Research Of Palm Image Feature Extraction Based On Palm Diagnosis

In recent years,the technology of biometrics identification was improved andapplied on many fields, but it is a new application of using biometrics identificationinto the area of the diagnosis of human diseases.Plam diagnosis as a kind oftraditional Chinese medicine,which reflects the person’s physical health through thechanges of palm color, texture, shape features.It has important implications to achieveautomated palm diagnosis useing computer, not only popularizing and themodernization of the Chinese traditional medicine but also can help discover ourdisease and cure it as early as possible.In palm diagnosis,the most important thing is to extract the palm of the textureand color features.It is called diagnostic features which appeared in the specialposition ofthe plam and make peaple easly illness. These special structures aredivided into closed special structure and unclosed special structure according to theirframe. Different special structures in different area means different disease andsymptom. When some unusual occurs in our body,the colour will changed in theparticular area,so it is vital to recognize these special structures and colour exactly inautomated palm diagnosis.The research works in this paper mainly focus on thefollowing three aspects:(1) We preprocess the collected images to extract the palm part of interestedarea.Referring to the knowledge of palm diagnosis and Eight Diagrams feature, wedefines some key points and the related line, and using the points and lines to segmentthe palm image.(2) The important feature is depend on the palm color spots to diagnosedisease.In order to extract the color feature of image, the image was changed fromRGB space to HSV space.But the color of the palm does not include all types of color in HSV color space, maximum value range of each component is determined throughthe experiment,then enlarged three ranges.In order to simplify the complexity ofprocessing, the quantization of color space is used in this paper.In order to extract thecolour spots,we adopt the watershed segmentation algorithm based on control markerto extract the spot on the palm,and classify the extractioned spots.(3) The special structures feature extraction and recognition:In this paper,according to the characteristics of the closed lines and the unclosed palmtexture,different methods are put forward to extraction and identification. Firstly,enhancement the palm image,then using the Hildich thinning algorithm to thin thepalmprint.According to the characteristics of the closed lines,the method to extract theclosed loopwas proposed.And according to the perimeter and area of the closed loop,and the relationship between the edges, to distinguish the different kinds of closedlines.For unclosed special structures’feature,first extract the grain line intersectionpoint,judgment the length of the different directions,and calculate the angle betweenthe lines,then realize these two unclosed structures’recognition.