Synthetic aerial image generation and runway segmentation

Abstract

Vision assisted navigation is an active area of research to assist pilots during bad weather conditions. However, these systems are not completely accurate. We propose 3D models to synthesize accurate 2D representations of the airport and the runway. A synthesized image sequence obtained from a 3D model of a view would be effective in conveying 3-D characteristics of the vanishing point (intersection between the horizon line and runway axis) and the beginning of the runway to the pilot. This can help to improve the pilot’s visual perception of the surroundings under adverse weather conditions, leading to a safer landing. We propose a system to generate 2D images of runway captured during takeoff and landing to provide better tracking. We analyze the results by segmenting the runway and comparing it with actual data captured by the aircraft.