极地研究

极地研究 ›› 2020, Vol. 32 ›› Issue (1): 112-120.DOI: 10.13679/j.jdyj.20190011

• 研究论文 • 上一篇    下一篇

基于TDD的科考船航迹分段方法研究

王建1,史景聪1,黄冬梅1,郑小罗1,何盛琪1,张北辰2   

  1. 1.上海海洋大学, 上海 201306;
    2.中国极地研究中心, 上海 200136
  • 收稿日期:2019-02-28 修回日期:2019-04-16 出版日期:2020-03-30 发布日期:2020-03-30
  • 通讯作者: 黄冬梅
  • 基金资助:
    极地环境观测/探测技术与装备研发(2016YFC1400300)、科技部专项(2016YFC1403200)、海洋大数据分析预报技术研发(2016YFC1401902)资助

Track segmentation method of polar research vessel based on time-domain difference

  1. 1.Shanghai Ocean University, Shanghai 201306, China;
    2.Polar Research Institute of China, Shanghai 200136, China)
  • Received:2019-02-28 Revised:2019-04-16 Online:2020-03-30 Published:2020-03-30

PDF (PC)

669

摘要： “雪龙”号极地科考船是推动我国极地科学考察事业发展的重要工具, “雪龙”号在数十次的极地科考过程中累积了大量的航迹数据, 其中蕴含的巨大价值亟须挖掘。针对科考船的航迹分段是将科考船移动轨迹分为停留与行驶两部分, 合理的分段方法可以分离出信息更丰富的航迹段, 有利于航迹知识提取。然而, 由于原始航迹信息密度分布不均等原因, 现有的航迹分段方法往往会造成分段过多等问题, 结果并不理想。本文针对该问题, 提出了一种针对科考航迹整体的时域差分(Time Domain Difference, TDD)分段方法。本方法基于时间域对航速进行差分处理, 有效降低了因为航速波动频繁对分段结果的影响。同时, 考虑到该方法的差分步长在航迹处理过程中的不明确性, 本文将差分后航迹的路程损失和航速波动幅值进行归一化处理, 提出了航迹差分时间步长的动态确定方法, 并以速率阈值对航迹进行分段。最后本文以第29次南极科考航迹数据为例, 将本方法与经典的具有噪声的基于密度的聚类方法DBSCAN(Density-Based Spatial Clustering of Applications with Noise)进行了比较, 实验结果表明本文提出的方法可有效降低航迹分段时分段过多的问题, 在分段准确性和时间效率等方面结果更优。

关键词: 科考航迹, 时域差分, 停留, 步长, DBSCAN

Abstract: The Xuelong research vessel is an important tool in the development of China's polar scientific research. A large quantity of ship track data has accumulated over dozens of voyages; they contain valuable information that needs to be mined. Track segmentation categorizes ship tracks according to the state of the ship’s propulsion system: at rest or engaged. Efficient segmentation algorithms can extract the segments containing large amounts of information. However, because of uneven distribution of track information, algorithms often cause problems such as excessive segmentation, and segmentation results are less than ideal. This paper proposes a segmentation method based on time-domain difference for the entire track. The method performs differential processing on speed in the time domain; because of frequent fluctuations in speed, influence of the segmentation result is effectively reduced. It normalizes path loss and speed fluctuation of the differential track, time step of track difference is determined dynamically, and finally, the track is divided by the rate threshold. Track of the 29th Antarctic expedition was segmented using the time-domain difference method and the classic Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm. Results from the time-domain difference method can effectively reduce excessive track segmentation, and are better than those obtained from DBSCAN in accuracy and time efficiency of track segmentation.

Key words: research vessel track, time-domain difference, stay point, step size, DBSCAN