Embedded Finger Vein Recognition System Using Raspberry Pi With Improved Region Of Interest
Finger Vein Recognition System (FVRS) is a biometric technology that identifies or verifies an individual based on unique vein patterns. Its performance is determined by the robustness of hardware and software development. According to the analysis of original FVRS, the software development stumbled...
Saved in:
| Main Author: | |
|---|---|
| Format: | Monograph |
| Language: | English |
| Published: |
Universiti Sains Malaysia
2017
|
| Subjects: | |
| Online Access: | http://eprints.usm.my/53033/1/Embedded%20Finger%20Vein%20Recognition%20System%20Using%20Raspberry%20Pi%20With%20Improved%20Region%20Of%20Interest_Lim%20Yuan%20Zhang_E3_2017.pdf http://eprints.usm.my/53033/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Finger Vein Recognition System (FVRS) is a biometric technology that identifies or verifies an individual based on unique vein patterns. Its performance is determined by the robustness of hardware and software development. According to the analysis of original FVRS, the software development stumbled upon image pre-processing stage particularly in the extraction of region of interest (ROI). The reason to the failure was because of the use of fixed window ROI extraction method, where the ROI was extracted using a predefined rectangle window. This method disregards the misplacement of finger and has no common reference point in extracting ROI. The objective of this project is to improve the accuracy of original FVRS by correcting the orientation of misaligned FV image and extracting ROI based on localised benchmark. In overall, this project contributed a new method of finger vein image segmentation using watershed segmentation with distance transform, then applied adapted methods to correct finger image orientation and extract consistent ROI using single sliding window based on phalangeal joints. OpenCV image processing library and C++ language were used in the development. The results proved that the improved system is able to correct the orientation of finger image regardless of finger placement during enrolment and verification, as well as obtaining ROI based on phalangeal joints. The performance evaluation shows that the verification accuracy of the improved system achieved 89.33%, an increase of 37.66% compared to original FVRS. |
|---|