|
| 基于深度极限学习机的高光谱遥感影像分类研究 |
| Hyperspectral Remote Sensing Image Classification Based on Deep Learning Extreme Learning Machine |
| 投稿时间：2017-11-29 修订日期：2017-12-21 |
| DOI： |
| 中文关键词: 高光谱遥感影像 深度学习 极限学习机 遥感影像分类 |
| 英文关键词: Hyperspectral Remote Sensing Image Deep Learning Extreme Learning Machine Remote Sensing Image Classification |
| 基金项目:国家自然科学基金项目（面上项目，重点项目，重大项目） |
|
| 摘要点击次数: 436 |
| 全文下载次数: 0 |
| 中文摘要: |
| 高光谱遥感数据越来越普及并为人们广泛使用, 基于高光谱数据的地面物体精确分类是高光谱遥感技术的核心应用之一. 针对高光谱遥感影像的分类问题, 提出一种基于深度极限学习机的分类方法. 该方法利用一种新的深度学习模型—深度极限学习机(D-ELM)对高光谱遥感影像进行分类, 并与基于极限学习机(ELM)分类、支持向量机(SVM)分类方法、基于核极限学习机(ELMK)分类方法进行了比较分析. 研究结果表明：相对于ELM、SVM、ELMK分类方法, D-ELM能够更加准确地挖掘高光谱遥感影像的空间分布规律, 提高分类的准确度. |
| 英文摘要: |
| For the classification of hyperspectral remote sensing image classification method is presented based on Extreme Learning Machine depth of the image calculated by nonsubsampled contourlet transform texture characteristics, using a new depth learning models - depth Extreme Learning Machine (D-ELM) hyperspectral remote sensing images based on spectral - texture features classification, and based on Extreme Learning Machine(ELM), based support vector machine (SVM) classification methods ,Kernal Extreme Learning Machine(ELMK) were compared. The results show that: with respect to the single-source spectral information using image spectral - texture features can effectively improve the classification accuracy of high resolution remote sensing imagery; with respect to ELM,SVM and ELMK, D-ELM can be more accurately and excavating the hyperspectral remote sensing spatial distribution of the image, improve the accuracy of classification. |
|
View Fulltext
查看/发表评论 下载PDF阅读器 |
| 关闭 |