五种分割算法在高分辨率遥感图像中的适用性分析  

Applicability analysis of five segmentation algorithms for high resolution remote sensing lmages

作　　者：戈乐乐[1,2] 赵银娣[1,2] 

机构地区：[1]中国矿业大学环境与测绘学院,江苏徐州221116 [2]国土环境与灾害监测国家测绘局重点实验室,江苏徐州221116

出　　处：《测绘科学》2012年第4期209-212,共4页Science of Surveying and Mapping

基　　金：国家自然科学基金资助项目(40901221);中国博士后科学基金资助项目(20090450182)

摘　　要：本文将5种图像分割算法应用在高分辨率遥感图像分割上,并利用图像分割评价指标,对5种分割算法进行了对比分析,评价了各种方法的优缺点,讨论了它们在高分辨率遥感图像分割中的适用性,明确了不同分割方法的适用条件。实验结果表明,改进的分水岭分割法与JSEG分割法在高分辨率遥感图像分割中的适用性比较强,对大小斑块分割结果都比较好,而其他3种方法不能兼顾不同等级的斑块。Five different segmentation algorithms proposed in the literature were applied to high resolution remote sensing image,including the improved watershed segmentation method,the JSEG method,the hill-climbing segmentation method,the super-pixel segmentation method and the topological derivative segmentation method.Comparisons among the above five segmentation approaches were performed using the segmentation qualitative evaluation indices,and the advantages and the disadvantages of each method were analyzed.Finally,the applicability of the five algorithms to high resolution remote sensing images was discussed,and applicable conditions of each segmentation algorithm were demonstrated.Experiment result showed that improved watershed segmentation method and JSEG segmentation method could not only obtain patterns with small sizes but also big size objects,and be better applied to high resolution remote sensing image,while the other three methods could not take enough attention to objects with different degrees.

关 键 词：高分辨率遥感图像 分割算法 评价指标 适用性 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]