Multitarget Real-Time Tracking Algorithm for UAV IoT

Unmanned aerial vehicles (UAVs) have increased the convenience of urban life. Representing recent rapid development drone technology, UAVs been widely used in fifth-generation (5G) cellular networks and Internet Things (IoT), such as photography, express delivery, traffic supervision. However, owing to low altitude speed, drones can only limitedly monitor detect small target objects, resulting frequent intrusion collision. Traditional methods monitoring safety are mostly expensive difficult implement. In smart city construction, a large number IoT cameras connected 5G installed city. Captured images transmitted cloud via high-speed low-latency network, machine learning algorithms for detection tracking. this study, we propose method real-time tracking targets by using existing network obtain real time employing deep which environments be guided. To achieve UAV targets, employed tracking-by-detection mode learning, with network-modified YOLOv3 (you look once v3) detector Deep SORT correlation algorithm. We established dataset that contains four types 2800 pictures different environments. The model trained achieved 94.4% accuracy speed 54 FPS. These results comprehensively demonstrate our achieves high-precision at reduced cost.