Combining multiscale features for classification of hyperspectral images: a sequence based kernel approach

TL;DR

This paper introduces a sequence structured kernel for hyperspectral image classification that leverages multiscale hierarchical features, demonstrating improved accuracy over traditional vector-based kernels in experiments.

Contribution

It proposes a novel sequence spectrum kernel approach that generalizes conventional stacked vector kernels for better hyperspectral image classification.

Findings

The sequence kernel outperforms traditional kernels on multiple datasets.

Hierarchical multiscale features enhance classification accuracy.

The conventional stacked vector kernel is a special case of the proposed spectrum kernel.

Abstract

Nowadays, hyperspectral image classification widely copes with spatial information to improve accuracy. One of the most popular way to integrate such information is to extract hierarchical features from a multiscale segmentation. In the classification context, the extracted features are commonly concatenated into a long vector (also called stacked vector), on which is applied a conventional vector-based machine learning technique (e.g. SVM with Gaussian kernel). In this paper, we rather propose to use a sequence structured kernel: the spectrum kernel. We show that the conventional stacked vector-based kernel is actually a special case of this kernel. Experiments conducted on various publicly available hyperspectral datasets illustrate the improvement of the proposed kernel w.r.t. conventional ones using the same hierarchical spatial features.