Abstract

Sparse representation classification (SRC) is being widely investigated on hyperspectral images (HSI). For SRC methods to achieve high classification performance, not only is the development of sparse representation models essential, the designing and learning of quality dictionaries also plays an important role. That is, a redundant dictionary with well-designated atoms is required in order to ensure low reconstruction error, high discriminative power, and stable sparsity. In this paper, we propose a new method to learn such dictionaries for HSI classification. We borrow the concept of joint sparse model (JSM) from SRC to dictionary learning. JSM assumes local smoothness and joint sparsity and was initially proposed for classification of HSI. We leverage JSM to develop an extension of discriminative K-SVD for learning a promising discriminative dictionary for HSI. Through a semi-supervised strategy, the new dictionary learning method, termed JSM-DKSVD, utilises all spectrums over the local neighbourhoods of labelled training pixels for discriminative dictionary learning. It can produce a redundant dictionary with rich spectral and spatial information as well as high discriminative power. The learned dictionary can then be compatibly used in conjunction with the established SRC methods, and can significantly improve their performance for HSI classification.

Highlights

  • Sparse representation classification (SRC), proposed in [1], is being widely investigated on hyperspectral images (HSI)

  • The main contribution of this research is that we introduce the structure information around a limited number of training pixels into the dictionary learning for HSI, establish a new discriminative optimisation function to jointly model the enriched information, and develop a joint sparse model (JSM)-constrained discriminative K-SVD (D-KSVD) algorithm to solve the optimisation problem and produce a desired discriminative dictionary

  • Dictionaries acquired from Draw, D-KSVD and the proposed JSM-DKSVD are used with three different SRC methods: 1) Sparse Model (SM), 2) JSM [2], and 3) non local weighting (NLW) [4]

Read more

Summary

Introduction

Sparse representation classification (SRC), proposed in [1], is being widely investigated on hyperspectral images (HSI). Wang et al follow the same TDDL framework and introduce a more explicitly formulated semi-supervised problem to the compact dictionary learning [19] In this context, we believe that, in order to develop a dictionary with high discriminative power for HSI classification but from only a limited number of labelled training samples, it is a promising direction to utilise the structure information as much as possible. We are highly impressed by the recent progress in HSI classification made by the JSM-based algorithms from its leveraging both spectral and spatial information in the representation of HSI pixels All these factors inspire us to develop a new dictionary learning approach for HSI classification, by enforcing the JSM assumption, of local smoothness and joint sparsity around the limited number of training sample, into D-KSVD through a semi-supervised fashion. The main contribution of this research is that we introduce the structure information around a limited number of training pixels into the dictionary learning for HSI, establish a new discriminative optimisation function to jointly model the enriched information, and develop a JSM-constrained D-KSVD algorithm to solve the optimisation problem and produce a desired discriminative dictionary

Joint sparse models for HSI classification
Discriminative dictionary learning algorithms
Classification approach
JSM-DKSVD
Motivation of JSM-DKSVD
Formulation of JSM-DKSVD
Algorithm of JSM-DKSVD
Iterative updating - sparse coding stage
Iterative updating – dictionary updating stage
Classification approach of JSM-DKSVD
Experimental studies
Parameter settings
AVIRIS dataset
ROSIS dataset
Findings
Discussion
Conclusion
Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call