摘要
利用全球海洋Agro观测计划提供的温、盐浮标资料,开展了印度洋海域水下环境特征提取与区划分析。在垂直方向运用Akima方法对浮标剖面进行插值,提取一组表征跃层、声速场的特征指标,并针对常规模糊C均值聚类算法中初始聚类数难以客观选取和聚类结果易陷入局部最优等问题,利用遗传算法的全局搜索能力对聚类算法作了改进。通过在遗传进化过程中引入动态变化的聚类中心解决了聚类数难以客观确定的问题,并在该算法的生存策略中引入Boltzmann选择机制,提高算法的收敛速度。在对印度洋海域温、盐跃层、声速分布及层结稳定度分析的基础上,利用改进的遗传聚类方法对印度洋海域水下环境特征进行聚类区划,得到一个基本的特征分类构型,结合各类构型的典型特征,分析了对水下潜器活动、声纳探测和水声通信等的影响。
Based on the temperature and salt buoy data provided by global ocean observation network of Agro data, character distilling and division analysis of underwater environment in the Indian Ocean were carried out. First, a group of indicators characterizing thermocline, pycnocline and sound velocity field were computed by using Akima interpolating method in the vertical direction. Then, this paper introduces two problems of conventional fuzzy c-means clustering algorithm. Addressing to hard selecting the initial cluster number objectively and clustering results easy being trapped into the local optimal, genetic clustering algorithm was improved by introducing simulation annealing ideal in the survival strategy to enhance its global search capability and dynamic changes ideal of cluster center in the evolutionary process to resolve the difficulty of identifying objective cluster number. On the basis of the analysis of thermocline, pycnocline, sound velocity distribution and stratification stability in the Indian Ocean, a general structure of character classification was obtained by dividing the characters of the underwater environment in the Indian Ocean in the method of the above improved genetic clustering algorithm. The influences of the typical features of various configurations on the Submarines movement, sonar detection and underwater communication, and other military activities were analyzed.
出处
《海洋通报》
CAS
CSCD
北大核心
2008年第6期27-33,共7页
Marine Science Bulletin
基金
中国博士后科学基金(No.2004036012)
江苏省博士后科研资助计划(0401068B)
关键词
遗传聚类
ARGO资料
跃层
genetic clustering algorithm
Argo data
ocean environment characteristics