Facial recognition in the heterogeneous range of angle using Speeded-Up Robust Features (SURF) / Nur Syafiqah Liyana Romely
Biometric system is a technique to identify a person by using their own body characteristics. It acts as identification and access control in computer science. There are two types of biometric identifiers, which are physiological identifier and behavioural identifier. This research identifier is bas...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2017
|
| Subjects: | |
| Online Access: | https://ir.uitm.edu.my/id/eprint/69258/1/69258.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Biometric system is a technique to identify a person by using their own body characteristics. It acts as identification and access control in computer science. There are two types of biometric identifiers, which are physiological identifier and behavioural identifier. This research identifier is based on physiological characteristic that are related to the shape of the body and is one of biometric system that has higher success as an identifier that is facial recognition. Over the decades, facial recognition has been broadly explored, but there is still room for improvement since the focus of the earlier approach is more on frontal facial images and non- ffontal facial images are extensively unexplored. Non-ffontal facial images have more hidden information compared to frontal facial images. Although the current facial recognitions approach have great performance to recognize faces, but current system are usually affected by several interruptions such as uneven illumination, expression and pose-variations. To solve the problem of pose variant, Speeded-Up Robust Features (SURF) has been chosen. SURF is one of techniques used to reduce the time computation which is good for a system to have a good computation performance. This research aims to develop a prototype that able to recognize faces in the heterogeneous range of angle using SURF. The accuracy of this prototype is about 70.00 %. For future work, enhance SURF algorithm maybe by combining SURF with other techniques to have a better recognition result. |
|---|