Deep Learning Hyperspectral Image Classification Using Multiple Class-based Denoising Autoencoders, Mixed Pixel Training Augmentation, and Morphological Operations

TL;DR

This paper introduces a novel hyperspectral image classification system combining class-based denoising autoencoders, a new data augmentation technique with pixel mixing, and morphological operations, achieving high accuracy on the Salinas dataset.

Contribution

It proposes a new hyperspectral classification framework integrating multiple autoencoders, a pixel mixing augmentation method, and morphological processing, which improves edge pixel classification.

Findings

High classification accuracy on Salinas dataset

Effective handling of mixed edge pixels

Robust image segmentation performance

Abstract

Herein, we present a system for hyperspectral image segmentation that utilizes multiple class--based denoising autoencoders which are efficiently trained. Moreover, we present a novel hyperspectral data augmentation method for labelled HSI data using linear mixtures of pixels from each class, which helps the system with edge pixels which are almost always mixed pixels. Finally, we utilize a deep neural network and morphological hole-filling to provide robust image classification. Results run on the Salinas dataset verify the high performance of the proposed algorithm.