典型遥感影像分类方法适用性分析  

Scalability analysis of typical remote sensing image classification method

作　　者：武英洁 冯勇 徐晓琳 刘思宇 朱辉 WU Yingjie;FENG Yong;XU Xiaolin;LIU Siyu;ZHU Hui(Key Laboratory for Meteorological Disaster Prevention and Mitigation of Shandong,Jinan 250031,China;Shandong Meteorological Data Center,Jinan 250031,China)

机构地区：[1]山东省气象防灾减灾重点实验室,山东济南250031 [2]山东省气象数据中心,山东济南250031

出　　处：《现代电子技术》2024年第6期137-141,共5页Modern Electronics Technique

基　　金：山东省气象局科研项目(2021SDQXZ06;2022SDQN04)。

摘　　要：分类技术是从遥感影像数据中提取信息必不可少的步骤,选择合适的分类器对提高分类精度至关重要,针对特定的研究如何选择适合的分类算法是一个亟需研究的问题。以北京市中心诚区中某一区域为研究区,应用“高分一号”(GF-1)数据和Landsat 8数据,分别采用最常用且分类精度相对较高的监督分类中的最小距离法、最大似然法、支持向量机法,将研究区分为林地、草地、水体、裸土、建筑物5种类型,并对分类结果进行空间分布、面积、精度三个方面的比对分析。结果表明,分类算法的选择主要取决于研究区的地物特点,其中最小距离法应用于植被覆盖面积较大的区域时精度较高,最大似然法适合于分类建筑物较多的区域,支持向量机法对各类地物的分类具有较高的普适性。The classification is an indispensable step to extract information from remote sensing images.Selecting appropriate classifier is very important to improve the classification accuracy.How to choose suitable classification algorithm for specific research is an urgent problem to be solved.An area in the central of Beijing is selected as the research area,the GF-1data and Landsat 8 data are applied,and the most commonly used and relatively high classification accuracy supervised classification methods such as minimum distance method,maximum likelihood method,and support vector machine method are applied,resperctively.The research is divided into five categories:forest land,grassland,water body,bare soil and buildings.The comparative analysis for the classification results in terms of spatial distribution,area,and accuracy are conducted.The results show that the selection of the classification algorithm mainly depends on the characteristics of the ground features in the study area,among which the minimum distance classification has a higher accuracy when applied to the area with large vegetation cover,the maximum likelihood classification is suitable for the area with many buildings,and the support vector machine has high universality to all kinds of objects.

关 键 词：遥感影像 分类技术 最小距离分类 最大似然分类 支持向量机 GF-1 Landsat 8 

分 类 号：TN929-34[电子电信—通信与信息系统] TP391[电子电信—信息与通信工程] P237[自动化与计算机技术—计算机应用技术]