一种融合超像素与最小生成树的高分辨率遥感影像分割方法

发布时间：2018-08-15 14:11

【摘要】：影像分割是面向对象高分辨率遥感影像分析的基础与关键。针对传统影像分割方法易受噪声影响,且难以确定合适的影像分割尺度的问题,本文提出了一种融合超像素与最小生成树的高分辨率遥感影像分割方法。首先用简单线性迭代聚类算法对影像进行过分割生成超像素;然后初始设定影像分割数,采用区域动态约束聚类算法对超像素进行合并,获得分割数-方差和、分割数-局部方差、分割数-局部方差变化率指标图,依据3个指标图确定合适的影像分割数;最后根据确定的合适影像分割数,采用区域动态约束聚类算法对超像素重新合并得到分割结果。定性对比试验和定量评价结果表明,本文方法可以有效地克服影像噪声对分割结果的影响,获得良好的影像分割结果。
[Abstract]:Image segmentation is the foundation and key of object-oriented high-resolution remote sensing image analysis. Aiming at the problem that the traditional image segmentation method is easily affected by noise and it is difficult to determine the appropriate image segmentation scale, a high resolution remote sensing image segmentation method combining super-pixel and minimum spanning tree is proposed in this paper. A simple linear iterative clustering algorithm is used to over-segment the image to generate super-pixels, and then the segmentation number is initially set, and the region dynamic constraint clustering algorithm is used to merge the super-pixels to obtain the sum of segmentation numbers and variances. Finally, according to the suitable image segmentation number, the appropriate image segmentation number is determined according to the appropriate image segmentation number. The region dynamic constraint clustering algorithm is used to recombine the hyperpixels to obtain the segmentation results. The qualitative contrast test and quantitative evaluation results show that the proposed method can effectively overcome the influence of image noise on segmentation results and obtain good image segmentation results.
【作者单位】：武汉大学测绘遥感信息工程国家重点实验室;地球空间信息协同创新中心;
【基金】：国家自然科学基金(91438203) 国家973计划(2014CB744201)~~
【分类号】：TP751

【相似文献】