舰船遥感图像的分类方法研究  

Research on classification method of ship remote sensing image

作　　者：张艳华[1] 王百勇[1] 王锦明 ZHANG Yan-hua;WANG Bai-yong;WANG Jin-ming(Shanxi Conservancy Technical Institute Mapping Engineering Department,Yuncheng 044000,China;Shanghai East Sea Marine Engineering Survey and Design Institude,Shanghai 200137,China)

机构地区：[1]山西水利职业技术学院测绘工程系,山西运城044000 [2]上海东海海洋工程勘察设计研究院,上海200137

出　　处：《舰船科学技术》2020年第4期52-54,共3页Ship Science and Technology

基　　金：2018年山西省教育科学“十三五”规划课题(GH-18217)

摘　　要：遥感图像分类一直是舰船应用领域的关键技术,由于遥感图像具有多波段、高维特征等特点,当前遥感图像分类技术面临一定的挑战。为了获得更优的舰船遥感图像分类结果,提出一种多分类器加权组合的舰船遥感图像分类方法。首先分析舰船遥感图像分类研究的历史,找到导致单分类器的舰船遥感图像分类错误率高的原因,然后引入双边滤波算法对原始舰船遥感图像进行去噪,并提取舰船遥感图像分类纹理特征,最后采用多种方法建立舰船遥感图像分类器,并对它们进行加权组合,输出舰船遥感图像的最终归属。仿真测试结果表明,本文方法获得了比单分类器更优的舰船遥感图像分类正确率,舰船遥感图像分类结果更加可靠。Remote sensing image classification is always the key technology in the field of ship application.Because of the characteristics of multi band and high dimension,the current remote sensing image classification technology is facing certain challenges.In order to obtain better results of ship remote sensing image classification,a ship remote sensing image classification method with multi classifier weighted combination is proposed.First of all,the history of ship remote sensing image classification is analyzed,and the reason of high error rate of ship remote sensing image classification caused by single classifier is found.Then,the bilateral filtering algorithm is introduced to denoise the original ship remote sensing image,and the texture features of ship remote sensing image classification are extracted.Finally,the ship remote sensing image classifier is established by various methods,and they are weighted and combined to output The result of simulation test shows that this method has better classification efficiency than single classifier,and the classification result is more reliable.

关 键 词：舰船工作环境 遥感技术 图像分类 噪声过滤 分类正确率 

分 类 号：TN215[电子电信—物理电子学]