Electronic Science and Technology ›› 2019, Vol. 32 ›› Issue (4): 33-39.doi: 10.16180/j.cnki.issn1007-7820.2019.04.008

Previous Articles     Next Articles

Hyperspectral Image Sub-Pixel Mapping Based on Sparse Unmixing of Endmember Dictionary

GU Zhengzhi1,2,3,WANG Suyu1,2,3   

  1. 1. Beijing Advanced Innovation Center for Future Internet Technology, Beijing 100124,China
    2. Beijing Engineering Research Center for IoT Software and Systems, Beijing 100124,China
    3. Faculty of Information Technology, Beijing University of Technology, Beijing 100124,China
  • Received:2018-03-18 Online:2019-04-15 Published:2019-03-27
  • Supported by:
    National Natural Science Foundation of China(61201361);Science Foundation of the Beijing Education Commission(KM201710005011);Training Program Foundation for the Talents in Beijing City(2013D005015000008)


In general, it is difficult to locate the spatial distribution of each endmember in the hyperspectral image. To solve the problem, this study proposed a K-SVD-based spectral unmixing algorithm whose result of the demixing was further used for performing the subpixel location. First, the mixed pixel and pure pixel were distinguished by KNN classification, and then the sparse decomposition correlation theory based on redundant dictionary was used for reference. Based on the standard spectral library, the K-SVD based dictionary training algorithm was used to train the most representative material spectral curves, and the endmember redundancy dictionary was subsequently constructed. The abundances were solved by the sparse decomposition algorithm based on K-SVD. Finally, the obtained abundance coefficient was used to locate the sub-pixel under the two spatial correlation constraints. Experimental results showed that the proposed algorithm had reliable performance for the sub-pixel mapping effects of simulated data and measured data.

Key words: hyperspectral image, spectral unmixing, sub-pixel mapping, K-SVD, sparse representation, redundant dictionary

CLC Number: 

  • TP391