基于纹理特征的遥感影像监督分类

The Research ON Remote Sensing Supervised Classification Based ON Texture Feature

影像分类技术是遥感影像分析与解译的重要基础。纹理特征是影像的重要特征,本文主要实现基于纹理特征的遥感影像监督分类。首先对地物样本进行提取,通过样本训练统计各类地物纹理特征向量,建立纹理特征库;然后以各类地物的特征向量作为基准,采用最短距离分类器对影像进行分类;最后采用混淆矩阵对分类结果进行精度评定,并与ERDAS专业软件分类结果进行对比分析。实验证明,本分方法取得了与ERDAS软件相当的分类效果,从而验证本文方法的可靠性。

Image classification technology is the important foundation of remote sensing image analysis and interpretation.Texture feature is important feature,this paper realizes supervised classification based on texture feature.Firstly the features samples were extracted,and statistics all kinds of terrain texture feature vector by sample training,establishes texture feature library;And then selects all kinds of terrain feature vector as a benchmark,the minimum distance classifier for image classification;Finally makes precision evaluation to the classification results using the confusion matrix,and compares with classification results obtained by the ERDAS professional software.Experiment results show that the method proposed in this paper obtained equivalent classification results with ERDAS software,thus verifying the reliability of the method.