摘要
高光谱影像分析是当前激光图像处理领域中的研究重点,高光谱影像具有海量性,影响后续高光谱影像处理的效率,为了提高海量高光谱影像分类结果,以获得最优特征为目标,提出了大数据分析技术的海量高光谱影像特征选择方法。首先采集海量高光谱影像,搭建大数据分析平台,将海量高光谱影像划分为多个小规模的高光谱影像集,并采用分布式计算技术提取高光谱影像特征,然后引入核主成分分析算法选择高光谱影像特征,最后根据选择的特征集合进行高光谱影像分类,并进行相关的仿真测试。结果表明,本方法可以在短时间选择最优的高光谱影像特征,可以保持高光谱影像分类正确率超过90%,而且高光谱影像特征选择效率高,可以满足现代高光谱影像发展的要求。
Hyperspectral image analysis is the research focus in the field of laser image processing. Hyperspectral image has a large number of features,which affects the efficiency of subsequent hyperspectral image processing. In order to improve the classification results of massive hyperspectral images and obtain the optimal features,a feature selection method of massive hyperspectral images based on big data analysis technology is proposed. Firstly,the massive hyperspectral images are collected,and the big data analysis platform is built. The massive hyperspectral images are divided into several small-scale hyperspectral image sets,and the hyperspectral image features are extracted by using distributed computing technology. Then,the hyperspectral image features are selected by the kernel principal component analysis algorithm. Finally,the hyperspectral images are classified according to the selected feature sets,and the relevant simulation is carried out True test. The results show that this method can select the best hyperspectral image features in a short time,and can maintain the classification accuracy of more than 90%. Moreover,the hyperspectral image feature selection efficiency is high,which can meet the requirements of modern hyperspectral image development.
作者
谭荣华
王俊
TAN Ronghua;WANG Jun(Yuzhang Normal University,Nanchang 330103,China)
出处
《激光杂志》
CAS
北大核心
2021年第9期90-93,共4页
Laser Journal
基金
江西省教育厅科学技术研究课题重点项目(No.GJJ171187)。
关键词
高光谱影像
海量变化特点
高维特征
大数据分析技术
核主成分分析
hyperspectral image
massive change characteristics
high dimensional features
big data analysis technology
kernel principal component analysis
