Distributed Compressed Hyperspectral Sensing Imaging Based on Spectral Unmixing.

Area of Science:

• Remote Sensing
• Signal Processing
• Data Compression

Background:

• Hyperspectral imagery generates massive data, requiring substantial computational resources, storage, and bandwidth.
• Compressed sensing offers a solution by reducing data acquisition, but efficient reconstruction frameworks are needed.

Purpose of the Study:

• To propose a distributed compressed sensing framework for hyperspectral imagery.
• To reduce the computational and storage costs associated with hyperspectral data acquisition.
• To enhance the reconstruction quality of hyperspectral data compared to existing methods.

Main Methods:

• A distributed compressed sensing framework separates hyperspectral data into key-band and compressed-sensing-band with varying sampling rates.
• Spectral unmixing is employed for reconstruction, involving endmember extraction from the compressed-sensing-band.
• Endmember prediction via interpolation and abundance estimation using sparse penalty, followed by linear mixing model recovery.

Main Results:

• The proposed framework effectively recovers original hyperspectral imagery from compressed measurements.
• Experimental results on real hyperspectral datasets demonstrate superior reconstruction performance.
• The reconstruction peak signal-to-noise ratio surpasses state-of-the-art methods.

Conclusions:

• The developed distributed compressed sensing framework offers an efficient approach for hyperspectral data acquisition and reconstruction.
• This method significantly alleviates the challenges posed by the large data volume of hyperspectral imagery.
• The integration of spectral unmixing provides a robust and effective reconstruction strategy.