基于超像素的高分遥感影像分割算法  

High-resolution remote sensing image segmentation algorithm based on superpixel

作　　者：向泽君 蔡怤晟[1,2,3] 楚恒 黄磊[1,2,3] XIANG Ze-jun;CAI Fu-sheng;CHU Heng;HUANG Lei(School of Telecommunication and Information Engineering,Chongqing University of Posts and Telecommunications,Chongqing 400065,China;Key Laboratory of Optical Communication and Network in Chongqing,Chongqing University of Posts and Telecommunications,Chongqing 400065,China;Key Laboratory of Ubiquitous Sensing and Networking in Chongqing,Chongqing University of Posts and Telecommunications,Chongqing 400065,China;Chongqing Urban Planning Bureau,Chongqing 401121,China;Chongqing Survey Institute,Chongqing 400020,China)

机构地区：[1]重庆邮电大学通信与信息工程学院,重庆400065 [2]重庆邮电大学重庆高校市级光通信与网络重点实验室,重庆400065 [3]重庆邮电大学泛在感知与互联重庆市重点实验室,重庆400065 [4]重庆市规划局,重庆401121 [5]重庆市勘测院,重庆400020

出　　处：《计算机工程与设计》2020年第5期1379-1384,共6页Computer Engineering and Design

基　　金：重庆高校创新团队建设计划基金项目(CXTDX201601020)。

摘　　要：针对高分遥感影像中存在地物数目多,特征信息复杂导致分割边缘不清晰、对象细节丢失等问题,提出一种改进的超像素分割和多特征结合的遥感影像分割合并算法。在对图像进行分割前的预处理阶段,使用超像素分割技术得到初始分割图像;区域合并过程中,基于对象间的异质性和对象内部的同质性,结合光谱、纹理和形状特征,对对象进行合并;通过调整全局分割参数来调整合并尺度,得到最终的影像分割结果。实验结果表明,所提方法能得到较好的影像分割效果。For the high score remote sensing image,there are many features,and the feature information is complicated,resulting in unclear edge and loss of object details.An improved superpixel segmentation and multi-feature combination remote sen-sing image segmentation and merging algorithm was proposed.In the pre-processing stage,the initial segmentation image was obtained using the superpixel segmentation technique.In the process of region merging,objects were merged based on the hete-rogeneity between objects and the homogeneity within the object,combined with spectral,texture and shape features.The merged scale was adjusted by adjusting the global segmentation parameters to obtain the final image segmentation result.Experimental results show that the proposed method can get better image segmentation effects.

关 键 词：多特征 超像素 异质性 合并策略 影像分割 

分 类 号：TN911.73[电子电信—通信与信息系统]