基于知识规则的遥感影像土地利用分类研究

以重庆北碚主城区为研究区域,提出一种基于遥感影像的光谱特征、灰度共生矩阵纹理特征和形状特征建立知识规则的方法,对该研究区进行土地利用分类。对QuickBird卫星影像进行分辨率融合,重点分析典型地物的光谱特征和纹理特征,并根据形状特征建立形状指数,综合上述知识建立提取规则对地物进行自动分类。分类结果表明,该方法能够消除单纯利用光谱信息分类的不足,有效解决地物混分现象,达到了较高的分类精度。

Land Use Classification of Remote Sensing Image Based on Knowledge Rules

Taking Beibei urban region as study area,and based on knowledge rules combining spectral feature,Gray Level Co-occurrence Matrices texture feature and shape feature,a classification method is presented in this paper.Firstly,the resolution fusion method is used for QuickBird PAN and multi-spectral images.Then,the spectrum feature and texture feature of the typical land-use types are analyzed,and shape index is established based on shape feature.Finally,the knowledge rules are established and applied to automatically classification of different land-use types,associated with an accuracy assessment.The classified results show that this method eliminates the deficiency of using spectral information only,and reaches high classification accuracy.