Landing pad analysis by point cloud method in unmanned air vehicle

Today, UAVs, which have started to be used in terms of fast transportation, low cost and reducing the loss of life, continue to develop rapidly due to the increase in their usage areas. Even though they are called unmanned, these aircraft must be controlled with the help of a pilot/flight attendant. According to the researches, the majority of UAV accidents occur during landing. Therefore, the detection of the landing strip with image processing methods and the analysis of the obstacles on the landing strip with the point cloud method during the landing stage, so as to reduce these accident rates, is a big step to be taken in this regard. In this thesis, applications based on deep learning networks are included in order to understand and facilitate the UAV landing systems. Since there are different types of UAVs with different aerodynamic properties depending on their wing structures, the construction phase and the design of the landing strip for Quadrotor, which is one of these types, were carried out in this study. As a result of taking the images of the runway on which the UAV will land with the designed UAV and processing these images in YOLOv8, the UAV landing strip is detected. During the determination of the landing strip, the presence and absence of obstacles on the runway were determined. In addition, information about point cloud, image processing, deep learning networks and deep learning architectures are given and explanations are presented. As a result, during the landing of the UAVs, the runway image is processed with image processing methods and the comparisons are evaluated by analyzing it with the help of point cloud method