基于LBP改进的MeanShift高分辨率遥感影像分割方法  

Segmentation method of improved MeanShift high resolution remote sensing image based on LBP

作　　者：柯佳宏 吴安东 李勇[1] 刘鑫[1] 

机构地区：[1]四川省地质矿产勘查开发局区域地质调查队,四川成都610213

出　　处：《矿山测量》2017年第6期64-68,共5页Mine Surveying

摘　　要：MeanShift算法广泛应用于自然场景图像和医学图像分割中,针对MeanShift算法分割遥感图像的边界模糊和精度不高的问题,文中提出一种全新的基于LBP纹理特征的MeanShift遥感图像分割方法。首先加入LBP纹理特征于MeanShift遥感图像分割过程中,并对分割后区域采用区域面积加权的区域相似度准则和最小对象面积参数对遥感影像进行区域合并。实验结果表明:本方法相比传统的MeanShift遥感影像分割方法能得到更好的分割效果。MeanShift algorithm was widely used in the nature scenes and the biomedical image segmentation. In the paper,a new remote sensing image segmentation method based on the texture feature of LBP was propose for the fuzzy boundary and the low precision of remote sensing image with MeanShift algorithm. Firstly,the LBP texture feature was added into the process of remote sensing image segmentation with MeanShift,and regional similarity criteria and the minimum object area parameters of area weighted were used to carry out the area amalgamation for the segmented area. Experimental result showed that: the method could get the better segmentation results comparing with the traditional MeanShift remote sensing image segmentation method.

关 键 词：遥感影像 MEANSHIFT LBP纹理特征 图像分割 

分 类 号：TD178[矿业工程—矿山地质测量]