Vision based landing for unmanned aerial vehicle

The objective of this work is to design an automated landing system for an unmanned aerial vehicle using image reference technique. The landing is aided by the comparison of the reference images previously captured and stored in the UAV memory. Each image is captured at specific parameters like time interval, angle of motion, speed, distance, altitude, image deflection angle, position etc. These parameters are of vital importance for the image comparison at real time operation capture. The images are captured at the landing phase, where the UAV glides through a specific angle for landing. At the landing phase the projection of the camera with respect to the runway must be aligned to the standard initial position, since the aircraft glides at a specific angle thereby effecting an angular change with respect to ground. This can be achieved by applying a feedback control loop using the signal response from the gyro. The UAV captures frames during its flight and compares them with the reference frames. By comparison of frames the corresponding feedback output is generated and applied to the UAV for it to align stable at the desired flight path and angle. The feedback output generation can be programmed by tools like SIFT or RANSAC algorithm using edge detection and homography techniques. Optical flow is computed either by Lukas-Kanade or Horn-Schunck method.