中文摘要：

针对均值漂移分割算法假设样本处于欧式空间,而遥感影像由于波段范围广,不能像普通彩色图像一样,能够由RGB空间转换到LUV空间来满足这种假设的问题,该文提出采用主成分分析（Principal Component Analysis,PCA）对遥感影像进行变换,以满足这种假设。对于输入的一幅遥感影像,首先采用PCA对该影像进行变换,并且将变换后的数据线性拉伸到0~255之间,然后替换原始影像数据,并调用均值漂移算法进行影像分割。实验结果表明,采用所提出方法得到的结果优于直接对原始影像进行均值漂移分割的结果。

英文摘要：

Mean Shift segmentation algorithm requires the validity of euclidean metric for the feature space. In general, we can convert RGB image from RGB space to LUV space to meet this requirement,but remote sensing image has a wide range of bands,this conversion method is no longer applicable. In order to solve this problem, this paper adopts principal component analysis （PCA） to transform remote sensing image to meet this requirement. First, use PCA to transform the input remote sensing image. Second, stretch the data to 0 - 255 by the linear transformation. And then replace the original image data. Finally,call the Mean Shift algorithm for image segmentation. Experimental results show the effectiveness of the proposed method.