Ground Truth Simulation for Deep Learning Classification of Mid-Resolution Venus Images Via Unmixing of High-Resolution Hyperspectral Fenix Data

TL;DR

This paper introduces a method to generate reliable ground truth data from high-resolution hyperspectral images to train deep learning models for classifying mid-resolution Venus images, overcoming data scarcity issues.

Contribution

The novel approach unmixes high-resolution hyperspectral data to simulate ground truth for training CNNs, enabling effective classification of lower-resolution Venus images.

Findings

Successful transfer of trained model to classify new Venus imagery.

Unmixing high-res hyperspectral images improves ground truth quality.

Deep learning classification performance is enhanced using simulated ground truth.

Abstract

Training a deep neural network for classification constitutes a major problem in remote sensing due to the lack of adequate field data. Acquiring high-resolution ground truth (GT) by human interpretation is both cost-ineffective and inconsistent. We propose, instead, to utilize high-resolution, hyperspectral images for solving this problem, by unmixing these images to obtain reliable GT for training a deep network. Specifically, we simulate GT from high-resolution, hyperspectral FENIX images, and use it for training a convolutional neural network (CNN) for pixel-based classification. We show how the model can be transferred successfully to classify new mid-resolution VENuS imagery.