Improving Hyperspectral Image Classification using Data Augmentation of Correlated Color Temperature

Abstract

Machines learning has a huge influence on object classification in the hyperspectral image. In order to obtain a satisfying result, machine learning needs large training data. However, a huge labelled sample for training purpose is hard to obtain. Data Augmentation (DA) is a strategy that can increase the quantity of training data and effective to overcome the limited training samples problem. On the other hand, color is one of the most important features that commonly used in object recognizing. In this study, we first explore how radiance manipulation in hyperspectral images using Correlated Color Temperature (CCT) can be used as the DA. Finally, using an ensemble method and a switching method to optimize the classification results. The experimental results demonstrate that the proposed technique can improve classification performance better than the recent feature selection technique.