Mathematical Methods for Video Registration in Autonomous Navigation Applications

"A system and the computational framework is proposed for the real-time video processing of single or multiple video camera feeds from the UAV platform, for the purpose of a computer aided ("pilot assistant") or the fully autonomous planning and executionof aircraft approach and landing procedures. While the system does not require any additional positioning instrumentation on board (e.g. GPS, IMU, etc.), the availibility of such data can be utilized effectively to increase the geometrical accuracy and thespeed of computation. The critical components of the system are Cognitech's fast contrast-invariant projective registration, a runway and the horizon extraction based on the recently developed theory of meaningful alignments, fully automatic cameracalibration in the presence of strong distortions, the non-linear multi-frame photogrammetry (videogrammetry) for an accurate geometrical estimation of runway vs. aircraft positioning and flight-path planning. The validation and calibration procedures arespecified for experimental verification of the accuracy of the geometric parameters estimation. Advanced automation for aircraft landing, collison avoidance, aerial traffic management systems, utilization of refurbished military aircraft for unmannedtasks. Surveying automation for geographical information systems."