摘要: |
该文基于高分卫星资料,通过基于规则的面向对象分类、基于最邻近法的监督分类及基于CART分类器的监督分类三种不同分类方法,对复杂山区光伏电站进行提取,对比三种分类提取方法结果并完成精度验证。结果表明,合理的分割参数有利于提高光伏电站提取精度;基于规则的面向对象分类法光伏电站提取精度最佳,最邻近分类法次之,CART分类器分类法最差,可利用基于规则的面向对象分类法较为准确地进行复杂山区光伏电站信息提取,为光伏产业健康、合理发展提供一定的数据支撑。 |
关键词: 面向对象;最邻近分类法;CART分类器;光伏电站 |
DOI: |
投稿时间:2022-08-04修订日期:2023-01-03 |
基金项目:贵州省气象局科研业务项目(黔气科登[2022]05-03号) |
|
Research on Extraction of Photovoltaic Power Stations in Mountains Based on GF Satellite Images |
liuyun,Song Shanhai,Li Huixuan,Tian Pengju,Wang Wei |
(GuiZhou ecological meteorological and Satellite Remote Sensing Center;The Engineering and Technology Center of Special Fisheries, Guizhou) |
Abstract: |
Using GF satellite data, the photovoltaic power stations in mountains area were extracted by three different classification methods: rule-based object-oriented classification, supervised classification based on K-Nearest Neighbor method and supervised classification based on cart classifier. Then, compare the results of three extraction methods and complete the accuracy verification.The results show that, reasonable segmentation parameters are helpful to improve the extraction accuracy of photovoltaic power plants.The rule-based object-oriented classification method has the best extraction accuracy, the nearest neighbor classification method takes the second place, and the cart classification method is the worst.The rule-based object-oriented classification method can be used to accurately extract the information of photovoltaic power stations in complex mountainous areas, providing certain data support for the healthy and reasonable development of photovoltaic industry. |
Key words: object-oriented; nearest neighbor classification; CART classifier; photovoltaic power station |