基于关联规则遥感影像的分类研究——以渭库绿洲为例

Study on Remote Sensing Image Classification Association Rules Based on Weigan and Kuqa Rivers Delta Oasis as an Example

对遥感影像的计算机图像处理与识别技术是现阶段遥感技术研究的热点内容,本文对新疆渭干–库车河流域不同分辨率的全色影像与多光谱影像融合并构建多源空间数据库,再通过与面向对象分类法相结合,进行图像分割、特征提取后将结果取值输入Matlab软件,编制相应的算法程序,构建关联规则分类器,探索分类规则并进行遥感影像的分类.同样基于Matlab软件平台,根据最大似然分类算法对同一研究区影像进行分类,并对分类结果进行精度检验.结果表明:相较传统分类方法,关联规则方法拓宽了传统逻辑推理分类方法中获取信息的途径,使分类更加智能化,提高并增强了影像分类速度与算法可靠性,分类精度得到显著提高,Kapp系数为0.83.因此,关联规则方法更适用于影像分类,可以作为研究区有效信息提取的支持手段.

The computer image processing and recognition technology of remote sensing image is the hotspot of remote sensing technology research at present. In this paper, we fused different resolution panchromatic images and multi spectral images to construct multi-source spatial database on Weigan-Kuqa River Basin of Xinjiang. Then we combined with object oriented classification method to segment and extract information. The results of the value were inputted in Matlab software and programming corresponding algorithm to build the association rule classifier, explored the classification rules and classified the remote sensing image. Based on the Matlab software platform, the images of the same study area are classified according to the maximum likelihood classification algorithm. And the accuracy of the classification results is verified. Results show that the association rule method can broaden the way of obtaining information compared to the traditional logical reasoning classification method. It made the classification more intelligent,improved and enhanced the image classification speed and algorithm reliability. The classification accuracy is improved remarkably and the Kapp coefficient is 0.83. Therefore, the association rule method is more suitable for image classification. The way can be used as a means of supporting the effective information extraction in the study area.