Loading...
Please wait, while we are loading the content...
Similar Documents
Low cost contactless fingerprinting
| Content Provider | Indraprastha Institute of Information Technology, Delhi |
|---|---|
| Author | Gaur, Prateek Ajmani, Stuti |
| Abstract | Identi cation using ngerprint is the most widely employed in the eld of biomet- rics. The existing ngerprint matching systems use especially designed ngerprint scanners for the purpose of capturing ngerprint images. Depending on the quality and the product, these scanners could be expensive and/or bulky. With the high penetration of mobile phones, webcams and digital cameras, it is highly likely that a biometric system using any of these devices could gain importance. There are a new set of problems related to such a system and images captured from such a device would be quite di erent than the ones captured from a dedicated ngerprint sensor. An algorithm to address all such problems in order to come up with such an inexpensive, easy to use, convenient, contactless ngerprint system is proposed. In such a system the ngerprint images could be captured using any device in hand: a mobile phone camera or a webcam. Images captured from such devices, are generally coloured, could have a lot of noise, erratic background thus rendering the quality of images captured much lower than the ones captured from dedicated sensors. Because of these di erences we propose a new contactless ngerprint matching algorithm for low resolution images. The key point is that we try to work with low resolution im- ages. Two datasets are used in this study. The Hong Kong Polytechnic University Low Resolution Fingerprint Database was used for training the system and then for testing as well. Another database collected by the students of IIIT Delhi was also used. This dataset was used for testing purposes. At rst image cropping, segmenta- tion and enhancement is performed as preprocessing steps. Next we extract features and use genetic algorithm and support vector machine as classi ers. We use GA to obtain the most optimized weights, for our fused scores and SVM as a classi er. Commendable results were attained by the fused score approach using GA, giving an accuracy of approximately 96 %. |
| File Format | |
| Language | English |
| Access Restriction | Authorized |
| Subject Keyword | Contactless Complexity Algorithms Image analysis Machine learning Ngerprints |
| Content Type | Text |
| Educational Degree | Bachelor of Technology (B.Tech.) |
| Resource Type | Thesis |
| Subject | Data processing & computer science |
