DSD Special Session

Scope

Hyperspectral imaging (HSI), also referred to as imaging spectroscopy, integrates conventional imaging and spectroscopy methods to obtain both spatial and spectral information of a scene. Unlike conventional RGB (red, green and blue) image, which only captures three diffuse Gaussian spectral bands in the visible spectrum (e.g., 380 – 740 nm), HSI increases the amount of data acquired beyond the capabilities of the human eye. Hyperspectral (HS) sensors measure the aggregate signal of reflected, absorbed and emitted radiance at specific wavelengths of the material that is being observed. These sensors are capable of capturing a very large number of contiguous spectral bands (also called spectral wavelengths or spectral channels) across the electromagnetic spectrum, obtaining a vector of radiance values for each pixel of the image that is commonly called spectral signature. Image processing algorithms make use of these spectral signatures to automatically differentiate the materials observed by the sensor at each pixel. These methods rely on the basis that different molecular compositions of each material present in the nature has different responses to the incident light.

HSI is a promising non-invasive and non-ionizing technique that supports rapid acquisition and analysis of diagnostic information in several fields, such as remote sensing, archeology, drug identification, forensics, defense and security, agriculture, food safety inspection and control, among many others. However, in general, the algorithms to process such types of images have high computational requirements to achieve real-time processing, being necessary in many cases to employ high performance computing platforms.

 

Topics

The following topics of interest are included, but not limited to:

  • Algorithms and methods for HS data processing.
  • Systems and architectures for real-time HS data processing.
  • Different applications of HSI (health, smart farming, remote sensing, food quality, archeology, etc.).
  • Artificial intelligence methods for HS data processing.
  • Low-power implementations of HS architectures.

Special Session Chairs

Technical Program Committee

TBD