Scholars Central

ArticlePublished 22 Nov 2024

Enhancing Hyperspectral Image Resolution with Deep Transfer Learning: A Case Study on Model Deployment and Evaluation

Authors: Nafis Khan, Siddharth Misra, James Omeke, Priyanka Chukka, Siddhanth Sirgapoor

Journal of Remote Sensing & GIS

Volume 13 · Issue 4 · Pages 1–14

Ask about this article

Abstract

Select any words in the title or abstract, or press this and choose a sentence.

Hyperspectral imaging holds immense potential for detailed land cover analysis due to its rich spectral information across the electromagnetic spectrum. However, the inherent trade-off between spatial and spectral resolution limits its applicability. The study explores the deployment and enhancement of Single Hyperspectral Image Super-Resolution (SSPSR) model, using the Spatial-Spectral Prior Network (SSPN), to notably improve the spatial and spectral quality of hyperspectral images. This model stands out for its ability to elevate image resolution without relying on supplementary hardware enhancements. Our research focused on training this model using the comprehensive Chikusei dataset, this dataset features 128 spectral bands from 363 nm to 1018 nm, captured over Chikusei, Japan, enhancing deep learning research in agricultural and urban land cover analysis, followed by fine-tuning with the Cave dataset, renowned for its hyperspectral indoor images across numerous spectral bands, to ensure adaptability to diverse real-world scenarios. A refined loss function was introduced to enhance image fidelity, spatial smoothness, spectral fidelity and perceptual quality. Through meticulous hyper parameter tuning and leveraging of public datasets, we significantly improved upon existing methodologies, as evidenced by quantitative performance measures. The fine-tuned SSPSR model achieved noteworthy gains in Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index (SSIM); thus, contributing to the precision and efficiency of hyperspectral image analysis. Our results demonstrate the model’s enhanced performance in generating high-resolution images from low-resolution data, promising substantial advancements in remote sensing and land mapping applications.

Ask about this article

Answers come only from this article's title, abstract and keywords, each sentence pointing to where it comes from. Your question is sent to OpenAI to write the answer. Scholars Central does not store it.

Try:

More from this journal

The newest in Journal of Remote Sensing & GIS.

All its articles

Conferences in this field

Upcoming conferences in this field.

Follow

New articles from what you follow go into your email digest, once you agree to alert emails.

  • JournalJournal of Remote Sensing & GIS
  • SubjectGeology and Earth Science
  • BylineArticles whose byline prints “Nafis Khan”
  • BylineArticles whose byline prints “Siddharth Misra”
  • BylineArticles whose byline prints “James Omeke”
  • BylineArticles whose byline prints “Priyanka Chukka”
  • BylineArticles whose byline prints “Siddhanth Sirgapoor”

For authors and organisers

Publish, present and keep up with your field

Find the right journal for your paper and submit your manuscript.

Photo: Pexels