关键词: 河流类型分类, 图像解译, 模糊支持向量机

Abstract: Due to low efficiency of river channel remote sensing images artificial interpretation, river classification of image recognition system was proposed. Sensitive factor combination was adopted, and multi-band combination method and region growing methods such as splitting and merging were taken to extract river from remote image, and mathematical morphology was applied to standard river channel. The type classification of river characteristics was determined on the obtained images, and the eigenvector method was given. As features vector received from extraction river was of a greater degree of aggregation and poor class, fuzzy support vector machine recognition was used. Fuzzy membership was reflected by the contribution of the sample properties, noise or outliers in the classification of samples were reduced. The experiments show using fuzzy support vector machine significantly improves the overall recognition rate.

Key words: river channel pattern classification, image interpretation, fuzzy support vector machine