Translator Disclaimer
22 August 2020 Drone patrol using thermal imaging for object detection
Author Affiliations +
Over the past decades, unmanned aerial vehicle (UAV) has been developing rapidly. The UAV can fly above most open area easily and understand the situation immediately. In this study, a restricted area that need to be controlled for 24 hours, the use of UAV with the thermal camera can be more efficient. This study applies thermal sensor fusion of an UAV for path planning, obstacle avoidance, and image processing for outdoor patrol. An UAV can follow a predefined path and search human target by thermal camera. Once the target is found, the UAV hovers above the object. Simultaneously, the location will be sent back to the control center. In the environment with known obstacles, the study uses the A* algorithm for path planning. For unknown obstacles, the UAV utilizes the depth camera and ultrasonic module to detect the obstacles. The FLIR DUO-R thermal camera is used to detect surroundings and checks whether if there is any desired target. About the image processing, the deep learning algorithm YOLOv3 is applied to identify human shape. Success detection rates of single object and multiple objects are both high accuracy.
© (2020) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Jih-Gau Juang, Guan-Ting Tu, Yu-Hsien Liao, Tsung-Hsien Huang, and Sheng-I Chang "Drone patrol using thermal imaging for object detection", Proc. SPIE 11503, Infrared Sensors, Devices, and Applications X, 115030U (22 August 2020);

Back to Top