Recently, great improvement has been achieved in the fingerprint sensing technology and automatic recognition algorithms. Various techniques, including fingerprint acquisition, classification, enhancement and matching, are highly advanced as well. However, there are still some challenging tasks in the practical application of fingerprint recognition, such as matching of poor-quality and partial fingerprints, fingerprint indexing, cross-matching of multi-type fingerprints, etc. These issues seriously affect the whole fingerprint recognition system's performance. In this paper, we will discuss these challenging issues. The main contributions of this thesis include: 1) We represent the fingerprint with a novel local feature descriptor, which is composed of minutia, the sample points on associated ridge and the adjacent orientation distribution. Then a novel fingerprint recognition method is proposed combining the orientation field and the local feature descriptor. The similarity score is evaluated by fusing the orientation field and the local feature descriptor. The experiments have been conducted on three large-scale databases. The comparison results approve that our algorithm is more accurate and robust for those poor-quality and partial fingerprints. 2) We propose a novel minutiae indexing method for fingerprint indexing, which can also speed up those minutiae structure-based fingerprint matching algorithms. An orderly sequence of features is extracted to describe each minutia and the indexing score is defined to select minutiae candidates from the query fingerprint for each input minutia. The proposed method is applied in both fingerprint identification and minutiae structure-based verification on large-distorted fingerprint databases (FVC2004). The experimental results confirm its validity in improving the accuracy and efficiency. 3) We develop a novel cross-matching algorithm for the fingerprints captured with multi-type sensors. A coarse-fine method is used to estimate the optimal scale between the input and template fingerprints. Then a hierarchical registration algorithm is proposed to handle the non-linear distortion among multiple kinds of sensors. Experimental results on the FINGERPASS cross-matching databases demonstrate that our algorithm effectively improve the performance of cross-matching. To sum up, in this thesis we have made a lot of fruitful attempts and significant progresses on some practical problems in fingerprint recognit...
修改评论