基于集成学习的海岸带变化检测方法研究  

Research on Coastal Zone Change Detection Method Based on Ensemble Learning

作　　者：王娟[1] 赵吉祥 单春芝[1] 高晓慧[1] WANG Juan;ZHAO Jixiang;SHAN Chunzhi;GAO Xiaohui(North China Sea Environmental Monitoring Center,SOA,Qingdao 266000,China;China University of Petroleum(East China),College of Oceanography and Space Informatics,Qingdao 266000,China)

机构地区：[1]国家海洋局北海环境监测中心,青岛266000 [2]中国石油大学(华东)海洋与空间信息学院,青岛266000

出　　处：《海洋开发与管理》2021年第7期48-54,共7页Ocean Development and Management

基　　金：国家重点研发计划(2018YFC1407605).

摘　　要：随着我国自主研发卫星组网的不断完善,利用遥感变化检测技术进行海岸带变化检测成为海岸带监测的重要手段。针对沿海地区的变化信息提取,文章首先利用多特征构建差异影像,在此基础上采用两种集成学习方式:随机森林(Random Forest)和极端梯度提升(XGBoost),进行试验区的变化检测,并与传统的机器学习SVM、经典的变化检测方法CVA和IR-MAD进行对比,结果表明集成学习进行变化信息提取效率远超其余方式,且XGBoost在变化信息提取精度方面具有一定优势。研究成果对海岸带及海域使用开展自动化变化监测和海岸带监督管理具有重要意义。It is of great significance to carry out automatic monitoring of coastal zone and sea area use for Coastal Zone Supervision and Management.With the continuous improvement of independent research and development of satellite network in China,remote sensing change detection technology has become an important means of coastal zone monitoring.According to the change information extraction of coastal areas,this paper first used multi features to construct different images,and then used two integrated learning methods:Random Forest and eXtreme Gradient Boosting(XGBoost)were used to detect changes in the experimental area,and compared with traditional machine learning SVM,CVA and IR-MAD.The results showed that the efficiency of change information extraction by integrated learning was far higher than that of other methods,and XGBoost had some advantages in the accuracy of change information extraction.

关 键 词：海岸带 变化检测 集成学习 随机森林 极端梯度提升 

分 类 号：P74[天文地球—海洋科学] TP751[自动化与计算机技术—检测技术与自动化装置]