AI-Based Aerial Image Object Detection and Classification for Autonomous UAV Navigation and Control

Abstract

The unmanned aerial vehicles (UAVS) have become the key platforms of surveillance, reconnaissance, and autonomous navigation. The given paper provides a detailed AI-based concept of aerial image object recognition and classification which is specifically intended to be used in autonomous UAV navigation and control systems. the suggested approach combines the newest deep learning models, such as YOLO-based detectors and attention-controlled convolutional neural networks, to develop real-time aerial object detection in complicated scenarios. We measure our method against three publicly available datasets: The UAV small object detection dataset, the UAV detection dataset and the drone dataset (UAV) of data set ninja. experiments indicate that the described framework can reach the mean average precision (map) of 94.7% and the processing time of more than 45 frames per second, which is appropriate to realize real-time UAV navigation. The system has adaptive control systems that process the outputs of detection into navigation commands, and they have autonomous capabilities of obstacle avoiding and path planning.