Real time compression for push broom hyperspectral imaging

Abstract

Hyperspectral imaging is useful for remote sensing, medicine, product inspection and other applications. Unfortunately, the large size of hyperspectral data presents acquisition challenges, especially in constrained flight-deployed hardware. Real time hyperspectral data acquisition and compression software is thus developed. The compression method takes advantage of the properties of naturally occurring spectra as captured by a push broom scanning hyperspectral imager. Principal component analysis is used to represent the spectral data into fewer coefficients, hence achieving a compression ratio of up to 30:1. By utilizing multiple CPU cores, image acquisition can operate up to 150 frames per second, allowing high resolution scanning. This paper presents the compression principle and implementation and operation in an optical setup simulating flight acquisition.