Development of 3D face recognition using microsoft kinect
Loading...
Date
2017-06
Authors
Loh, Qing Rong
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
3D face recognition system has become a famous research over recent years. Due to rapid population growth, automatic face recognition system is needed for a lot of security applications. This research aims to develop a working, low-cost and precise 3D face recognition system using Microsoft Kinect. By making this system available at a cheap price, the overall security level can be increased. In this project, input data preparation is done using Microsoft Kinect. The data are preprocessed into the desired form before any further process. In previous work, jawline is used as a boundary to remove outliers. However, some information may be treated as outliers and removed by the algorithm. So, a different way of outliers removing is proposed in this project and compare the performance of the system respectively. After some image processing is performed on the image collected, the next step would be matching, which included ICP algorithm. The different configuration in ICP is used and analyzed. 10 images and 10 3D data are captured for each subject. 70% are used as a database. The rest of it are test images. Every image undergoes the identical process for denoising and facial extraction. After testing, the system achieved 87.18% of accuracy by giving the correct result in recognition with input data prepared at the same session as the database. On the other hand, the system achieved recognition rate of 68.2% with the input data prepared at a different session or distinct in the head pose.