Drone design and simulation

With developing technology, the use of drones, which are unmanned aerial vehicles, is rapidly increasing. Drone technology has a rich development both in physical structure and hardware depending on the area where it will be used.On the other hand, artificial intelligence technology is also developing rapidly. Many studies have been done in both technologies. The hardware technology used along with the studies carried out in artificial intelligence technology has also become very advanced. One of the results of this development has been the development of supervised learning algorithms. Using advanced hardware and algorithms in the field of computer vision, open access visual datasets such as COCO and PASCAL, which contain image information of many different structured classes, have been created. However, these data sets do not contain images taken at a certain height needed for image processing by drone aircraft. In this thesis study, object recognition is aimed using real-time image processing methods with drone aircraft. For this purpose, a data set consisting of image information obtained using drone aircraft was created. This dataset was created and trained with YOLOv3, a convolutional neural network with real time image processing speed. With this developed application, image information from the drone aircraft was processed using the YOLOv3 Model, and real-time object recognition was made in the "motor vehicle" and "human" classes in the data set.