Course

Syllabus

Hyperspectral Image Cube Formation Multispectral vs Hyperspectral Imaging Electromagnetic Spectrum Sensor Architectures Radiometric and Geometric Calibration Noise Sources Hyperspectral image Acquisition using FS23 and ADC Considerations for Embedded Systems.Radiometric Normalization Spectral Filtering Spectral Signature Extraction Dimensionality Reduction: PCA, ICA, MNF Band Selection vs Band Extraction Feature Visualization Embedded Optimization of Preprocessing Pipelines.Spectral mapping Pixel Purity Index (PPI) Minimum Noise Fraction (MNF) Mixture Tuned Matched Filtering (MTMF) Review of Classification Techniques: Supervised Unsupervised Hybrid – Quantification in Hyperspectral Images Using Classical Least Squares Models – Spectral Angle Mapper (SAM) Accuracy Assessment and Performance Metrics Case Studies in Agriculture and Remote Sensing.

Text Books / References

1. Chein-I Chang, “Hyperspectral Data Exploitation: Theory and Applications”, Wiley, First Edition, 2007.
2. Da-Wen Sun, “Hyperspectral Imaging for Food Quality Analysis and Control”, Academic Press, First Edition, 2010.
3. John A. Richards and Xiuping Jia, “Remote Sensing Digital Image Analysis: An Introduction”, Springer, Fifth Edition, 2013.
4. Chein-I Chang, “Hyperspectral Imaging: Techniques for Spectral Detection and Classification”, Springer, Second Edition, 2015.
5. Sudeep Jayasumana, “Machine Learning for Hyperspectral Data Processing”, Springer, First Edition, 2021.

Pre-requisite: Nil

Course Objectives:

• To introduce the fundamentals of hyperspectral imaging, sensor technology, and the structure of hyperspectral data cubes.
• To impart knowledge of preprocessing techniques and dimensionality reduction methods for effective hyperspectral data handling.
• To enable students to understand and implement various spectral classification algorithms suitable for real-time decision-making.
• To develop skills for deploying hyperspectral data processing techniques on low-power embedded systems for field applications.

Course Outcomes:

• CO1: Explain the fundamentals of hyperspectral imaging and data cube characteristics.
• CO2: Apply preprocessing and dimensionality reduction techniques to enhance data interpretability.
• CO3: Implement spectral classification algorithms and evaluate their performance on real-world data.
• CO4: Design and optimize hyperspectral data processing pipelines for deployment on embedded systems.

CO-PO Mapping: