| 融合谱聚类和多特征的遥感图像分割 |
| |
| 引用本文： | 冒,伟.融合谱聚类和多特征的遥感图像分割[J].教育技术导刊,2020,19(3):248-251. |
| |
| 作者姓名： | 冒 伟 |
| |
| 作者单位： | 上海理工大学 光电信息与计算机工程学院，上海 200093 |
| |
| 摘 要： | 为解决传统谱聚类算法在图像分割时计算量大、使用单一特征分割的局限性问题，设计一种融合谱聚类和多特征的图像分割算法。首先进行超像素分割以减少计算量，分别提取每个超像素的颜色特征和纹理特征，构建超像素相似度矩阵|然后采用特征加权方法线性融合颜色和纹理特征的超像素相似度矩阵|最后采用谱聚类算法进行聚类分割。在UCMerced_LandUse和Berkeley数据集上进行实验测试，并与现有方法进行比较。实验结果表明，大多数实验图像IOU指标均在90%以上，相比于传统方法有了显著提高。
|
| 关 键 词： | 谱聚类 超像素 图像分割 特征提取 |
| 收稿时间： | 2019-05-06 |
Fusion Spectral Clustering and Multi-feature Remote Sensing Image Segmentation |
| |
| MAO Wei.Fusion Spectral Clustering and Multi-feature Remote Sensing Image Segmentation[J].Introduction of Educational Technology,2020,19(3):248-251. |
| |
| Authors: | MAO Wei |
| |
| Institution: | School of Optical-Electrical and Computer Engineering，University of Shanghai for Science and Technology，Shanghai 200093，China |
| |
| Abstract: | In order to solve the problems like large amount of computation and segmentation limitation based on single feature of traditional spectral clustering algorithm in image segmentation， this paper design a fusion spectral clustering and feature of image segmentation algorithm， first of all to pixel division to reduce amount of calculation， respectively to extract each pixel color features and texture features， build super pixel similarity matrix， and then adopt the method of feature weighted linear fusion of color and texture feature of pixel similarity matrix， then using spectral clustering algorithm for clustering segmentation. We performed experimental tests on the ucmercedes d_landuse and Berkeley data sets and compared them with the existing methods. The experimental results showed that for most experimental images， the IOU index of our method was above 90%， which was significantly improved compared with the traditional methods since multiple features of the images were considered simultaneously. |
| |
| Keywords: | spectral clustering super pixel image segmentation feature extraction |
|
| 点击此处可从《教育技术导刊》浏览原始摘要信息 |
| 点击此处可从《教育技术导刊》下载免费的PDF全文 |
