针对无人船载双雷达目标跟踪问题，提出一种基于全局最近邻关联（GNN）的无人船雷达目标跟踪算法。建立随船正北坐标系，对传感器的测量结果进行坐标变换，利用GNN算法对目标进行跟踪。针对GNN算法在虚警、漏报情况下易发生误关联的缺点，提出一种跟踪门的设置方法，利用关联代价与模糊隶属度对目标进行2次滤除来提高GNN的关联正确率。最后，利用安装有激光雷达与毫米波雷达的无人船在实际海上环境进行试验，试验结果表明，与未改进的跟踪算法相比，本文提出的改进跟踪算法正确率提高了25.87%，验证了算法的可行性和有效性。
Aiming at the problem of Unmanned Surface Vehicle(USV) dual radar target tracking, an algorithm of USV radar target tracking based on Global Nearest Neighbor Association (GNN) is proposed. According to the coordinate system of the USV and the angle of the bow of the ship, coordinate the measurement results of the sensor are transformed; the GNN algorithm is used to track the target. Aiming at the shortcomings of the GNN algorithm that false associations easily occur in the case of false alarms and false alarms, a method for setting association gates is proposed, which uses association cost and fuzzy membership to filter the target twice to improve the accuracy of GNN association. Finally, an USV equipped with lidar and millimeter-wave radar was used to conduct experiments in the actual marine environment. The experimental results show that: compared with the unimproved tracking algorithm, the correct rate of the improved tracking algorithm proposed in this paper is increased by 20%, which proves The feasibility and effectiveness of the algorithm.
2022,44(22): 58-62 收稿日期：2021-04-14
DOI：10.3404/j.issn.1672-7649.2022.22.011
分类号：TN953+.6
基金项目：国家重点研发计划(2017YFC1405***)；中央高校基本科研业务费专项资助(19CX05003A-1)
作者简介：王宁(1997-),男,硕士研究生,研究方向为多传感器数据融合
参考文献：
[1] 李小毛, 张鑫, 王文涛, 等. 基于3D激光雷达的无人水面艇海上目标检测[J]. 上海大学学报(自然科学版), 2017, 23(1): 27–36
[2] 李瑞伟, 李立刚, 金久才, 等. 基于欧氏距离的无人艇载毫米波雷达点迹凝聚方法[J]. 水下无人系统学报, 2020, 28(6): 604–610
[3] 胡朋启, 李蔚清. 异类传感器数据融合的最近邻-点拓扑关联法[J]. 计算机与数字工程, 2019, 47(6): 1347–1350+1465
[4] SHI D U, REN J R, GUO J. Design and research on tracking accuracy of 4D trajectory of aircraft[J]. Computer Simulation, 2019.
[5] HE S, HS SHIN, TSOURDOS A. Distributed multiple model joint probabilistic data association with gibbs sampling-aided implementation[J]. Information Fusion, 2020: 64
[6] 马娟, 许厚棣, 张瑞国. 一种复杂环境下的多假设分支跟踪方法[J]. 雷达科学与技术, 2020, 18(4): 399–403+416
[7] 王鹏宇, 赵世杰, 马天飞, 等. 基于联合概率数据关联的车用多传感器目标跟踪融合算法[J]. 吉林大学学报(工学版), 2019, 49(5): 1420–1427
[8] 贺丰收, 缪礼锋. 一种交互式多模型全局最近邻关联算法[J]. 火力与指挥控制, 2013(10): 30–33
[9] AZIZ A M. A new nearest-neighbor association approach based on fuzzy clustering[J]. Aerospace ence& Technology, 2013, 26(1): 87–97
[10] 戴永寿, 马鹏, 孙伟峰, 等. 基于JVC的紧凑型地波雷达海上目标点迹-航迹最优关联方法[J]. 电子与信息学报, 2021, 43: 1–8
[11] 田威, 黄高明, 彭华甫. 多传感器数据融合中错误关联的生成机理及影响分析[J]. 指挥与控制学报, 2018, 4(2): 160–164
[12] 韩崇昭, 朱洪艳, 段战胜. 多源信息融合[M]. 北京: 清华大学出版社, 2010.
[13] HAN J, KIM J, SON N S. Persistent automatic tracking of multiple surface vessels by fusing radar and lidar[C]//Oceans. IEEE, 2017: 1−5.
