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

Abstract

Unmanned aerial vehicles (UAVs) have become key platforms for surveillance, reconnaissance, and autonomous navigation. This paper presents a detailed AI-based framework for aerial image object detection and classification, intended for autonomous UAV navigation and control systems. The proposed approach combines recent deep learning models, namely YOLO-based detectors and attention-enhanced convolutional neural networks, to achieve real-time aerial object detection in complex scenarios. We evaluate our method on three publicly available datasets: the UAV Small Object Detection dataset, the UAV Detection dataset and the Drone Dataset (UAV) from DatasetNinja. Experiments indicate that the framework reaches a mean average precision (mAP) of 94.7% at a processing speed of about 45 frames per second, which is suitable for real-time UAV navigation. The system includes adaptive control components that convert detection outputs into navigation commands and provide autonomous obstacle avoidance and path planning.