Research On Iris Cypts And Plaques Detection Methods

On the surface of human iris,there are a lot of different kinds of feature textures,such as plaques,crypts,furrows and circle lines,the spatial information of which,including size,location and shape,on the one hand,can be used as auxiliary information for iris recognition to further improve the iris recognition accuracy and speed,on the other hand,can also be used as the basic information for iris diagnosis.By now,the existing research on iris feature texture extraction is still in the exploratory stage,and the extraction accuracy and extraction speed can not meet the needs of practical applications.Aiming at the existing problems in the method of iris feature textures extraction,by taking the visible light iris image as the research object,and taking iris database built by ourselves as basis,this paper carried out the research from two aspects.On the one hand,in order to ensure the accuracy of iris texture extraction,an in-depth study on the key problem of image preprocessing in complex visible light iris was conducted,including iris inner and outer circle positioning problems and actual iris visible region segmentation problem;On the other hand,after the preprocessing of iris image,another targeted study how to extract plaques and crypts was carried out.The main research contents and contributions are as follows:(1)To solve the problem of iris effective region segmentation in complex visible light iris images containing light spot,eyelids,eyelashes and a variety of texture features,a method based on Union-and-Search set was proposed.For the inter circle localization,firstly,the pupil region was segmented coarsely by using the method of Union-and-Search set.Secondly,the edge contour of the pupil region was obtained by edge tracking method.Finally,the Hough transform method was applied to complete accurate inner circle positioning;For the outer circle localization,firstly,a coarse positioning of iris outer circle was carried out according to the gray scale characteristics of the iris edge.And then,the canny operator was used to complete iris binarization operation.Finally,the accurate outer circle was obtained according to the density distribution near the outer edge of the iris.The test results show that this method can overcome the interference of light spot,eyelid,eyelash and other characteristic texture,and has high positioning accuracy and positioning speed.After successful positioning of iris inner and outer circle,the method of Union-and-Find setwas used again to complete the rough segmentation of the effective iris region.After that,the method of rotating triangle filled was used to achieve precise positioning of the effective iris region.Compared with the existing iris visible region extraction methods,both the extraction accuracy and the extraction speed are all obviously improved.(2)Aiming at the problem of iris plaques extraction,a new method of plaques extraction based on bilinear templates and block strategy was proposed.First of all,according to the size range of plaques in the gallery,two linear templates were designed to extract all the textures.Secondly,all the target textures extracted were located by the Branch-and-Bound method.And then,the target textures was partitioned into blocks and the gray value of each block was calculated,thus,a feature vector representing the characteristics of the gray distribution of the goal texture was gotten.Finally,the neural network method was applied to test plaques.The experimental results show that the proposed method can be used to detect the plaques effectively in the complex iris images with various iris feature textures and interferences.(3)In order to extract crypts in the iris,this paper presents two kinds of extraction methods.Method one: the fuzzy set theory was introduced to extract the crypts in the iris.First of all,a one-dimensional linear texture template was designed to extract all the "narrow" strip texture according to the shape characteristics of plaque and non-plaque.Secondly,a set of fuzzy rules was designed to eliminate plaque;And then,according to the characteristics of crypts,another set of fuzzy rules was designed to identify crypts.The test results show that this method of crypts detection have a better effect and faster speed.Method two: in the process of crypts extraction in method one,the light spot would lead to the increase of false detection rate.In order to solve this problem,a method based on morphology was proposed to detect crypts.First of all,the structure element was designed according to the maximum width of the crypts in gallery and top hat operation in gray scale morphology was used to extract all the "narrow" textures.And then,the close operation in two-value morphology was applied to connect two cracked textures cased by the operation of gray morphological,thus,completing the extraction of all the "narrow" texture.After that,a feature vector was defined to reflect the different features of texture features and a recognition method based on the support vector machine was used to detect crypts and furrows.Finally,according to the different characteristics of crypts and furrows,the finalcrypts were obtained by using threshold method.The test results show that this method can overcome the interference of light spot and further reduce the false detection rate of the crypts.