Edge AI Empowered UAV with Onboard Semantic Extraction for Monitoring Applications

UAV platforms can provide new services on the move that go beyond pure communication. This work integrates remote sensing, embedded processing, and 5G connectivity on a UAV platform as an end-to-end intelligent monitoring solution. The system leverages RGB and multispectral (MS) sensors to execute AI tasks (i.e., object detection, MS-RGB fusion, semantic segmentation, and spectral index extraction) directly onboard. Processed telemetry and AI outputs feed a high-fidelity Digital Twin for operational replication and synthetic data generation. Experimental results in the field confirm that the integrated edge AI approach significantly reduces bandwidth requirements by 5x, 8x, and 10x, respectively, for semantic segmentation, MS indexing, and object detection, thus allowing efficient monitoring even under low connectivity and power limitations.