遥感图像非监督计算机分类方法的研究

Research on Methods of Computer Unsupervised Classification to Remote Sensing Image

通过对几种常用的非监督计算机遥感图像分类方法,如k-means、层次聚类和神经网络的分析研究发现,由于这些方法不能克服数据噪声点的影响,输出结果对输入参数依赖性较大,使其对图像的分类效果受到影响。为了提高图像的非监督分类效果,本文提出了一种基于密度和自适应密度可达聚类算法。实验分析表明,与常用的分类方法相比,该算法具有良好的分类效果。

Through the analysis of commonly used methods of computer unsupervised classification to remote sensing image, such as k-means, hierarchical clustering and neural network clustering, it＇ s discovered that the classification effectiveness of the methods are affected by noise data points and input parameters of the methods. In order to improve the effectiveness of computer unsupervised classification, a clustering algorithm based on density and adaptive density-reachable is presented. Experimental Analysis shows that the classification effectiveness of the algorithm is better than the commonly used methods.