基于决策树算法的Landsat 8 OLI影像海岸类型自动识别方法  

Automatic Recognition of Coast Types in Landsat 8 OLI Images by Using Decision Tree Algorithm

作　　者：王常颖[1] 谭萌[2] 

机构地区：[1]青岛大学信息工程学院,山东青岛266071 [2]国家海洋局北海信息中心,山东青岛266033

出　　处：《海岸工程》2015年第1期47-54,共8页Coastal Engineering

基　　金：山东省科技发展计划--基于数据挖掘的TM影像山东省海岸类型自动识别及海岸线快速提取技术研究(2011YD15005);国家自然科学基金--海洋灾害大数据分析的系统模型研究及应用(41476101);山东省自然科学基金重点项目--复杂海量数据可计算建模及基于数据挖掘的机理发现方法的研究(ZR2012FZ003);青岛市科技计划--可计算复杂网络模型研究及其在海洋生态灾害预警预测与海洋执法中的应用(13-1-4-121-jch)

摘　　要：针对基岩、人工、砂质和淤泥质海岸四种海岸类型,采用数据挖掘算法中的C4.5决策树算法对Landsat 8OLI多光谱影像数据进行分析。首先得出4种海岸类型中所包含的海水、陆地、植被、养殖区、淤泥、沙滩六种地物的识别规则;然后利用不同海岸类型所包含的地物类型差异,分别提出了基岩、人工建筑、人工养殖、沙滩、淤泥五种海岸类型的自动识别规则。实例验证结果表明,对于地物特征明显的人工海岸和基岩海岸,我们提出的海岸类型自动识别方法的精度可达到100%,而对于光谱相近的淤泥质海岸和人工养殖海岸其识别精度较低,大于80%。For recognizing the following four types of coasts including bedrock coast,artificial coast,sandy coast and mud coast,the Landsat 8OLI multispectral images are analyzed by means of C4.5decision tree algorithm of data mining technique.Firstly,the rules for recognizing the land features such as seawater,land,vegetation,aquaculture,mud and beach which are contained in the four coast types are worked out,and then the rules for automatically recognizing five types of coasts including bedrock coast,man-made structure coast,artificial aquaculture coast,sandy coast and mud coast are respectively proposed according to the differences in land features contained in different types of coasts.The experimental results show that the accuracy of the proposed automatic recognition rules is 100%for the artificial and the bedrock coasts with obvious land features and greater than 80%for the mud and the artificial aquaculture coasts with similar spectra.

关 键 词：LANDSAT 8 海岸类型 自动识别 数据挖掘 C4.5决策树算法 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]