遥感分类技术一直是林草行业应用的热点和难点,2021年开展的国家林草综合生态年度监测开始广泛应用遥感技术开展林地、草地和湿地图斑变化判读,地类前后变化的识别精度是其难点和关键。通过利用国内首颗分辨率优于0.5m的高分多模卫星,在湖南省桃源县、吉首市2个试验区分别开展林业主要地类识别和树种精细识别应用实践研究。结果表明:随机森林方法在林业地类识别中表现较好,林地、湿地、其他林地等主要地类的总体分类精度为89.56%,Kappa系数为0.733;K最邻近分类法对杉木、马尾松、灌木组、柑桔4个主要树种的总体识别精度为77.58%,Kappa系数为0.697。总体而言,高分多模卫星遥感分类和目标识别能力较好,在林草调查监测工作中应用潜力较大。

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.