关键词: 高分多模卫星, 林业地类识别, 树种识别, 随机森林, K最邻近分类法

Abstract:

Remote sensing classification technology has always been the hotspot and difficulty in forest and grassland monitoring.Since 2021,the annual comprehensive ecological monitoring of forest and grassland has been widely applied remote sensing classification technology in the interpretation of patch types and changes of forest land,grassland and wetland.GFDM Satellite is the first civil optical remote sensing satellite with the resolution better than 0.5 meters in our country.In this paper,we used the GFDM satellite data to identify forest land types and tree species in Taoyuan County and Jishou City of Hunan Province.The results showed that the random forest method performed well in the identification of forest land types.The overall classification accuracy of forest land,wetland and other forest land was 89.56%,and the Kappa coefficient was 0.733.The K-nearest neighbor method was used to identify Chinese fir,Masson pine,shrub group and citrus.The overall recognition accuracy of the four main tree species was 77.58%,and the Kappa coefficient was 0.697.In general,GFDM satellite has good ability in remote sensing classification and target recognition,and has great application potential in forest and grassland monitoring.

Key words: high resolution multi-mode satellite, forest land type identification, tree species identification, random forest, K-nearest neighbor