Estimation of UAV Count Using Thermal Imaging and Lightweight CNN

Illegal and improper use of UAVs has damaged public property and challenged the safety and security of the civilian population. Due to their small form factor, UAVs are undetectable using conventional aircraft detecting methods. In this work, we have addressed this issue by utilizing thermal images to detect and estimate the UAV count in a multi-UAV setting. Thermal imaging-based detection provides a number of advantages, including night vision, temperature sensitivity, low visibility, camouflage penetration, and non-invasiveness. It is a non-contact and non-intrusive detecting method that can detect hidden objects or people even in low-visibility environments such as smoke and fog. Experiments were carried out by capturing thermal images of a multi-UAV setting, where an arbitrary number of UAVs are flying in a random manner. Further, a UAV-thermal dataset is also developed so as to facilitate further research. Extensive experiments were carried out and the reported results show that the proposed model accurately estimates the number of UAVs with an accuracy 99.9%.