联合特征选择与改进集成学习模型的海底底质分类  

Joint feature selection and improved ensemble learning model for seabed sediments classification

作　　者：逄岩 许枫[1] 刘佳[1] 李益丞 赵越[1] PANG Yan;XU Feng;LIU Jia;LI Yicheng;ZHAO Yue(Ocean Acoustic Technology Laboratory,Institute of Acoustics,Chinese Academy of Sciences,Beijing 100190;University of Chinese Academy of Sciences,Beijing 100049)

机构地区：[1]中国科学院声学研究所海洋声学技术实验室,北京100190 [2]中国科学院大学,北京100049

出　　处：《声学学报》2023年第1期83-92,共10页Acta Acustica

基　　金：中国科学院青年创新促进会项目(2020023)资助。

摘　　要：针对侧扫声呐获取类型复杂的海底底质数据分类问题,提出联合特征选择与改进Stacking模型的数据自驱动分类方法。该方法首先在海底散射数据多域态特征的基础上采用ReliefF算法提取有效的低维度特征,然后将人工鱼群算法与Stacking模型结合形成改进集成学习分类器,完成海底底质分类。海上数据处理结果表明该方法可对多种海底底质类型进行分类,分类准确率、Kappa系数和F1-score分别达到85.55%,0.857和0.887,证明了该方法的有效性。Aiming at the classification of seabed sediments data with complex type acquired by side scan sonar,a data self-driven method based on joint feature selection and improved Stacking model is proposed to accomplish the classification of seabed sediments.The method firstly uses the ReliefF algorithm to extract effective low-dimensional features based on the multi-domain features of the seabed scattering data,and then combines the artificial fish swarm algorithm with Stacking model to form improved ensemble learning classifier to complete the classification of seabed sediments.The results of marine data processing show that the method can classify various types of seabed sediments,and the classification accuracy,Kappa coefficient and F1-score reach 88.55%,0.857 and 0.887,respectively,indicating the effectiveness of this method.

关 键 词：人工鱼群算法 分类准确率 数据分类 学习分类器 海底底质分类 Kappa系数 侧扫声呐 自驱动 

分 类 号：U666.7[交通运输工程—船舶及航道工程]