Journal of Aerospace Information Systems
A widely used machine vision pipeline based on the Speeded-Up Robust Features feature detector was applied to the problem of identifying a runway from a universe of known runways, which was constructed using video records of 19 straight-in glidepath approaches to nine runways. The recordings studied included visible, short-wave infrared, and long-wave infrared videos in clear conditions, rain, and fog. Both daytime and nighttime runway approaches were used. High detection specificity (identification of the runway approached and rejection of the other runways in the universe) was observed in all conditions (greater than 90% Bayesian posterior probability). In the visible band, repeatability (identification of a given runway across multiple videos of it) was observed only if illumination (day versus night) was the same and approach visibility was good. Some repeatability was found across visible and shortwave sensor bands. Camera-based geolocation during aircraft landing was compared to the standard Charted Visual Approach Procedure.
Original Publication Citation
Moore, A. J., Schubert, M., Dolph, C., & Woodell, G. (2016). Machine vision identification of airport runways with visible and infrared videos. Journal of Aerospace Information Systems, 13(7), 266-277. doi:10.2514/1.i010405
Moore, Andrew J.; Schubert, Matthew; Dolph, Chester; and Woodell, Glenn, "Machine Vision Identification of Airport Runways With Visible and Infrared Videos" (2016). Electrical & Computer Engineering Faculty Publications. 176.