People detection using fused hyperspectral and thermal imagery

摘要

Hyperspectral imaging is used in various applications to identify and classify the materials present in a scene. Often the requirement is to detect a specific target material, a single style of red shirt for example, and spectral processing selects only that singular object. This paper explains a novel new approach to identifying much larger and more spectrally varied classes of objects, using the specific example of 'people', based on a series of spectral matched filter style algorithms. This new approach has successfully been demonstrated against a range of targets in various scenarios alongside current thermal imaging techniques for person identification.