Study On3D Reconstruction And Fast Imaging System Of Subcutaneous Veins

With the development of Wise Medical and Biometrics Identification Technology,analysis of subcutaneous venous has attracted wide attention to the application ofidentity recognition and clinical medical assisted intravenous injection. Depended onthe infrared spectrum absorption characteristics of the hemoglobin of venous vesselblood, infrared imaging is widely used in subcutaneous vein imaging. However, due tothe low contrast of the blood vessel, blur of the bifurcation and asymmetrical of thevessel matching in infrared venous image, segmentation and3D reconstruction of thesubcutaneous vein are still urgent issues in the field of vein analysis.In this thesis, base on the reflective Near Infrared (NIR) imaging technique, twomethods, including vein segmentation method based on multiple features classification,and vascular matching method of veins based on hybrid constraints and hierarchicaloptimization, have been proposed. Those two methods can rapidly and robustly realizesegmentation, parameter analysis, matching and3D reconstruction of the vein. Todisplay the depth and the location of the vein intuitively, the skin surface isreconstructed by using Gray code structured light to realize the fusion and display ofthe3D structure of the vein and skin surface, and the reconstructed vein are backprojected to the skin surface to achieve the contrast enhancement of the vein. The maincontributions of this thesis are as follows:(1) A local adaptive subcutaneous vein segmentation method is proposed by multiplefeatures classification. It overcomes the inaccurate segmentation in bifurcations andtails. First, enhancement, feature map extraction and rough segmentation of veinsare achieved by Wavelet Transform and Hessian Matrix. Then, vascular branchesare repaired by the vessel direction map. Meanwhile, their topological structure anddirections are optimized. Finally, by extracting vascular features by Matched-Filterin the profile space, K-Means are used to achieve the accurate segmentation of veins.Meanwhile, venous radii and centerlines are further optimized.(2) A novel vein matching and3D reconstruction method is proposed based on the epi-polar constraint and homographic property. It overcomes the matching problemscaused by the incomplete structure of vein segmentation and the asymmetricstructure if left and right views. First, the matching candidate point sets and thehomographic matrix of two images are obtained by epi-polar constraint and SURF.Then, the complex centerlines of veins are stably and accurately matched by bidirectional Homographic Point to Point Matching(HPPM), the optimization ofhomographic matrix and branch to branch matching. Finally,3D structure of thesubcutaneous vein is obtained according to the triangulation principle andcalibration procedures.(3) A prototype of the3D reconstruction and rapid imaging system of the subcutaneousvein is developed. By utilizing the spectral response characteristics of the cameraand the light source control circuit, the system implements the time-sharingmultiplex between near infrared imaging and visible imaging. The calculation speedof the system is increased by using multi-thread computing. Finally, the system iswith the functions of the real-time segmentation, parameter analysis of veins,3Dreconstruction of the vein and the skin, and back-projecting the vein onto the skin.