联合支持向量机和增强学习算法的多波束声学底质分类  

Multibeam acoustic seabed classification combining SVM and adaptive boosting algorithm

作　　者：纪雪 唐秋华 陈义兰 李杰 丁德秋 JI Xue;TANG Qiuhua;CHEN Yilan;LI Jie;DING Deqiu(State Key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing, Wuhan 430079, China;First Institute of Oceanography, Ministry of Natural Resources, Qingdao 266061, China;College of Geodesy and Geomatics, Shandong University of Science and Technology, Qingdao 266590, China)

机构地区：[1]测绘遥感信息工程国家重点实验室,湖北武汉430079 [2]自然资源部第一海洋研究所,山东青岛266061 [3]山东科技大学测绘科学与工程学院,山东青岛266590

出　　处：《测绘学报》2021年第7期972-981,共10页Acta Geodaetica et Cartographica Sinica

基　　金：国家自然科学基金(41876111);国家海洋局极地考察办公室项目(IRASCC2020-2022)。

摘　　要：基于多波束的声学底质分类是近年来快速发展起来的新型海底底质探测技术。针对多波束声学底质分类中底质类型多样化、类型之间差异较小等多分类难点问题,本文提出一种GA-SVM-AdaBoost算法。利用自适应性和全局搜索能力强的遗传算法(genetic algorithm,GA)去优化支持向量机(support vector machines,SVM),以获得最优模型初始参数,并将多个GA优化后的SVM作为弱分类器组成AdaBoost强分类器。对胶州湾采集到的多波束反向散射强度数据,经过精细处理后生成海底声呐镶嵌图;构建SVM-RFE-CBR算法从提取的36维声强空间特征中筛选出10维优势特征,将其输入到GA-SVM-AdaBoost模型中进行分类识别。通过与SVM、GA-SVM、基于单层决策树的AdaBoost分类模型对比,GA-SVM-AdaBoost算法的总体分类精度高达92.19%,优于另外3种模型,证明GA-SVM-AdaBoost分类模型可有效应用于高精度海底底质类型识别。As a new technology,multibeam acoustic classification has been rapidly developed in recent years.A seabed sediment classification approach,GA-SVM-AdaBoost algorithm,is proposed by using the genetic algorithm(GA)optimized support vector machines(SVM)classifier as the AdaBoost weak classifier to solve the multi-classification problem in multibeam acoustic seabed classification.The sonar mosaic is obtained from multibeam echo sounder backscatter data collected in the Jiaozhou Bay within fine processing.The 10 dimensions advantage features are selected by SVM-RFE-CBR algorithm before input GA-SVM-AdaBoost classification model.Compared with SVM,GA-SVM and AdaBoost based on single-layer decision tree,the classification results of GA-SVM-AdaBoost algorithm are more satisfactory.The total classification accuracy is as high as 92.19%,which is better than the other three models.It is proved that the proposed method can be effectively applied to high precision seabed sediment identification.

关 键 词：多波束测深系统 声学底质分类 支持向量机 遗传算法 ADABOOST算法 

分 类 号：P229[天文地球—大地测量学与测量工程]