基于贝叶斯信息准则的多波束海底底质非监督分类  

Unsupervised classification of multibeam seafloor sediment based on Bayesian information criterion

作　　者：牛樱晓 崔晓东[1] 胡海洋 丁继胜 阳凡林[1,3] NIU Yingxiao;CUI Xiaodong;HU Haiyang;DING Jisheng;YANG Fanlin(College of Geodesy and Geomatics,Shandong University of Science and Technology,Qingdao 266590,China;First Institute of Oceanography,Ministry of Natural Resources,Qingdao 266061,China;Key Laboratory of Ocean Geomatics,Ministry of Natural Resources,Qingdao 266590,China)

机构地区：[1]山东科技大学测绘与空间信息学院,山东青岛266590 [2]自然资源部第一海洋研究所,山东青岛266061 [3]自然资源部海洋测绘重点实验室,山东青岛266590

出　　处：《山东科技大学学报（自然科学版）》2023年第2期22-32,43,共12页Journal of Shandong University of Science and Technology(Natural Science)

基　　金：国家自然科学基金项目(41930535,41906165,52201400);山东省自然科学基金项目(ZR2022QD043);山东省研究生教育创新计划建设项目(SDYJG19083);山东科技大学科研创新团队支持计划项目(2019TDJH103)。

摘　　要：海底底质分类作为海洋测绘的主要研究方向,适用于无先验样本的非监督底质分类技术是重要的研究内容。传统K-均值聚类方法存在主观设置参数的局限性,无法准确高效判别最优聚类数。因此,本研究首先依据多波束反向散射强度角度响应曲线数据,结合贝叶斯信息准则构建特征概率密度模型,实现聚类数的准确寻优;其次,融合多波束强度特征和地形特征构建基于K-均值的海底底质非监督分类模型。实验表明:该方法准确解释了实际底质的分布情况,总体分类精度和Kappa系数分别为77.97%和0.67,相较于最大最小值法和中心点法分别提高了13.93%、0.22和1.69%、0.02,收敛迭代次数比中心点法减少67次,提高了分类效率,为海底底质的自适应聚类提供参考。Seafloor sediment classification is a main research direction of marinesurveying and mapping and unsupervised substrate classification techniques applicable to no priori samples are important research contents.With the limitation of subjective setting of parameters,the traditional K-means clustering method cannot accurately and efficiently discern the optimal number of clusters.This paper first constructed a feature probability density model based on multibeam backscattered intensity angular response curve data combined with Bayesian Information Criterion to achieve the accurate clustering number finding.Secondly,an unsupervised seafloor sediment classification model based on K-means was constructed by integrating multibeam intensity features and topographic features.The experiments show that the proposed method accurately explains the actual sediment distribution.The overall classification accuracy and Kappa coefficient are 77.97%and 0.67 respectively,which are 13.93%,0.22 and 1.69%,0.02 up respectively compared with the MaxMin and the K-medoids method.The number of convergence iterations is 67 times less than that of the K-medoids.The proposed method can improve the classification efficiency and provide a reference for adaptive clustering of submarine substrates.

关 键 词：角度响应 反向散射数据 贝叶斯信息准则 K-均值聚类 底质分类 

分 类 号：P717[天文地球—海洋科学] P229.1