Extension of Facial Recognition Abilities of the Humanoid
Humanoid robots become increasingly important in today’s world. One such robot called Pepper was used in this thesis. To allow a more personalized interaction between the robot and humans, Peppers facial recognition abilities were extended by the usage of a computer vision library called OpenCV.
Daniel Hertig, 2018
Bachelor Thesis, University of Applied Sciences and Arts Northwestern Switzerland
Betreuende Dozierende: Achim Dannecker
Keywords: robotics, facial recognition
Views: 10 - Downloads: 0
The Institute for Information Systems at the University of Applied Sciences and Arts Northwestern Switzerland does research and runs projects with humanoid robots. The most advanced among them is called Pepper, which is produced by Softbank Robotics. Pepper is researched to support several use cases. One of the functionalities these use cases require is the detection and recognition of human faces. This thesis was assigned to allow a management of the faces that Pepper detected and recognized.
Pepper has a built-in solution that can handle the basic facial detection and recognition tasks. The management of the data used for the built-in facial detection and recognition is done on Pepper itself and no interface is provided. The main task was to extend the existing face recognition solution. Additionally, a backend system in form of a web application to manage the data collected by Pepper was to be developed. To store the data from Pepper and the backend system in a structured form, databases were to be created.
Since the initial research showed that the built-in face recognition solution is rather limited and cannot be used to fulfill the goals of this thesis, OpenCV, a well-known library for image manipulation, was used to implement an own facial recognition solution. To manage the data collected by Pepper, a web application based on the Python micro framework Flask was developed. To be able to move data from Pepper to the web application and vice versa, an interface between the databases of the two platforms was established.
Once a prototype of the solution was implemented, it was tested on the use case of identifying and greeting employees. At the beginning of the test phase, the recognition rate was rather low but increased steadily during the course of the test because Pepper was able to collect more faces to train its recognizer.
Studiengang: Business Information Technology (Bachelor)
Fachbereich der Arbeit: Business Information System & IT-Management