Vision Based Automatic Landing with Runway Identification and Tracking

Vision based autolanding of a fixed wing unmanned aerial vehicle (UAV) is presented in this paper. The UAV initially searches for a nearby unknown road or runway from the terrain image frames captured by onboard stereovision camera. The captured unknown road or runway are evaluated to be a possible candidate for autolanding of a UAV. Six degree of freedom ($6DOF$) simulink model of a UAV with landing and navigation autopilot is integrated with flightgear flight simulation software. Flightgear FG camera is used to capture image frames of the nearyby terrain while UAV is in auto navigation mode and performing a level flight with various heading angles to look out for all possible terrain scenario. Image processing techniques are applied on the captured video to identify a smooth and straight road or runway for landing and calculates the relative distance and orientation of the runway or road with respect to UAV's own location. Image processing techniques such as hough transform and random Sample consensus (RANSAC) are used to detect the road strip lines. Once the desired touchdown point on the road is computed as well as road heading is also computed, UAV begins the autolanding maneuver. The image processing techniques are used to identify the candidate road or runway. The heading of the runway and cross track of the runway is computed and autopilot is designed to bring the aircraft at desired touch down point on the road or runway.