Iris Recognition Method And Its Implementation Based On K-NN Classification Matching

Biological feature recognition technology is to recognize personal identity based on physiological or behavioral characteristics by combing computer with high science technology means such as optical, sound science, biology sensor and biology statistical principle etc. It is safer, more secret and more convenient than traditional identity recognition means. It has some advantages such as being difficult to forget, preventing counterfeit and easy carrying .It is widely used in government, troop, bank, social welfare undertake, electronic business etc.Iris includes abundant information. Study indicates that each iris has high exquisite and unique texture, and this uniqueness is decided by difference between embryo growth circumstances. Iris has features of high stability and reliability compared with other biometrics feature, and also non-invasion. Iris recognition can reach very high accuracy by using good algorithms, it has wide application prospect.Iris recognition system includes iris image acquiring, image preprocessing, feature extraction, coding and recognition and so on. This paper proposes some improved methods based on generalizing present iris recognition technologies and has definite reference to iris recognition technology.On iris image preprocessing, it uses binary valued threshold,corrosion and expansion to segment pupil, and uses projection method to locate the centre and the radius of pupil; It locates outside edge by means of combing circle detected operators with improved Hough transform. This method is easy to realize and needs less memory space and less time-consuming. Normalization is achieved by coordinate transform, and enhancing image with histogram equilibrium.On feature extraction, it discusses algorithm to extract texture feature of iris image based on Gabor transform, extracting phase information of iris image by Gabor filter and constructing 256 bytes binary coding; Because Gabor filter has a shortcoming of existing DC component, therefore the paper using Log-Gabor filter to extract the phase information of iris feature, and to code by binary circular coding. This algorithm is not subject to bandwidth without DC components. The experiment result indicates that the algorithm has a good effect.The paper analyzes algorithm of matching and recognition based on hamming distance and judging similar degree of two iris codes by calculating hamming distance. In order to eliminate the effect of rotation of iris image on iris recognition, iris codes acquired must be dislocated matched. In regard to the effect of rotation of iris image on iris recognition in the algorithm, the paper puts forward a improving fast k-NN search algorithm, with matching and recognition. The experiment results indicate that recognition rate is achieving 95.2% when choosing k=3 neighbors. In the last, using MATLAB software realizes the algorithm of iris recognition; experiment has performed on CASIA iris image database and has got a good effect.