An Efficient Infravision Thermal-Infrared Human Detection System for Low-Visibility Surveillance

Current detection systems struggle to provide real-time, accurate, and location-aware human detection in environments where illumination is poor or visibility is obstructed. Thermal imaging technology has become one of the prominent imaging technologies to capture and visualize the thermal radiation emitted from the body of living beings irrespective of visibility conditions. This work discusses the design and implementation of a low-cost, small and portable thermal imaging based human detection system to assist in low visibility environments such as wildlife conservation areas, harsh search and rescue zones, etc. The system integrates Raspberry Pi 4B with FLIR Lepton 3.5 thermal camera and Neo-6M GPS module, powered by YOLOV3 model to perform human detection on captured images. Our newly trained YOLOv3 model demonstrated good performance, achieving an Average Precision at IoU of 0.5 (AP50) of 0.9404 and a recall of 0.6974. This small and portable system can be easily mounted on a UAV or UGV to support a wide range of crucial and challenging surveillance tasks such as anti-poaching. A Flutter-based mobile application is also developed for real time alert notification with GPS coordinates and images captured, enabling rapid and informed decision making. Hence, the proposed holistic approach of efficient infravision human detection based on thermal imaging significantly enhances security and is directly applicable to crucial domains including anti-poaching, search and rescue and disaster management.