Research On Fringe Projection 3D-Colored Tongue Diagnostic System

Tongue diagnosis is an important part of Traditional Chinese Medicine(TCM)inspection and plays an important role in TCM clinical diagnosis.At present,the research on the objectification of TCM tongue diagnosis is mainly aimed at the twodimensional tongue diagnosis system.It cannot obtain three-dimensional information such as the depth of the tooth marks of the tongue and the contour of the tongue.Aiming at the acquisition of color 3D data on the tongue surface,this paper designs and develops a color 3D tongue diagnostic system based on fringe projection,and realizes the automatic segmentation of the tongue surface.The research contents of this paper mainly include:1.A dual-camera-single-projector fringe projection color 3D tongue diagnosis system was designed.The projector and the left and right cameras are used to form two pseudo-binocular systems,which supplement the three-dimensional information that cannot be obtained by the traditional binocular system,and realize the complete acquisition of the color point cloud of the tongue.Reasonable equipment selection and system hardware design were carried out for the color three-dimensional tongue diagnosis instrument.The upper computer software is designed to realize the functions of system calibration,color calibration,image acquisition,three-dimensional calculation and point cloud display.2.A process optimization method of epipolar and order hybrid constrained stereo matching is proposed,which realizes the simultaneous acquisition of the point cloud of the true and false hybrid binocular system.Starting from the ideal raster image,the matching points of the left and right cameras are obtained respectively,and three stereo matching point sets of the binocular system are obtained at the same time to realize 3D reconstruction.The polynomial color correction algorithm was studied,and the polynomial color correction algorithms of various combinations of terms were compared,and 9 polynomials were selected for color correction of tongue pictures.Finally,the color information of the 2D image is mapped to the 3D point cloud using sub-channel cubic interpolation,and the complete acquisition of the 3D color tongue point cloud is realized.3.An algorithm is proposed to automatically segment the tongue surface by combining 2D mask map and 3D spatial geometric features.The color and hue constraints and the morphological hole filling algorithm are used to realize the rough segmentation of the two-dimensional tongue image,establish a mask map,realize the rough localization of the tongue,and improve the efficiency of three-dimensional reconstruction.The characteristics of the rough segmented tongue point cloud obtained by 3D measurement were analyzed,and different segmentation and recognition algorithms were compared.