关键词: 关键词：指纹识别；指纹图像分割；半监督学习；协同训练；CoSeg

Abstract:

Abstract: Fingerprint segmentation is one of the key preprocessing steps in an automated fingerprint identification

system (AFIS). Effective segmentation of the fingerprint from the background could speed up following processes and improve

recognition accuracy. However, in traditional segmentation algorithms, it is essential to obtain lots of labeled fingerprints,

which are usually more expensive than unlabeled ones. To incorporate labeled and unlabeled data together, this paper proposed

CoSeg, a semisupervised fingerprint segmentation algorithm. Under the view of pixellevel features, i.e. Coherence, Mean and

Variance (CMV), CoSeg employs Label Box and SVM as two base learners and trains the final model for segmentation based on a co

training style algorithm. Experiments performed on FVC 2002 databases show that CoSeg can effectively exploit unlabeled data

with limited labeled data, and the proposed method is also robust when dealing with lowquality fingerprints.

Key words: Key words: fingerprint recognition； fingerprint segmentation； semisupervised learning； cotraining; cotraining based