Research On Technologies Of Palm Principal-line Features Extraction

Palmprint is a sort of hereditary feature controlled by polygenes.Once the palmprint is developed,it never changes and it is individually unique.As a kind of important biological characteristics,it has been widely used in the identification field.On the other hand,biological research demonstrates there is a correlation between palm principal line and diseases,especially congenital diseases and hereditary diseases,such as tumors.Therefore,in order to study the correlation of palm principal line and diseases,we need to extract the complete palmprint principal lines.Currently,palm principal line is mainly used for identification that needs a part of feature points instead the complete principal line.Therefore,the methods used to extract the complete principal line are rare.In this paper,we aim at extracting complete,pure and natural palm principal line.We mainly focus on the research and experiments on palmprint image acquisition,image preprocessing and principal line extraction.The contribution is concluded as follows.(1)A normalization preprocessing on the palmprint image is implemented.Contactless palmprint image captured by digital camera in this paper cannot be directly used for feature extraction because of its ununiformed size and qualities.Therefore,we implement a normalization preprocessing on the palmprint image.Firstly,we extract the palm contour,then locate the finger valley points using the improved disk method.Subsequently a normalized directional correction is performed on the palmprint image using these points to make the principal line trend towards the same direction.Finally,we capture the region of interest(ROI)using the method of searching the maximum inscribed circle,and normalize the ROI image into the same size.Experimental results show the preprocessing method can accurately capture the principal line region,and the direction of principal line among different ROI images are roughly the same,which improves the robustness of the principal line extraction algorithm.(2)An improved principal line extraction algorithm based on mathematical morphology and template matching is proposed.Since the existing algorithm using 4 nonspecific directional templates to extract palm principal line,which is apt to extract wrinkles and noises in all directions.Therefore,this paper improves the current algorithm according to the directional regularity of the principal line.Firstly,we divide the principal line into different parts according to its trends,and incorporate these part-lines into 3 specific directions,then we use a new smoothing template to smooth the image,and replace the 4 nonspecific directional template with 3 specific ones to extract the principal line.The experimental results show that the improved method can shield the interference from the different directions of the principal line,and effectively avoid the error extraction of wrinkles and noises.(3)A novel principal line tracking algorithm is proposed.Since the principal line extraction method based on morphology is sensitive to directional noise,it cannot eliminate interference from noise in the same direction of the structural elements.So we proposed a principal line extraction algorithm based on neighborhood searching.Firstly,we use mathematical morphology to extract the principal line from the ROI image roughly,and then determine a seed point grown on the principal line,and use this point to start the neighborhood searching algorithm to extract the complete principal line in the ROI of the gray scale image.The experimental results show that this method is able to extract complete,pure and natural palm principal line.