了一套简易、可行性高的UUV视觉接驳装置;其次,对大津法进行改进,解决了在亮度整体较暗、低对比度、细节较模糊条件下,导引灯阵目标的特征提取问题;再次,利用二进制大对象(BLOB)分析描述特征结合改进多尺度核相关滤波器(KCF)跟踪算法,解决了视觉跟踪过程中出现伪光源而导致跟踪失败的情况;最后,结合超短基线(USBL)与人工标定数据进行水池实验,验证了在近距离定位过程中,该方法相比USBL定位更加有效与稳定,误差补偿后,定位偏差<0.1m。
A novel monocular visual guided light array positioning method of unmanned undersea vehicle(UUV) dock- ing system is proposed for the fold-carrying-pole recovery. First, according to the mechanism characteristics of UUV docking and the visual docking requirements, a simple and feasible UUV visual docking device is designed. Second, the Otsu method is improved to extract target features of the guide light array under the conditions of darkness, low contrast, and blurry details. Third, using the binary large objects(BLOB) analysis to describe features and adopting the improved multi-scale kernel correlation filter(KCF) tracking algorithm, the tracking fail due to appearance of pseudo light source during the tracking process is avoided. Pool test combined with ultra-short baseline(USBL) and manual calibration data verifies that this positioning method is more effective and stable than the USBL method for short-distance positioning, and the error is within 0.1 m after error compensation.
作者
张伟
李子轩
宫鹏
赵汉卿
曾宁
ZHANG Wei;LI Zi-xuan;GONG Peng;ZHAO Han-qing;ZENG ning(College of Automation,Harbin Engineering University,Harbin 150001,China)
出处
《水下无人系统学报》
北大核心
2019年第3期313-318,共6页
Journal of Unmanned Undersea Systems
基金
国家自然科学基金资助(HEUCFP201736、HEUCFM171011、E091002/51309067)
黑龙江省科学基金项目(E2016020)
关键词
无人水下航行器
自动接驳
视觉跟踪
特征提取
unmanned undersea vehicle(UUV)
automatic docking
visual tracking
feature extraction
