Purchase this article with an account.
Gislin Dagnelie, Avi Caspi, Michael P Barry, Paul Gibson, Greg Seifert, Arup Roy; Thermal imaging prototype enhances person identification and warm object localization by Argus II wearers. Invest. Ophthalmol. Vis. Sci. 2016;57(12):5167.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To explore the potential of thermal imaging to improve the ability of Argus II users to locate people and objects in cluttered environments.
Four Argus II wearers used either thermal or visible imagery in two test scenarios. To present thermal imagery, a Lepton 60x80 pixel thermal-IR camera (8.6 frames/s) was mounted on the Argus II glasses, and the USB output was converted to NTSC and supplied to the video processing unit (VPU) instead of the Argus II camera output. Thermal imagery was thresholded either at an absolute level or at a fraction of the thermal range available in the image. Both settings were compared to visible imagery in 2 tasks. In the Social Interaction (SI) task the subject was asked to indicate the presence of one or two assistants in the left and right locations of the scene, as well as the stance (stand, sit, absent) of either. In the Object Localization (OL) task, a cup of hot water was present in one of 3 positions (left/center/right), against 4 different backdrops. Scores and times for different cameras and settings were compared using sign tests; trials with 0 difference between matching trials were ignored. Probabilities of one image type having the observed count of higher scores or times than another, assuming no true difference, were calculated using binomial distributions with success probabilities of 0.5. P < 0.05 (one-tailed) was considered significant for scores and p < 0.025 (two-tailed) was considered significant for times.
In the SI task, all 4 subjects performed between 80% and 100% correct with thermal imagery, using one or both threshold settings; 3 subjects performed at chance with visible imagery, while the 4th performed near 70% correct, but took 3 times longer than with thermal imagery. In the OL task, all subjects were near chance with visible imagery, and 2 with the relative threshold thermal image; these 2 were near 100% correct with the absolute threshold thermal image, as were the other 2 subjects with either thermal image. All score differences and the described time difference were highly significant (p < 0.01).
These results suggest that for social interactions, as well as temperature-based object localization, Argus II wearers can make effective use of thermal imagery. Construction of a prototype with user-adjustable mixed visible-thermal camera imagery is under way.
This is an abstract that was submitted for the 2016 ARVO Annual Meeting, held in Seattle, Wash., May 1-5, 2016.
This PDF is available to Subscribers Only