面向UUV回收过程的单目视觉导引灯阵跟踪方法  

Monocular Vision Guided Lamp Array Tracking Method for the UUV Recovery Process

作　　者：张伟[1] 潘珺 宫鹏 李子轩 郭帅克 高赛博 ZHANG Wei;PAN Jun;GONG Peng;LI Zi-xuan;GUO Shuai-ke;GAO Sai-bo(College of Automation,Harbin Engineering University,Harbin 150001,China)

机构地区：[1]哈尔滨工程大学智能科学与工程学院,黑龙江哈尔滨150001

出　　处：《水下无人系统学报》2021年第4期435-441,共7页Journal of Unmanned Undersea Systems

基　　金：国家自然科学基金(5207110590).

摘　　要：无人水下航行器(UUV)在进行水下作业过程中,受到自身可携带载荷和环境的制约,导致其作业时间不能过长和作业距离不能过远,需要在一定时间内回收以供给能量、获得数据并布置新任务。深入研究UUV视觉跟踪问题,对实现UUV快速、准确、安全的回收有着重要意义。基于此,文中对UUV回收的单目视觉多目标跟踪问题进行了研究。以非线性卡尔曼滤波器为基础,提出基于非对称光阵列的UUV回收跟踪方法。首先,对单目摄像机采集的水下图像进行预处理,提取光源阵列的特征信息。其次,初始化目标轨迹,提出了基于非对称L型光源阵列的轨迹确定方法。再次,权衡观测器与滤波器设计了寻优代价矩阵,应用改进的匈牙利算法将检测与目标轨迹相匹配,并根据匹配结果对卡尔曼滤波器的先验结果进行修正。最后,通过水下目标跟踪实验,验证了所提方法的工程有效性。An unmanned undersea vehicle(UUV)is restricted by its carrying load and environment during underwater operation,reducing the maximum operation time and distance.As a result,periodical UUV recovery is necessary to supply it with energy,obtain data,and arrange new tasks.Further study on UUV vision tracking is important to realize a rapid,accurate,and safe UUV recovery.In this study,problems caused by the monocular visual multi-target tracking for UUV recovery are investigated.A recovery tracking method of UUVs based on the nonlinear Kalman filter and asymmetric light array is proposed.First,underwater images collected by the monocular camera are preprocessed to extract characteristic information of the light source array.Afterward,the target trajectory is determined through a method based on the asymmetric L-shaped light source array.The optimal cost matrix is thereafter established by weighing the observer and the filter,and the detection is matched to the target trajectory using the improved Hungarian algorithm.Prior results of the Kalman filter are modified according to these results.Finally,the engineering effectiveness of the proposed method is verified through underwater target tracking experiments.

关 键 词：无人水下航行器 回收 多目标跟踪 单目视觉 非线性卡尔曼滤波器 

分 类 号：TJ630.33[兵器科学与技术—武器系统与运用工程] TP242.62[自动化与计算机技术—检测技术与自动化装置]