Real-time online fingerprint image classification using adaptive hybrid techniques

Abstrak

This paper presents three different hybrid classification techniques applied for the first time in real-time online fingerprint classification. Classification of online real time fingerprints is a complex task as it involves adaptation and tuning of classifier parameters for better classification accuracy. To accomplish the optimal adaptation of parameters of functional link artificial neural network (FLANN) for real-time online fingerprint classification, proven and established optimizers, such as Biogeography based optimizer (BBO), Genetic algorithm (GA), and Particle swarm optimizer (PSO) are intelligently infused with it to form hybrid classifiers. The global features of the real-time fingerprints are extracted using a Gabor filter-bank and then passed into adaptive hybrid classifiers for the desired classification as per the Henry system. Three hybrid classifiers, the optimized weight adapted Biogeography based optimized functional link artificial neural network (BBO-FLANN), Genetic algorithm based functional link artificial neural network (GA-FLANN) and Particle swarm optimized functional link artificial neural network (PSO-FLANN), are explored for real-time online fingerprint classification, where the PSO-FLANN technique  is showing superior performance as compared to GA-FLANN and BBO-FLANN techniques. The best accuracy observed by the application of PSO-FLANN, is 98% for real-time online fingerprint classification.