Research On Alogrithms Of Feature Extraction And Matching Of Fingerprint Images

At present, the fingerprint recognition technology is becoming a hot spot which iswidely concerned and applicated at home and abroad because of its advantages such asconvenient to collect, stable efficiency for recognition and low system cost. However,there is room for improving of current fingerprint feature extraction and matchingalgorithms. To improve accuracy and reliability of matching methods, the main researchcontents of this paper include the following aspects:Firstly, the development history of fingerprint recognition technology is introduced,and the development status of feature extraction and matching algorithms for fingerprintimages is studied with emphasis based on the analysis of development status of fingerprintrecognition technology at home and abroad. Meanwhile, the composition of fingerprintrecognition system is mentioned.Secondly, the basic theory and process of preprocessing of fingerprint images iselaborated, and thorough study at several common binarization and thinning algorithmsfor fingerprint images is carried on. Then, to make a comparison among the algorithms,emulation for them is conducted. In addition, an optimized thinning algorithm is proposedon account of the disadvantage of advanced OPTA algorithm.Thirdly, a pseudo minutiae removal algorithm based on the local structuralinformation combined with two feature extraction algorithms is put forward afteranalyzing and studying the template method and ridge line tracing method which is basedon the eight neighborhood. Among them, the method combining the factors of distanceand structure is carried on during the process of pseudo removal, and ridge line tracingmethod is used to get the minutiae information including the sample information in thecorresponding ridge and the direction information.Finally, the fingerprint matching algorithm based on congruent triangles, thefingerprint matching algorithm based on ridge line correction and sectional fingerprintmatching algorithm based on local feature vector are used to match the sets of minutiaeextracted, and running time and false rejection rate(FRR) and false acceptance rate(FAR)of the three kinds of fingerprint matching algorithms are compared. At last, the best threshold scores are determined by the images of the false rejection rate and false acceptrate test results.