Joint Collaborative Representation With Multitask Learning for Hyperspectral Image Classification

• Published in: IEEE transactions on geoscience and remote sensing Vol. 52; no. 9; pp. 5923 - 5936
• Main Authors: Li, Jiayi, Zhang, Hongyan, Zhang, Liangpei, Huang, Xin, Zhang, Lefei
• Format: Journal Article
• Language: English
• Published: New York IEEE 01.09.2014; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Classification; Classifiers; Dictionaries; Feature extraction; hyperspectral imagery (HSI); Hyperspectral imaging; Image classification; joint collaborative representation (CR) model; Joints; Learning; Mathematical analysis; multitask learning; Representations; sparse representation; State of the art; Training; Vectors; Vectors (mathematics); hyperspectral imagery (HSI); multitask learning; Classification; joint collaborative representation (CR) model; sparse representation
• ISSN: 0196-2892; 1558-0644
• DOI: 10.1109/TGRS.2013.2293732

In this paper, we propose a joint collaborative representation (CR) classification method with multitask learning for hyperspectral imagery. The proposed approach consists of the following aspects. First, several complementary features are extracted from the hyperspectral image. We next apply these features into the unified multitask-learning-based CR framework to acquire a representation vector and adaptive weight for each feature. Finally, the contextual neighborhood information of the image is incorporated into each feature to further improve the classification performance. The experimental results suggest that the proposed algorithm obtains a competitive performance and outperforms other state-of-the-art regression-based classifiers and the classical support vector machine classifier.