Improving execution time for supervised sparse representation classification of hyperspectral images using the Moore-Penrose pseudoinverse

被引:5
作者
Arias, Fernando X. [1 ,2 ]
Sierra, Heidy [1 ,3 ]
Arzuaga, Emmanuel [1 ,2 ,3 ]
机构
[1] Lab Appl Remote Sensing Imaging & Photon, Mayaguez, PR 00681 USA
[2] Univ Puerto Rico, Dept Elect & Comp Engn, Mayaguez, PR 00682 USA
[3] Univ Puerto Rico, Dept Comp Sci & Engn, Mayaguez, PR 00682 USA
基金
美国国家科学基金会;
关键词
sparse representation classification; hyperspectral imaging; remote sensing; sparse signal processing; compressive sensing;
D O I
10.1117/1.JRS.13.026512
中图分类号
X [环境科学、安全科学];
学科分类号
08 ; 0830 ;
摘要
Hyperspectral images (HSIs) contain spectral information on the order of hundreds of different wavelengths, providing information beyond the visible range. Such spectral sensitivity is often used for the classification of objects of interest within a spatial scene in fields, such as studies of the atmosphere, vegetation and agriculture, and coastal environments. The classification task involves the processing of high-dimensional data which fuels the need for efficient algorithms that better use computational resources. Classification algorithms based on sparse representation classification perform classification with high accuracy by incorporating all the relevant information of a given scene in a sparse domain. However, such an approach requires solving a computationally expensive optimization problem with time complexity Omega(n(2)). We propose a method that approximates the least squares solution of the sparse representation classification problem for HSIs using the Moore-Penrose pseudoinverse. The resulting time complexity of this approach reduces to O(n(2)). The impact on the classification accuracy and execution time is compared to the state-of-the-art methods for three varied datasets. Our experimental results show that it is possible to obtain comparable classification performance current methods, with as much as 82% of a reduction in execution time, opening the door for the adoption of this technology in scenarios where classification of high-dimensional data is time critical. (C) 2019 Society of Photo-Optical Instrumentation Engineers (SPIE)
引用
收藏
页数:15
相关论文
共 27 条
[1]  
Aggarwal CC, 2001, LECT NOTES COMPUT SC, V1973, P420
[2]  
[Anonymous], 2005, Signal theory methods in multispectral remote sensing
[3]  
[Anonymous], 1996, NUMERICAL METHODS LE
[4]  
[Anonymous], 2012, MATRIX COMPUTATIONS
[5]  
[Anonymous], 1994, INTRO CONJUGATE GRAD
[6]  
[Anonymous], 2015, 220 BAND AVIRIS HYPE
[7]  
Arias F. X., 2017, INT C PATT REC SYST
[8]  
Chang CC, 2003, ACAD STUD ASIAN ECON, P1
[9]   Hyperspectral Image Classification Using Dictionary-Based Sparse Representation [J].
Chen, Yi ;
Nasrabadi, Nasser M. ;
Tran, Trac D. .
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, 2011, 49 (10) :3973-3985
[10]   Deep Feature Extraction and Classification of Hyperspectral Images Based on Convolutional Neural Networks [J].
Chen, Yushi ;
Jiang, Hanlu ;
Li, Chunyang ;
Jia, Xiuping ;
Ghamisi, Pedram .
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, 2016, 54 (10) :6232-6251