A Brief Introduction to Node Extraction and Filtering of Fingerprint Feature Values in Fingerprint Scanners

Node extraction and filtering in fingerprint scanners are generally difficult. The usual node extraction process goes through texture direction calculation, fingerprint segmentation, fingerprint enhancement, texture extraction and binarization, texture refinement, and finally the refined texture. Detect nodes in the image (mainly refer to endpoints and bifurcation points), in the refined texture image, for the endpoint, there is only one adjacent point on the texture; for the bifurcation point, there are and only three adjacent points On the texture, the second type of node extraction method is based on the gray image to track the texture, and detects the node while tracking the texture, and then uses the asymmetry of the space at the node to detect the node from the gray image. The node extraction algorithm often misses some real nodes, and also generates some wrong nodes. Improving the correct string of node detection depends on the effectiveness of various algorithms in the node detection process. In addition, In order to reduce wrong nodes, Yiyin adopts the ridge enhancement algorithm and node filtering algorithm to remove the wrong nodes. After the nodes are detected, the machine learning method is used to verify the correctness of each node and correct the type of each node. B5mM is called based on The knowledge-based method enhances the ridges, thereby reducing the detection of wrong nodes. Yin Quimin adopts the knowledge-based method to delete the wrong nodes.

The field of texture counting is relatively simple, and mainly depends on the correct calculation of the direction and the correct extraction of the texture. Due to its simplicity, there are few reports on the calculation method of the texture, and it is assumed that the texture period is constant. Automatic number of lines is calculated, and some fingerprint matching algorithms use the number of lines as a special envoy.
Image quality calculation, it is very important to calculate the image quality in the automatic fingerprint identification system. By calculating the image quality, it can prevent low-quality fingerprint images from being registered in the database, so that the correctness of the system can be improved. The calculation method of fingerprint image quality is based on Method for the ratio of directional and non-directional regions J. Method 6 based on gamma filter, method based on wavelet compression, although several methods for calculating the quality of fingerprint images have been proposed, there is no recognized standard that can measure the quality of a calculation method.