河流形状类型的自动图像解译分类

Interpretation classification of river channel images

针对河流遥感图像形状类型人工图像解译效率低,提出了河流图像自动识别分类方法。利用敏感因子组合条件,采用多波段组合法和区域生长分裂合并等方法提取河流并应用数学形态学方法进行处理;对得到河流图像确定类型分类特征,并给出了特征向量计算方法;河流类型特征向量散布程度大、类内聚集性较差,在支持向量机的基础上引入模糊隶属度函数,通过模糊隶属度反映样本的贡献属性,减弱噪声或野值样本对分类的影响。实验结果显示,采用模糊支持向量机有效地提高了识别准确度。

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.