University of Twente Student Theses

Login

Creating 3D images for facial recognition using the RealSense SR300

Manterola, Nahuel (2017) Creating 3D images for facial recognition using the RealSense SR300.

[img] PDF
4MB
Abstract:3D face recognition is becoming a viable alternative to traditional 2D face recognition. It is more robust in terms of changed lighting and rotation of the face. The RealSense SR300 is a depth camera which makes use of structured light. It acquires the depth data using an infrared projector and camera. This paper presents a method to create a 3D point cloud of a face using this camera. Several settings in the software environment are treated to obtain an image with the best quality. Measurements are done on the depth resolution and the per-pixel depth deviation. A Face verification algorithm is used to assess the quality of the generated point cloud and determine a relation between False Accept Rate and Verification Rate using two depth images of 22 subjects. The effects of lighting conditions and the distance between a subject and the camera are tested. The depth resolution is very dependent on the distance between subject and camera and is optimal at 25 cm. The perpixel depth deviation is deemed too small to have a large impact at close distances. At larger distances averaging of multiple frames is suggested. Using the verification algorithm, a VR of 0.991 is found at FAR=0. At FAR=0.05 the VR = 1. The camera performs very poorly in outside conditions but can achieve acceptable results when increasing the projector brightness. Inside, the best results are achieved with a lower projector brightness. The maximum distance between the camera and the subject is 40 cm. Beyond this length, the face model becomes unreliable in terms of verification. This method enables the RealSense SR300 to be used as a reliable, cheap face recognition system for any indoor applications. As this camera comes built into various laptops and tablets, it can be used for various security applications. Several potential recommendations for further research are given to increase the performance of the camera.
Item Type:Essay (Bachelor)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:50 technical science in general, 53 electrotechnology, 54 computer science
Programme:Electrical Engineering BSc (56953)
Link to this item:https://purl.utwente.nl/essays/75772
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page