Research On Key Techniques Of Hand Dorsal Vein Injection Based On Machine Vision

Since the outbreak of COVID-19,more than 230,000 health care workers worldwide have contracted novel coronavirus infections as a result of inadequate protection.Non-human contact way of the automatic injection of palm-dorsal vein image,with its non-human nursing ways of contact and the function of the nursing process for automatic injection,was welcomed by many health care workers,and between medical staff and patients to non-contact automatic injection of palm-dorsal vein image,the main content is: the detection and segmentation of palm-dorsal vein image,and injected into the needle point location decision feedback.In this paper,an image processing algorithm AT-U-Net(Attention-U-Net)based on improved U-Net guiding Attention mechanism is proposed for the detection and segmentation of dorsal hand veins.By introducing the attention mechanism,long-distance detection and segmentation of dorsal hand veins can be realized,and clear images of dorsal hand veins and vessels can be provided for subsequent decision of needle insertion location.The experiment was carried out according to the self-built dorsal hand vein database,and the accuracy rate reached 93.6%.According to the detection and segmentation results of dorsal hand vein,a PT-Pruning(Point-Pruning)method based on improved Pruning algorithm was proposed in this paper.PT-Pruning was used to extract the main lines of dorsal hand vein.Considering the cross-sectional area and bending value of each vein,the best matching insertion point of dorsal hand vein was obtained comprehensively.Compared with the self-built dorsal hand vein injection point database,the detection accuracy of effective injection point injection area reached 96.73%,and the detection accuracy of the most matched injection point injection area reached 96.5%.In this paper,two kinds of detection methods are proposed for the feedback detection problem of hand back vein.One is to judge whether the puncture is successful by detecting whether there is blood return through penetrating a certain depth of the skin.The other is to judge whether the injection is successful by detecting whether there is bulge in the process of injection.Firstly,aiming at the problem of blood return detection,a hand intravenous blood return detection device based on near infrared light was proposed,which could prompt doctors and patients to observe blood return situation through multiple data fusion.Through experimental verification,the accuracy of blood return detection is 81% Secondly,to solve the problem of swelling detection,based on the self-built data set of dorsal hand vein swelling,a swelling detection algorithm based on improved Yolo V4 was proposed.The change of swelling was observed through the linear change of laser bar at the insertion point in the dorsal hand region.If the swelling area appeared,the injection was immediately stopped.Through experimental verification,the detection accuracy of bulge is 99.36%,which lays a solid theoretical foundation for the subsequent mechanical automatic injection.