摘要
本研究以北京市昌平区作为研究区域,基于2.5 m空间分辨率的SPOT5遥感影像,同时利用SEa TH算法和CART决策树两种分类方法,在自动获取分类规则的基础上,实现了对耕地信息的快速提取。结果表明,两种分类方法的总体精度均在90%以上,KAPPA系数均能达到0.80;SEa TH算法与CART决策树相比,在从高分辨率遥感影像中快速提取耕地专题信息时,耕地的漏分现象得到明显改善,且分类的稳定性更好。
In this paper,Changping District of Beijing City was selected as research area,and its farmland information was extracted from 2. 5-meter SPOT5 remotely-sensed images using SEa TH and CART methods on the basis of automatic access to classification rules. The results showed that the overall accuracy of the two classification methods were above 90%,and the KAPPA coefficient could reach 0. 80. Compared with CART method,the leakage of farmland was decreased obviously and the classification stability was better using SEa TH method.
出处
《山东农业科学》
2018年第3期132-136,141,共6页
Shandong Agricultural Sciences
基金
山东省农业科学院青年科研基金项目(2015YQN47)
国家重点研发计划项目(2017YFD0301004)
关键词
高分辨率遥感影像
耕地
信息提取
面向对象影像分析
High-resolution remotely-sensed image
Farmland
Information extraction
Object-based image analysis