Research On Iris Cracks And Annular Lines Detection Methods

There are plenty of textures on the surface of irises,which are consisted of cracks,annular lines,crypts,plaques and the textures inside collarettes.The feature information contained in iris textures is the significant basis for iris identification and evaluation of human health status.However the existing iris textures detection methods have higher detection rate and false detection rate and cannot offer the detailed characteristics of the iris textures,but.these characteristics are the effective supplement of the existing feature of iris identification systems and can also provide the texture features required for human health status evaluation.Based on iris crack and annular line as test object,the study of the detection method and the description method of detected textures in this paper were proposed.The main research work and contributions are as follows:(1)In order to solve the difficult problem that accurately detecting iris cracks by eliminating non-cracks in rich iris textures,a detection method of iris cracks from coarse to fine was proposed.This method can be used to detect the cracks in the iris images with complex background.Firstly,according to the characteristics of Gray level difference between cracks and background regions,a line gray level difference template in horizontal direction was designed to search for possible areas of cracks in iris images that after being preprocessed and extended and normalized.The step could detect all cracks while detect simlar non-cracks,which including the textures inside collarettes,crypts,and plaques.Secondly,according to the similarity of textures inside collarettes in length and location while specificity of cracks in length and location,a Local density maximum clustering method to eliminate the textures inside collarettes was proposed.The method could classify the textures inside collarettes and outside collarettes and eliminate the textures inside collarettes.The test result showed that the method proposed was effective for eliminating the textures inside collarettes,and the accuracy was 92.21%.which provided a good foundation for the following steps.Finally,a method was proposed based on ridge line operator and energy density.The method can eliminate other non-cracks.According to the characteristics of the different feature textures having different extreme points in a certain region,a ridge line operator was designed to meet the characteristics of the crack ridge.The operator was only a certain direction of the crack with a significant detection,but for other non-crack textures after the ridge detection didn't have significant line characteristics and there would be a multiregional characteristics.This method firstly obtained binary image after the two steps,then found the corresponding original image regions with the binary image area as the center.The ridge operator was used in the region,then the information of fitting ridge shape was detected in the image.The energy density in the paper was used to determine the number of pixels in the vicinity of the central line and the proportion of the total area.The small value was crack,and non-cracks would be emilinated,thereby achieved the detection of cracks.The experimental result showed that the accuracy can reach 94.09%.In this paper,the ridge detection operator and energy density can reflect the characteristic information of the crack,and eliminate the non-cracks,and achieve the detection of cracks.(2)In order to solve the difficult problem that accurately detecting iris annular lines in rich iris textures and signal intensity of annular line is very weak.A detection method of iris annular lines was proposed based on vector weighted line detection operator.The dominant signal of the annular line will appear on a channel of the three component of the RGB space.But which component is random.The existing detection methods of annular lines are detected based on gray images.In the process of conversion from vector image to gray image,the traditional method is adopted fixed three components weighting,which makes the signal conversion of the dominant signal decrease.An adaptive weighting method is designed to maximize the weight of the dominant signal,and the most prominent single channel image is obtained.On the basis of this,the two value image is obtained by line detection,and then the shape factor is designed according to the shape feature of the circular line to find the area to meet the requirements of circular line detection.The experimental result in ROI images shows that the accuracy of this method is 91.78%,and the detection rate is increased.In order to obtain the feature information of the iris textures,take the cracks and annular lines as objects,according to their length,width,area and so on were selected to be feature parameters to describe the informations of cracks and annular lines.The computational formulas of these parameters were proposed,and the feature vector and matrix of iris cracks and annular lines textures was defined.Moreover,the establishment of the classification gallery for the algorithm testing and evaluation and the preprocessed of iris images were also discussed.