作者: Louis Coetzee , Elizabeth C. Botha
DOI: 10.1016/0031-3203(93)90151-L
关键词:
摘要: Abstract Algorithms are identified which best suited for an automatic fingerprint recognition system operating on low quality images. New preprocessing algorithms noise removal and binarization described. Three approaches to classification investigated: a correlation classifier, two feature-based schemes. The results database of 80 fingerprints obtained with spatial-frequency features. classifiers (neural net, linear classifier nearest neighbour) using these features successful in identifying independent test set. Details the shown. In conclusion suggestions made concerning most suitable each processing steps.