热带海洋学报 ›› 2013, Vol. 32 ›› Issue (5): 73-78.doi: 10.11978/j.issn.1009-5470.2013.05.010

• 海洋物理学 • 上一篇    下一篇

基于动态聚类的单声脉冲多波束测深数据滤波

陈小龙1, 庞永杰2, 李晔2   

  1. 1. 中国舰船研究设计中心, 湖北 武汉 430064; 2. 哈尔滨工程大学水下机器人技术重点实验室, 黑龙江 哈尔滨 150001
  • 收稿日期:2012-05-30 修回日期:2012-08-24 出版日期:2013-11-21 发布日期:2013-11-21
  • 作者简介:陈小龙(1982—), 男, 湖北省麻城市人, 博士, 研究方向为水下机器人控制与导航。E-mail: xiaolong1133@163.com
  • 基金资助:
    国家自然科学基金项目(50909025、51179035)

Single ping filtering of multi-beam echo sounder data based on dynamic clustering

CHEN Xiao-long1, PANG Yong-jie2, LI Ye2   

  1. 1. China Ship Development and Design Center, Wuhan 430064, China; 2. State Key Laboratory of Autonomous Underwater Vehicle, Harbin Engineering University, Harbin 150001, China
  • Received:2012-05-30 Revised:2012-08-24 Online:2013-11-21 Published:2013-11-21

摘要: 针对相干型多波束测深数据的特点, 提出了一种基于动态聚类的单声脉冲多波束测深数据实时滤波算法。利用地形的连续性特性, 将测深数据的异常值检测问题转化为真实地形的聚类问题, 通过不断地聚类提取真实的地形数据, 对异常值进行剔除。在聚类过程中, 由于数据量很大, 对聚类集合进行划分后采用动态聚类的方式, 同时引入地形趋势变化调节因子, 选定地形特征域, 对聚类的方向进行判断, 最后利用改进后的k均值法进行聚类目标输出。对GeoSwath多波束测深系统的真实海上试验数据的处理结果表明, 该算法对地形特征具有较强的适应能力, 且实现简单, 可用于多波束的在线滤波以及测深数据的后处理。

关键词: 多波束, 动态聚类, 特征域, k均值法

Abstract: For the data characteristics of interferometric multi-beam echo sounder (MBES), a single ping filtering method of MBES based on dynamic clustering is proposed. Considering the continuity of real terrain, the problem of outlier detection is transformed into clustering of real terrain data. Through continuous clustering of real terrain data, the outliers in data are eliminated. For the large data size in the process of clustering, dynamic clustering is adopted after partitioning clustering sets. Simultaneously, a trend adjusted factor is introduced for the feature domain selection, which is helpful for the decision of clustering direction. At last, the improved k-means method is utilized for output of clustering object. The results from processing sea test data of GeoSwath MBES show that the algorithm has good adaptability for different terrain characteristics, and is simple for implementation, which can be used for real-time filtering and post-processing of MBES data.

Key words: multi-beam echo sounder, dynamic clustering, feature domain, k-means method

中图分类号: 

  • P733.246