Fingerprint Quality Classiifcation Method Based On Feature And Its Application Research

With high frequency of security threat and information leak, people put forward moreattention and requirements towards identity authentication system. Develop as the leap ofelectronic information technology provides a new technical approach and measure for thediversity and high safety of identity authentication. Biometirc recognition technologies havereplaced traditional status authentication as the foundation and core of new identityauthentication system. Fingerprint achieves a better balance of many performance indicatorsthan other biometrics which makes fingerprint become the first selection for biometricrecognition technologies, so it has been widely recognized and used. Fingerprint recognitiontechnology has developed mature, but there are some technology defects and automaticfingerprint identification system still needs further perfection. The existing technologyproblems mainly include fingerprint quality control, slippage region detection, deformationfingerprint recognition and latent fingerprints processing. Fingerprint quality evaluation andcontrol is in the front of recongnition system, and it is the foundation of the followingprocessing. Effective fingerprint quality evaluation and control can decrease the difficulty ofthe following processing and promote the performance of system, so it has very importantsignificance and actual using value to do research in evaluating fingerprint quality.Fingerprint collection is time-variant, so fingerprint quality evaluating belongs tono-reference images quality assessment. Meanwhile, the special texture structure makes thetraditional evaluation standards unsuitable for evaluating fingerprint quality. The majorresearch of this paper is fingerprint quality and fingerprint quality classification is realizedwith the combination of features which reflect fingerprint texture structure effectively, andquality classification is applied to enhancement processing. Firstly, we extract global qualityfeatures and local quality features to classify fingerprint images quality with methods such asquality feature threshold, multi-features weighted combination and K-means clusteringrespectively. By the analysis of the results, we point out disadvantages of these methods andintroduce BP neural network to quality classification process, sequentially suggest afingerprint quality classification method based on BP neural network. The proposed method istested on international fingerprint identification competition database FVC2004DB4and theresult of simulation demonstrates that the method could establish the nonlinearity relationship between multi-features and subjective perception effectively, so it achieves higherclassification accuracy. Second, simulations of three classical fingerprint image enhancementalgorithms are carried out, and we analyze the advantages, disadvantages and the applicableobjects of these enhancement algorithms combined with the results of simulaton. A methodcombining quality classification and fingerprint enhancement algorithm is proposed whichcould remove low quality fingerprint images and choose suitable algorithm to enhancefingerprint images based on their quality level, and it promotes subjective visual perception.At last, simulations of several binaryzation, thinning and feature extraction algorithms aredone, and we respectively use Gabor enhancement, STFT enhancement and the methodproposed in this paper to enhance sixty and five different quality fingerprint images. Throughcalculating and comparing performance parameter, it is comfirmed further that theeffectiveness and feasibility of quality classification technology in fingerprint identificationsystem.