ABSTRACT
Footage that is captured through unmanned aerial vehicles (UAVs) provides the sufficient information for identification of a person. This idea gives limelight in the direction of surveillance with the help of drones. Considering the areas which are inaccessible for humans in an easy way, UAVs or so-called drones can make a great impact by making it easy for humans to reach those particular areas. In this chapter, two algorithms for face detections are used which are Viola Jones and Tine Face. For face recognition, Histogram of Oriented Gradients (HOG), Local Binary Pattern (LBP),VGG-Face and Commercial-Off-The-Shelf System (COTS) are used. According to the result, tiny faces give better precision and recall as compared to viola jones. This result is justifiable because the tiny face detector is trained for low-resolution data which increases its performance. Viola Jones does not give good precision and recall because it needs a proper frontal view of the face. For face recognition, VGG-Face algorithm gave the best results with 14.36% in identification performance. The LBP model gave the accuracy of 5.08%; on the other hand, 5.08% accuracy was achieved with the HOG algorithm. The COTS algorithm gave the worst results in comparison to others by giving the accuracy of less than 7%. Current face recognition techniques do not perform well in the given data set.
