A Programming Based Boosting in Super-Classifier for Fingerprint Recognition
A super-classifier with programming based boosting has been designed and established for fingerprint recognition. This multiple classifier set is comprised of three different classifiers. The first classifier is an OCA based modified RBFN with BP learning
- PDF / 214,971 Bytes
- 11 Pages / 439.37 x 666.142 pts Page_size
- 20 Downloads / 196 Views
Abstract A super-classifier with programming based boosting has been designed and established for fingerprint recognition. This multiple classifier set is comprised of three different classifiers. The first classifier is an OCA based modified RBFN with BP learning, second classifier is a combination of Malsburg learning and BP Network and third classifier is a SOM based modified RBFN with BP learning. These three individual classifiers perform fingerprint identification separately and these are fused together in a super-classifier which integrates the different conclusions using programming based boosting to perform the final decision regarding recognition. The learning of the system is efficient and effective. Also the performance measurement of the system in terms of accuracy, TPR, FPR and FNR of the classifier are substantially high and the recognition time of fingerprints are quite affordable.
⋅
⋅
Keywords Fingerprint recognition OCA Malsburg learning BPN RBFN Programming based boosting Holdout method FPR FNR
⋅ ⋅
⋅
⋅
⋅
⋅
SOM TPR
⋅ ⋅
1 Introduction Biometric identification is a technology, which identifies a person based on their physiology or behavioral characteristics. Fingerprint recognition is reliable and accurate biometric method that has been extensively used in a number of applications for a person’s identity authentication. Fingerprint recognition is very effective in fields such as improving airport security, strengthening the national borders, in travel documents, in preventing ID theft, person authentication, access control system and retrieval of an identity from a database for criminal investigation S. Kundu (✉) ⋅ G. Sarker Computer Science and Engineering Department, NIT Durgapur, Durgapur, India e-mail: [email protected] G. Sarker e-mail: [email protected] © Springer Science+Business Media Singapore 2017 S.K. Sahana and S.K. Saha (eds.), Advances in Computational Intelligence, Advances in Intelligent Systems and Computing 509, DOI 10.1007/978-981-10-2525-9_31
319
320
S. Kundu and G. Sarker
etc. Fingerprint recognition is a difficult task because the fingerprints diverge highly in terms of quality, size, shape, rotation and occlusion. A RBFN with Optimal Clustering Algorithm (OCA) was established in [1, 2], for clear and occluded fingerprint identification and localization. A combination of Malsburg learning and Back propagation Network (BPN) was established in [3], for clear, occluded and rotated fingerprint recognition and rotation and location invariance localization of the fingerprints in different fingerprint image frames. Many fingerprint identification methods which also already established are based on feature (minutiae) extraction and minutiae matching. This method mentioned in [4]. This methodology mainly involves extraction of minutiae points from the sample fingerprint images and then performing fingerprint matching based on the number of minutiae pairings among two fingerprints. A technique for fingerprint identification by minutiae feature extraction using back-propagation algorithm has been appro
Data Loading...
