基于可变尺度Mean-Shift的农田高分遥感影像分割算法

发布时间：2018-03-07 13:04

本文选题：可变尺度　切入点：Mean-Shift　出处：《国土资源遥感》2017年03期 　论文类型：期刊论文

【摘要】：为了提升农田高分遥感影像(high spatial resolution remote sensing image,HRI)的信息提取效果,提出了一种新的农田HRI分割算法。传统的Mean-Shift(MS)HRI分割算法仅利用全局或单一的尺度参数;而常规可变尺度MS算法在尺度参数估算中也只考虑光谱信息。这些都导致其分割结果难以完整地展现不同尺度的农田区域。针对该问题,在MS算法的基础上进行了改进:第一,提出了一种局部可变尺度参数的估计方法;第二,提出了利用局部可变尺度进行MS滤波的模型函数。该改进算法主要包含3步:(1)为了全面考虑不同波段的响应变化,在MS滤波核函数中采用了对角化的尺度参数矩阵,并将其与采样点密度估计模型相结合,导出了一种可变尺度MS滤波的迭代函数;(2)为了提高算法的自动化程度,利用局部光谱变化与边界强度信息,提出了一种新的局部尺度参数估算方法;(3)将MS滤波结果输入到基于分形网络演化方法(fractal net evolution approach,FNEA)的空间聚类算法中,得到最终的分割结果。利用Rapid Eye与OrbView3的2景HRI进行了算法验证。实验结果表明,所提出的改进算法能够优化农田HRI分割的精度。
[Abstract]:In order to improve the information extraction effect of high spatial resolution remote sensing image, a new field HRI segmentation algorithm is proposed. The traditional Mean-Shift(MS)HRI segmentation algorithm only uses global or single scale parameters. However, the conventional variable scale MS algorithm only considers spectral information in the estimation of scale parameters. These results make it difficult for the segmentation results to show the different scale farmland regions completely. To solve this problem, we improve the algorithm based on the MS algorithm: first, A local variable scale parameter estimation method is proposed. Secondly, a model function using local variable scale for MS filtering is proposed. The improved algorithm mainly includes 3 steps: 1) in order to take into account the response variation of different bands. In order to improve the automatization of the algorithm, the diagonalized scale parameter matrix is used in the MS filter kernel function, and combined with the sampling point density estimation model, an iterative function with variable scale MS filter is derived. Based on the local spectral variation and boundary strength information, a new local scale parameter estimation method is proposed, in which MS filtering results are input into the spatial clustering algorithm based on fractal net evolution approximation FNEA (fractal network evolution method). Finally, the final segmentation results are obtained. The algorithm is verified by Rapid Eye and OrbView3's 2 scene HRI. The experimental results show that the proposed improved algorithm can optimize the precision of HRI segmentation in farmland.
【作者单位】：内蒙古农业大学水利与土木建筑工程学院;
【基金】：国家自然科学基金项目“科尔沁沙地典型生态系统水热通量传输机理及其与植被耦合关系试验和模拟研究”(编号:51569017)、“内蒙古典型草原水文过程及其扰动与触发草地退化的水文临界条件实验与模拟研究”(编号:51269014) 内蒙古自然科学基金项目“半干旱区沙地典型生态系统水热通量传输机理研究”(编号:2015MS0514) 中国博士后科学基金面上资助项目“西部地区博士后人才资助计划”(编号:2015M572630XB)共同资助
【分类号】：TP751

【相似文献】