特徵點比對是指紋辨識的主流方法。特徵點比對法是在兩枚指紋中匹配出相符的特徵點，並根據匹配出的特徵點數目判定兩枚指紋是否相同。由於指紋影像常伴隨著雜訊。若影像前處理時無法將雜訊濾除，則擷取出的特徵點亦會摻雜著偽特徵點，影響特徵點比對的正確性。因此近年有越來越多方法提倡結合多種指紋特徵來進行辨識。本研究提出結合LBP紋理特徵的特徵點比對方法，透過比較特徵點鄰近區域的LBP紋理特徵，進而篩選出辨識過程中所需的參考點以確保辨識的精確性。實驗結果顯示本研究提出的方法能夠確實改善傳統特徵點辨識方法的性能，在FVC2000與我們自建的指紋資料庫中實驗出EER結果皆優於傳統特徵點辨識方法。使用本研究的LBP紋理特徵比較來篩選參考點，能改善特徵點匹配錯誤的情況，並且獲得更佳的辨識性能。;The minutiae-based matching is the main stream method in fingerprint recognition which is an algorithm tries to find out the matched minutiae between two fingerprints. According to the amount of matched Minutiae, we can determine whether these two fingerprints are same or not. Due to the noise in the fingerprint images, image pre-processing has to be proceeded in order to reduce the noises, otherwise those extracted spurious Minutiae will influence the correctness of results. Therefore, recently, there are plenty of methods proposing the combination of multiple fingerprint features to identify two fingerprints. In this paper, we propose an improved minutiae matching method by integrating the Local Binary Pattern (LBP) feature, which compares LBP texture features in the adjacent region of Minutiae in order to select the reference points we want and guarantee the precision of identification. The experiment results show that our method can outperform the traditional Minutiae-based method. In FVC2000 and the fingerprint database we build, both Error Equal Rate (EER) of the outcomes are better than the traditional way.
