基于特征组合的无人机影像城市地表要素提取  

作　　者：宋素素 范丽丽[1] 朱静 

机构地区：[1]滁州职业技术学院,安徽滁州239000

出　　处：《科技创新与应用》2025年第7期85-88,共4页Technology Innovation and Application

基　　金：滁州职业技术学院校级科研项目(ZKZ-2024-5、ZKY-2024-7)。

摘　　要：城市地表要素是城市管理和规划的基础构成,准确获取其信息对于城市规划及社会经济分析具有重要意义。该文基于金寨县部分无人机影像数据,采用面向对象的分类方法,研究城市地表要素提取,并探究特征数量及不同特征组合对分类结果的影响。结果表明,第一,充分考虑特征的选择,并非特征数量选择越多,结果精度越高;第二,在城市地物提取中,贝叶斯模型比KNN模型精度高,总体精度可达84%,Kappa系数为0.78。Urban surface elements are the basic components of urban management and planning,and accurate acquisition of their information is of great significance for urban planning and social and economic analysis.Based on some drone image data in Jinzhai County,this paper uses an object-oriented classification method to study the extraction of urban surface elements,and explores the impact of the number of features and different feature combinations on classification results.The results show that,the selection of features is fully considered,and the more the number of features is selected,the higher the accuracy of the result;in urban feature extraction,the Bayesian model has higher accuracy than the KNN model,with the overall accuracy reaching 84%,and the Kappa coefficient is 0.78.

关 键 词：无人机 城市地表要素 面向对象 特征优选 精度 

分 类 号：P217[天文地球—测绘科学与技术]