Research On The Design Of Near Infrared Blood Vessel Imaging System And Image Processing Algorithms

Venipuncture is one of the most common medical means in clinical medicine,However,in real life,owing to factors such as skin color,age,obesity and so on,venipuncture failure often occurs.Therefore,it has much realistic meaning and will be of great importance to develop a set of near-infrared blood vessel imaging system and image processing algorithms with simple structure and clear imaging.In this thesis,firstly,according to the theory of near-infrared vein imaging,the overall structure design of the system is completed,and according to the specific parameter requirements,the selection of the key components of the system and the construction and debugging of the system are completed.After collecting the blur vein images,image enhancement and image segmentation algorithms are used to extract the clear vein track.Through the projection correction algorithm,vein edge images are 1:1 in-situ projected onto the measured part,which shows a great application prospect in assisting doctors to conduct vein puncture.The main contents of this thesis are as follows:(1)Design and construction of near infrared blood vessel imaging systemIn order to realize in-situ projection of vein image on the measured part,most products on the market adopt spectroscope.The imaging system designed in this thesis removes this component and directly uses projection correction algorithm to adjust the misplaced image projected to the measured part,which not only simplifies the system structure,but also reduces the cost;in addition,through the parameter requirement analysis of key components,the component selection is completed,and finally the whole system prototype is built and debugged.(2)Research on vein image enhancement algorithmsDue to the inevitable hardware and environmental factors,the vein images collected by near-infrared vascular imaging system usually exists some problems such as fuzzy texture and discontinuity.In order to improve the overall clarity of vein image and the effect of image recognition,the principle of classical adaptive histogram equalization algorithm,the limited contrast adaptive histogram equalization algorithm and the guided filtering algorithm were firstly analyzed.Inspired by the weighted guided filtering algorithm,an improved adaptive Canny edge detection operator in the weighting factor was proposed.Experimental results indicate that the enhanced effect of the improved weighted guided filtering algorithm proposed in this thesis is better than that of other contrast methods.(3)Research on vein image segmentation algorithmsIn recent years,U-Net network is one of the best networks in the field of medical image segmentation.In order to extract more clear vein edge information,the U-Net network was deeply studied.For the first time,the U-Net network is applied to the dataset of hand vein image.Firstly,the improved weighted guided filtering algorithm and Labelme software are used to complete the dataset of hand vein image.Then,aiming at the defects of the classical U-Net,such as the loss of detail information and the large semantic gap between encoder and decoder,an improved RRU-Net network is proposed,which adds Res Block and Res Path structure to the classic U-Net network.Experimental results show that the improved network can not only retain the image structure information,but also improve the reliability of network training and reduce the background noise.