Regarding the fingerprint algorithm used by the fingerprint punching machine in the optical fingerprint scanner, for the processing of low-quality fingerprint images, we have deeply studied the calculation and image segmentation method of the low-quality fingerprint image texture direction, and proposed a neural network based on the correctness of the texture direction. The method of training and learning, and the fingerprint segmentation method based on this, the correct calculation of the texture direction of the low-quality fingerprint image is the basis for the correct feature extraction and matching, aiming at the shortcomings of the existing gradient-based and low-pass filtering methods, On the basis of calculating the direction of the texture with the gradient method, we train and calculate the correctness of the preliminary calculation result of the direction in combination with the fingerprint segmentation, so as to perform the fingerprint segmentation according to the correctness of the direction and correct the wrong direction according to the correct direction. The network has different response results for specific image blocks in different directions. According to these response results, the texture direction of the image block can also be determined and fingerprint segmentation is performed. The experimental results show that these methods can effectively improve the accuracy of low-quality fingerprint image feature extraction.
In-depth analysis of the types of fingerprint image regions, a method of secondary segmentation of fingerprint images to remove residual lines is proposed. Many fingerprint segmentation algorithms can effectively separate regions without lines and lines whose structure cannot be recovered, but cannot effectively separate the structure. Clear residual texture area, the secondary segmentation method is based on the initial segmentation of the fingerprint to separate the area without texture and the texture area where the texture structure cannot be recovered, analyze the remaining area, and separate the residual texture area, thereby reducing the extraction of erroneous features .
The effective processing of low-quality fingerprint images in the fingerprint algorithm greatly improves the accuracy of fingerprint recognition and the speed of fingerprint matching.