Research On Detection And Recognition Of Human Nail Feature Information Based On Multispectral Imaging

As a rather important tissue organ,the health of the fingernail is easy to be ignored.When it comes to the illness for human beings,its relative features can be gained by the fingernail.Therefore,traditional Chinese medicine recognizes body health according to the diagnosis from one's fingernails.Nevertheless,the features of shape,color,luster and concave change can present different results as for various people,which can bring certain difficulties to the doctor and lead to misdiagnosis.Instead,multispectral imaging technology is an optical sensor with combination of the image and spectrum,which can gain the targeted images and spectral information at the same time.Compared with other sensors,it has the advantages of more wave bands,sharp resolution and high space resolution,ect.This paper attempts to make use of multispectral imaging technology to test and recognize the common features of the fingernails,in order to provide an assistant tool for traditional Chinese medicine's diagnosis.The specific research contains following aspects:(1)Research the spectral features of fingernails.In the test of fingernails' characteristic information,the fingernails' color,shape and texture should be known at first.Besides,it's important to extract and analyze the spectral characteristic curves of several kinds of common fingernails' features by visual inspection,such as crescent white,white spots,dark strips and yellow color.(2)Set up the multispectral imaging system which combines the CCD camera and LCTF according to the principle of multispectral images.After comparing the advantages and disadvantages of two compound modes,the test applies the way that LCTF lies before the lens and camera,which is benificial to gain the clear images and simple installation and adjustment.In addition,the operating principle,location and correction and data presentationof this system are also introduced in details.(3)Select the characteristic wave bands of fingernails.It needs to gain the multispectral data of the fingernails' samples by multispectral imaging system,so as to preprocess the images.After analyzing the necessity and ways of the region of interest in the data processing,it comes to reset the reflectivity to the original gray images of the fingernails applying the method of comparative measurement.Meanwhile,the image segmentation of the fingernails realizes by watershed algorithm.What' s more,it uses band index method based on band selection to research several kinds of characteristic bands respectively.Furthermore,it applies the PCA based on the band extraction to select the optimal band combination which can distinguish these features.(4)Take advantage of the maximum likelihood and SAM respectively to test and classify these features of the fingernails based on the multispectral data and characteristic bands above.It verifies the validity of the above characteristic bands and analyzes the classified accuracy of two classification methods.So it turns out that the SAM is more effective for the classification of human fingernails.The paper illustrates that the testing method of fingernails' characteristic information based on multispectral imaging can classify the common features of fingernails accurately and it proves multispectral imaging technology has certain potentials in the field of fingernails' diagnosis.