Drone Based Face Recognition System using MTCNN and Facenet in ArduPilot Software Platform

Abstract

The security and surveillance industries have seen significant changes as a result of drones, often known as unmanned aerial vehicles, or UAVs. In recent years, integrating facial recognition technology with drones has emerged as the best way to improve real-time identifying capabilities. A crucial field of study in computer vision and artificial intelligence is face recognition. In this paper, we have used MTCNN (Multi-task Cascaded Convolutional Networks) and Facenet for face recognition. Additionally, we compare the performance of the MTCNN method using an existing HOG (Histogram Of Gradient) method. For the
simulation of the drone based face recognition system, we have used the ArduPilot software platform. Tools used for simulation purpose are Dronekit, Mavproxy and Mission Planner. The comparison results sheds some light on the algorithm’s adaptability, accuracy, and detection rates as well as its resilience.