A Comprehensive Taxonomy of UAVs: A Comparative Feature Matrix for Hardware, Capabilities, and Payloads

This survey consolidates the state of the art in vision-based Unmanned Aerial Vehicle (UAV) systems and introduces a unifying taxonomy and feature matrix that relate platform form-factors, onboard capabilities, and payload characteristics to application domains. We systematically review and classify recent literature across agriculture, transportation and logistics, infrastructure inspection, search and rescue, environmental monitoring, emergency response, surveying/mapping, and surveillance, and analyze operational trade-offs of prominent platform classes—multi-rotor, fixed-wing, and hybrid UAVs—with respect to endurance, payload capacity, maneuverability, sensing modalities (i.e., Red-Green-Blue (RGB) cameras and thermal sensors), and onboard computation. Algorithmic trends are synthesized, highlighting advances in lightweight Deep Learning (DL) techniques (notably Convolutional Neural Network (CNN)-based perception), robust state estimation via Visual-Inertial Odometry (VIO) and Simultaneous Localization and Mapping (SLAM), and multisensor fusion. Recurring challenges are identified, including limited flight endurance, resilient operation in Global Positioning System (GPS)-denied or cluttered environments, scarcity of domainspecific datasets for vision tasks, and the absence of standardized benchmarks and safety frameworks. Finally, we outline research directions to promote resilient, explainable, and energy-aware UAVs and to accelerate the transition from experimental prototypes to operational deployments.