【摘要】 采用环境减灾二号A/B卫星数据，在福建武夷山国家公园和东北虎豹国家公园2个试验区开展森林树种识别的应用研究。分别提取2个试验区影像的光谱特征、归一化植被指数(NDVI)和基于主成分第一分量的8个纹理特征信息，采用支持向量机(SVM)和原型网络分类对森林主要树种进行识别，并利用验证样本进行精度评价。结果表明：在福建武夷山国家公园试验区，SVM和原型网络分类的总精度分别为86.37%和91.11%,Kappa系数分别为0.83和0.90。在东北虎豹国家公园试验区，SVM和原型网络分类的总精度分别为91.77%和91.34%,Kappa系数均为0.90。总体来说，环境减灾二号A/B卫星数据在2个试验区的主要树种识别精度较好，后续能较好地满足林业行业相关业务应用需求。更多还原

【Abstract】 The application research of forest tree species identification is carried out in Fujian Wuyishan National Park and Northeast Hubao National Park by using HJ-2 A/B satellites data. Spectral features, NDVI(normalized difference vegetation index) and eight texture feature information based on the first component of the principal components are extracted from the images of the two test areas, and the main tree species are identified by SVM(support vector machines) and prototype network classification methods. The accuracy evaluation is performed using the va lidation samples. The results show that: in Wuyishan National Park, the total accuracy of SVM and prototype network classification are 86.37% and 91.11%, respectively, and the Kappa coefficients are 0.83 and 0.90, respectively; in Hubao National Park, the total accuracy of SVM and prototype network classification are 91.77% and 91.34%, respectively, and the Kappa coefficients are 0.90. In general, the identification accuracy of the main tree species is relatively good, and the satellites can meet the needs of related business applications in the forestry industry in the future.更多还原