Improving UAV autonomous landing on target using combination of FAST & SURF

Abstract

UAVs are being one of the most interesting and useful topics during past years and it has high scientific value and high potential for military and discovery issues. Many research labs around the world have been developing their UAVs for military and civilian, surveillance, construction, rescue, exploration and etc. Also, UAV usage in many factories, hospitals, hotels and industrial area is growing. The challenging task among the AI is optimization of navigating UAV in the arbitrary environment including obstacle avoidance. In order to land on a platform, there are approaches like graph based and geometric estimation. Time complexity in geometric approaches would increase when the degree of freedom (DOF) is increasing. Thus geometric approaches are useful in less DOF systems. To obtain this purpose, using several approaches based on image processing, artificial intelligence and machine learning has recommended. So that these approaches generally are evaluated with the speed and precision of finding the objects. Therefor they will guarantee finding the best way of navigation and performing precision landing on landing platform. In this thesis, we’ve presented a novel method based on combination of FAST and SURF, which is an efficient solution in order to detect the object and performing the precision landing on it. First of all, we design the system and necessary equipment, then in order to reduce the color dimension and segmentation of the image, an algorithm is presented and developed. Finally, in order to perform object detection and autonomous precision landing, the combination of algorithms is introduced. According to the results from the simulation of the purposed solution and evaluations, the purposed solution has performed better than the similar solutions. In comparison with the fastest solution, which is ORB, the processing speed improved 16.8% and in comparison, with the most precise solution, which is SIFT, the processing speed improved 82.3%.