融合边界信息的高分辨率遥感影像分割优化算法

发布时间：2018-06-22 05:38

本文选题：高分辨率影像 + 影像分割　；参考：《中国图象图形学报》2016年08期

【摘要】：目的针对目前区域分割算法获取的区域边界与真实地物边界不一致问题,利用高分辨率遥感影像地物内具有均质性和地物间边缘信息突出的特点,提出一种融合边界信息的高分辨率遥感影像分割优化算法。方法首先采用Canny算法对遥感影像进行边缘提取并进行边缘连接处理,产生闭合边界;然后将边界与初始分割结果进行融合处理,获得新的分割结果;最后在闭合边界约束下,基于灰度相似性准则对新的分割结果进行区域合并,获得优化后的最终分割结果。结果采用本文提出的分割优化算法对Mean Shift算法和eCognition软件获得的分割结果进行优化处理,优化后的分割结果与初始分割结果相比正确分割率(RR)平均提高了4%,验证了本文算法的有效性。结论该优化算法适用性广,可优化基于区域、基于边界和基于聚类等多种分割方法,同时该算法既能保持高分辨率遥感影像分割的区域完整性,又能保持地物边缘细节特征,提高了分割精度。
[Abstract]:Aim to solve the problem of the inconsistency between the region boundary and the real ground object boundary obtained by the current region segmentation algorithm, using the characteristics of homogeneity and prominent edge information between the ground objects in the high-resolution remote sensing image. An optimal segmentation algorithm for high resolution remote sensing images is proposed. Methods the edge of remote sensing image was extracted by Canny algorithm and the edge was connected to produce the closed boundary. Then the edge was fused with the initial segmentation result to obtain a new segmentation result. Finally, under the constraint of closed boundary, a new segmentation result was obtained. Based on the gray similarity criterion, the new segmentation results are merged and the final segmentation results are obtained. Results using the segmentation optimization algorithm proposed in this paper, the mean shift algorithm and the segmentation results obtained by the software of Cognition were optimized. Compared with the original segmentation results, the optimized segmentation results increase the average correct segmentation rate (RR) by 4%, which verifies the effectiveness of the proposed algorithm. Conclusion this algorithm can be used to optimize the segmentation methods based on region, boundary and clustering. The algorithm can not only maintain the regional integrity of high resolution remote sensing image segmentation, but also preserve the feature of ground object edge detail. The segmentation accuracy is improved.
【作者单位】：云南师范大学旅游与地理科学学院;西部资源环境地理信息技术教育部工程研究中心;中南大学地球科学与信息物理学院地理信息系;
【基金】：国家自然科学基金项目(41201463,41201428) 国家重点基础研究发展计划(973)基金项目(2012CB719906) 国家高技术研究发展计划(863)基金项目(2012AA121400) 江苏省资源环境信息工程重点实验室开放基金资助项目(JS201301) 云南省教育厅基金项目(011Y311)~~
【分类号】：TP751

【相似文献】