Research On Fast Identification Method Of Low Quality Damaged Fingerprint

As a biometric recognition technology,fingerprint is widely used in criminal investigation and civil field.Fingerprints are stable,universal and easy to obtain,it has become the most effective and convenient recognition solution and has legal effect.Automatic fingerprint identification System has been put into use in our country since 1990 s.More than 20 years passed,studies found that only about twenty percent fingerprints are of high quality in fingerprint database,while the rest are various types of low-quality fingerprints.High quality fingerprint recognition technology has been developed for a long time and many important achievements have been made,however,there is not a very mature model to deal with the low quality and damaged fingerprint.The key point is to improve the accuracy of the low quality fingerprint recognition and shorten the recognition time.Low quality fingerprint appeared to be a damaged image due to peeling,scars,stains,sweating and any other circumstance caused by bad finger situation.The situation mentioned above will lead to the fingerprint lack of enough texture in different degrees and form an unrecoverable region.The information that can be obtained from low quality damaged fingerprint is limited,how to eliminate the unreliable information,enhance the reliable information effectively and extract features for matching detection are the key problems of low quality fingerprint matching.This paper has three main research contents,while innovative work focus on pre-processing and feature matching.In the pre-processing stage,the adaptive fractional differential enhancement method is added to the basic traditional spatial filtering,which can effectively improve the reliability of texture area.According to the gray scale modulus of each image and the information entropy of the image,the differential order is reasonably selected.After enhancement,the texture information of the recoverable region is adaptively enhanced,and the edge of the fingerprint can be suitable smoothing.In the feature extraction phase,WDF(Walking Direction Field)singular tracking method is introduced.This method can reduce the number of search points and the lower the false detection rate.After the feature extraction,several rules are formulated to remove the spurious minutiae,which can prevent invalid minutiae being brought into feature matching steps.In the phase of feature matching,a two layer feature matching method is proposed.A triangular fuse feature structure descriptor is constructed by pre-matching and local matching steps.The feature structure is high anti-interference and high discriminative.We add bounding box mechanism in two-step matching decision,which can effectively solve the matching problem of fingerprint which has nonlinear deformation.A one to one matching verification test is carried out on the official data set of international fingerprint verification competition.False reject rate,false acceptance rate,equal error rate and processing speed have analyzed.Compared with some heuristic methods,the experimental results verify the applicability of the proposed method on low quality damaged fingerprints.This method can effectively reduce the recognition error and also the processing time,it's good for application of real-time detection of automatic fingerprint identification.