Fingerprint Image Quality Classification Based on Feature Extraction

Fingerprint recognition technology has been widely used in criminal investigation, attendance system, security testing and other fields and has become one of the most mature biometric technologies. Since fingerprint image quality affects heavily the performance of fingerprint recognition system, accurate evaluation of fingerprint image quality has great value in improving the performance of automatic fingerprint identification system and applicability of fingerprint recognition algorithms. In this paper, we mainly investigate fingerprint image quality classification approaches based on feature extraction. We extract six groups of quality features including frequency domain features and spatial domain features, and respectively use methods such as individual quality feature parameter, linear weighted sum, wavelet domain energy, Kmeans clustering, Support Vector Machine and BP neural network to classify fingerprint images into three types of high quality, medium quality and low quality images. Experimental results indicate that classification accuracy of the method combining six groups of quality feature vector with BP neural network is higher than other methods.