中文摘要：

为探讨高分辨率遥感图像用于中小尺度森林分类的模式，利用SPOT5遥感数据、地面样地调查数据和前期森林资源规划设计调查GIS资料，以图像的光谱和纹理信息为主、历史调查数据的知识为辅构建专家知识分类系统对SPOT5图像进行森林分类，并探讨了历史调查数据在该模式中的贡献率。结果表明，对于所选取的8个类别，总体分类精度达到了92．97％，各类别的分类精度均达到87％以上，分类效果良好；历史调查数据在分类过程中的总体贡献率为11．55％，对提高SPOT5图像分类有较大的帮助作用，尤其对竹林、八角和玉桂、灌木林分类的辅助作用表现更为明显。

英文摘要：

To work out a classification routine of forestland in local level using high-resolution satellite imagery, SPOT5 data, field sampling plot data and （;IS data of historic forest resources inventory were used to build an expert classification system to classify the forestland with SPOT5 imagery, which was composed of information of spectrum and texture, and the knowledge mined from the historie inventory data. The results indicated that when the imagery was classed to 8 types by this system, the total accuracy of classification was 92.97%, the accuracy of all classes were higher than87% ; and the historic inventory data played 11.55% contribution in this system, it was helpful to improve the classification accuracy of SPOT5 imagery, especially to classification of bamboo, truestar anisetree and cassiabarktree, and shrub.