Runway detection from map, video and aircraft navigational data

Abstract

As part of the reinforcement of operations performed by the Mexican Navy, unmanned aerial vehicles (UAV) have been equipped with daylight and infrared cameras. Processing the video information obtained from these devices opens the door to a number of strategic opportunities. By recognizing patterns in visual sources, we address one problem in particular: how to achieve corresponding runways using a map and an individual frame of video. An approach to runway detection using two tools is presented in this thesis. The first tool is a geographical information program, which is used to set the runway map we want to detect. The second tool is a video frame of the same runway, recorded in a camera mounted on a UAV. The needed equations and related algorithms are developed and tested on a simulation that reconstructs the three-dimensional view of the aircraft camera employing the map and navigational data. Next, the algorithms are implemented using actual video frames. Finally, mismatches in the runway detection due to sensor noise and to an assumption made on the aircraft roll orientation angle are corrected using image-processing techniques, such as the Hough transform for linear features.