Research On Acquisition And Recognition Of Tongue Image

Tongue diagnosis is the essence of the traditional Chinese medicine. It is a clinicalexamining way which works now as well. The traditional method of obtaining tonguediagnosis information is finished by observing tongue body and tongue-coating. Sothe result of this method is necessarily affected by the surrounding factors such aslight, the posture of the patient. Moreover, the experience of the doctor affects theaccuracy of the diagnosis. These entire problems baffle the developing of tonguediagnosis. According to these problems, the major goal of this thesis is to design aset of tongue image analysis instrument which has the following functions: obtainingthe origin tongue image, finding the edge of the tongue body, and recognizing theclass of tongue image by its feature. The instrument can help doctor improve theability to examine.This paper implements an integral system for acquisition and recognition of tongueimage. It is composed of two parts: the hardware and the software. The hardware partincludes acquisition device,computer and output device. The acquisition device ismade up of LED, digital camera, dark box and mechanical components. The softwarepart implements the following functions: main program, image process, edgedetection,image segmentation and patten recognition. The tongue image edgedetectionis implemented by canny algorithm and then segments the tongue body andthe tongue coating by the method of color threshold. In the HSV color space, thecolors of tongue and tongue coating are identified and diagnosed based on thestatistical pattern recognition theory.It can reduce the volume of apparatus and increase the electrical security by usingthe LED as the mainly light source. The LED increases the uniformity of the light onthe integrating spheres as illuminant geometry condition as well. We compare theresult of tongue image analysis instrument with the sample and the accuracy of thisinstrument is over 80 percent.