Design and implementation of robotic vision for face recognition

Robotic vision is an ideal sensor for many robot platforms. By making robots perform adverse tasks, engineers today are working towards the eternal goal of bringing robots closer to human life. One such task is recognition or authentication of a person which is essential in both social and industrial domains. Upon finding a face, the face recognition robot either recognizes it to be one from the database, or in case of a new person, adds it to the database. Indeed, there are a number of existing algorithms that have been used to achieve this goal. For vision-based autonomous robots in dynamic domain it is crucial that the processing algorithms are fast in addition to being robust. This paper compares the efficiency of three algorithms - Eigenfaces, Fisherfaces and Local Binary Patterns Histograms. It also compares the implementation of these algorithms on a Raspberry Pi against that on a PC. Empirical results demonstrating the robotic platform performing face recognition under various circumstances, justify the validity of the proposed design of a face recognition robot.