Share Email Print
cover

Proceedings Paper

Minimal entropy reconstructions of thermal images for emissivity correction
Author(s): Lloyd G. Allred
Format Member Price Non-Member Price
PDF $14.40 $18.00
cover GOOD NEWS! Your organization subscribes to the SPIE Digital Library. You may be able to download this paper for free. Check Access

Paper Abstract

Low emissivity with corresponding low thermal emission is a problem which has long afflicted infrared thermography. The problem is aggravated by reflected thermal energy which increases as the emissivity decreases, thus reducing the net signal-to-noise ratio, which degrades the resulting temperature reconstructions. Additional errors are introduced from the traditional emissivity-correction approaches, wherein one attempts to correct for emissivity either using thermocouples or using one or more baseline images, collected at known temperatures. These corrections are numerically equivalent to image differencing. Errors in the baseline images are therefore additive, causing the resulting measurement error to either double or triple. The practical application of thermal imagery usually entails coating the objective surface to increase the emissivity to a uniform and repeatable value. While the author recommends that the thermographer still adhere to this practice, he has devised a minimal entropy reconstructions which not only correct for emissivity variations, but also corrects for variations in sensor response, using the baseline images at known temperatures to correct for these values. The minimal energy reconstruction is actually based on a modified Hopfield neural network which finds the resulting image which best explains the observed data and baseline data, having minimal entropy change between adjacent pixels. The autocorrelation of temperatures between adjacent pixels is a feature of most close-up thermal images. A surprising result from transient heating data indicates that the resulting corrected thermal images have less measurement error and are closer to the situational truth than the original data.

Paper Details

Date Published: 19 March 1999
PDF: 8 pages
Proc. SPIE 3700, Thermosense XXI, (19 March 1999); doi: 10.1117/12.342310
Show Author Affiliations
Lloyd G. Allred, U.S. Air Force (United States)


Published in SPIE Proceedings Vol. 3700:
Thermosense XXI
Dennis H. LeMieux; John R. Snell, Editor(s)

© SPIE. Terms of Use
Back to Top