For automatic classification of satellite imagery by unsupervised method, category assignment has been an important but difficult task. In this paper, we propose an automatic assignment method using a vegetation map with seven merged categories. The merged raster image with UTM coordinate system and 30 m spatial resolution is obtained by converting digital vector vegetation data edited by Japan Environmental Agency. Unsupervised cluster image data are compared with the vegetation categories to produce a contingency matrix and an assignment priority table. The contingency matrix is used for evaluation of maximum classification accuracy, and priority table is used to combine a relationship between cluster and category. At the Kesennuma city 500 × 500 pixels study area, the overall accuracy is greater than 0.7.