Drone-Based Adaptive Hybrid Techniques for Improving Face Detection and Recognition

Abstract

With the rapid development of Unmanned Aerial Vehicle (UAV) related technologies, Drones gained the opportunity to search locations that are dangerous or difficult for humans to reach. In this regard, Drones technology has been widely used to detect and recognize humans on the ground. In order to identify their potential role in counter-terrorism operations, the aim of this research study is to design and develop a reliable aerial system capable of identifying people with a high detection rate. Furthermore, it presents a solution to all challenges that degrade the resolution when query images are taken from high altitudes, different angles, and long distances. In the context of improving the accuracy of facial detection and recognition, two adaptive techniques are proposed: The first approach is to adaptively adjust the Drone’s altitude, speed, and attitude based on two parameters: Ground sampling distance (GSD), and confidence level of the recognized face. The second approach is to adaptively adjust the resolution of images captured by drone based on the size of the detected facial area. Face detection and recognition are done using the Haar Cascade classifier and Local Binary Pattern Histogram (LBPH) algorithm. The entire system was developed, implemented, and tested using a Hexacopter Drone and onboard vision system. The testing results corroborate the system's practicality and demonstrate that the prototype may be simply implemented to track dangerous people with criminal records.