基于半监督算法的高光谱影像特征提取仿真
Extract Simulation Based on High Spectrum Imaging Features Based on Semi-Supervised Algorithms
万露 1武天 1刘纬 1王宽田2
作者信息
- 1. 电子科技大学成都学院,四川 成都 611731
- 2. 桂林电子科技大学海洋工程学院,广西 北海 536000
- 折叠
摘要
高光谱影像包括待测物的空间、光谱和辐射三重信息,且图像信息具有维度高、空间相关性弱、特征非线性强的特点,导致其空间特征序列混乱,特征提取难度大.于是提出基于半监督算法的高光谱影像特征提取方法.应用半监督算法对高光谱图像中的高维数据降维处理,并基于降维结果完成高光谱图像的去模糊.高光谱图像完成降维去模糊后,根据特征学习模型学习高光谱影像数据,获取图像深层特征.在像元空间内对深度特征以及空间信息完成空、谱的联合,实现高光谱影像特征的提取.实验结果表明,所提方法应用下影像特征点在特征空间内聚类效果好,查全率和查准率均能达到 95%以上,说明上述方法的应用性能更优.
Abstract
Hyperspectral image includes the information of space,spectrum and radiation of the object to be meas-ured.In addition,the image information has the characteristics of high dimension,weak spatial correlation and strong nonlinearity,leading to the confusion of spatial feature sequence and the difficulty of feature extraction.Therefore,this paper puts forward a method of extracting hyperspectral image features based on semi-supervised algorithm.At first,the semi-supervised algorithm was adopted to reduce the dimension of high-dimensional data in hyperspectral images,and then the image deblurring was completed based on the result.After that,hyperspectral image data were learned by the feature learning model,thus obtaining the deep features of the image.In the pixel space,the depth fea-tures were combined with spatial information.Finally,the extraction of hyperspectral image features was completed.Experimental results show that the proposed method has good clustering effect of feature points in feature space.Meanwhile,the recall and precision can be more than 95%,indicating that the application performance of the method is better.
关键词
半监督算法/高光谱图像/图像去模糊/数据降维/特征提取方法Key words
Semi-supervision algorithm/Hyperspectral image/Image deblurring/Data reduction/Feature ex-traction引用本文复制引用
基金项目
四川省科普培训项目(2020JDKP0041)
四川省科技厅项目(2021JDKP0077)
出版年
2024
NETL
NSTL
万方数据