基于对象特征的遥感影像分类

为了有效地实现分类识别,采用多尺度分割后影像为数据源,通过计算影像对象特征,实现特征集构建,并对特征集进行特征提取,辅助决策树的特征选择方法,达到特征优选的目的,针对影像对象的特征实现对TM影像的分类,实验表明该方法较一般的特征提取方法精度大幅提高,且完成可视化特征提取的重要任务。

Classification of Remote Sensing Image Based on Object Features

In order to achieve the classification, the multistate segmentation image as data source, through the calculation of image ob-ject features, building the feature set, and extracting characteristics of the feature set, advance the feature selection method for deci-sion tree, to achieve the purpose of feature selection, according to the characteristics of image objects are implemented on the TM im-age classification, experiments show that the method is more improved than a general method for extracting accuracy, and achieve the important task of visual feature extraction.