Description:The method of fingerprint identification described in figure 1 and figure 2 is intended to identify or authenticate the user even during poor skin condition. The block schematic shown in figure 1 consists of two process database creation and fingerprint matching. All users who are required to prove authentication have to register using their fingerprint. The fingerprint acquisition block 101 accepts the I/P from the fingerprint scanner and generates the fingerprint image. The fingerprint image is then preprocessed for image enhancement and minutiae features required for matching are extracted (102). These features are useful for fingerprint recognition to authenticate the user. The extracted features for each user are stored in the database 103 for further use during the fingerprint matching process. The extracted minutiae not only consist of number of minutiae points but also consists of type of minutiae with its position on the fingerprint skin. Each user has to register five patterns of their fingerprints with a small change in placement of fingerprints on the scanner in block 101.
A separate 9x9 pattern of various standard minutiae points and image patches with poor skin conditions are created which are later useful for fingerprint reconstruction. During the fingerprint matching process, the user has to prove his/her authenticity by giving their fingerprint by fingerprint scanner 104. They may sometimes fail to prove the authenticity due to their poor skin conditions. To avoid this before matching at 107 each accessed fingerprint is predicted for damage if any using a machine learning algorithm and then reconstructing the damaged parts of fingerprint images using fingerprint reconstruction model 105. Then reconstructed fingerprint is further processed for extraction for its minutiae at 106 then this feature is matched with database 103 for fingerprint recognition.
Figure 2 disclosed the model of fingerprint reconstruction. 201 accepts the I/P from fingerprint image scanned by fingerprint scanner 104. Block 202 predicts this image for damage with poor skin condition using machine learning algorithm. Linear PDE equation can identify the linear scars on finger print and standard templates for various finger skin damage like burns or seasonal scars can be used for predicting the other damaged areas. Once the region of interest (ROI) is marked for construction then block 203 performs segmentation of this target area and identify vicinity patches (refer reference 2 for this). Each pixel in the target regions is recovered (block 204) using a 9x9 window.
The prediction for best match of pixel value is based on the access from source patch and the stored minutiae templates of the 9x9 window. The KNN-SVD algorithm from reference 2 shall be used to predict the best suitable substitute for the pixel. The best predicted match updates the pixel value in the vicinity patch. The process of reconstructing the pixels in the target region is repeated by 205 till the recovery of all pixels in the ROI. This reconstructed image is then given to block 106 for feature extraction, which extracts the minutiae's and matches with database 103 to authenticate the user. The method can be used for authentication of any personal electronic gadgets used by the user.
, Claims:1. A fingerprint matching method comprising two process database creation and fingerprint matching; fingerprint matching process consists of reconstruction of fingerprint image which is damaged due to poor skin condition further to reconstruct the fingerprint image pixels in the damaged areas are first marked by prediction and then reconstructed by replacing a pixel from 9x9 window from target patch with best suitable predicted match; further for best prediction the knowledge of source patch and standard 9x9 minutiae template points are to be used.