A deep learning approach to object detection in quadcopter based reconnaissance

Abstract

This paper presents a reconnaissance system utilizing a remote-controlled quadcopter drone, designed to transmit long-range, and real-time video feeds for target object detection. The system enables continuous surveillance and rapid identification of targets in remote or high-risk areas. It employs high-resolution imaging and video capture through a camera mounted on a stabilized gimbal, along with long-range radio frequency communication to ensure high-quality video transmission over extended distances. For object detection and classification, the system integrates the CiRA CORE software, which processes video input from a universal serial bus website camera (USB webcam) quickly and efficiently. Field tests demonstrate the system's effectiveness in dynamic and changing environments, highlighting its strong capability to detect and distinguish targets. The results confirm the feasibility of reliable, real-time target detection from a long-range unmanned aerial surveillance platform.