Abstract:
Thermal cameras have an advantage in feature extraction, namely the ability
to show only objects with/according to temperature; for example, a person
with a higher temperature than the immediate surroundings could be
extracted from the background with simple machine vision techniques. The
problem with industrial thermal cameras for security use is that they are very
expensive. A solution would be to obtain less expensive security thermal
cameras. However, a problem inherent to these cameras is their automatic
self-calibration system. This is problematic in fixed constant comparator
programs.
This paper proposes a calibration system to regulate the optimal exposure
parameter for the analysis of a single, possibly under- or over-exposed image.
This calibration system provides an optimal exposure reference for the
camera, based on image clarity and reduced image noise. The phenomenon
of image noise is caused by under- or over-exposed images. To estimate the
exposure quality in the presence of saturated and unsaturated pixels, a
temperature-controlled surface is introduced into the camera's field of view.
The camera's calibrating point is aimed at this surface; therefore, this is a
reference temperature for the camera. Experimental results are presented
comparing different reference temperatures to target visibility. The experiment
was conducted in a controlled environment.
