Design of Multi-sensor Vein Data Fusion Blood Sampling Robot Based on Deep Learning

Abstract

In 2020, the new crown epidemic swept the world, and the world's medical resources will be significantly challenged. Medical staff manual venipuncture has a low accuracy rate and is prone to cause cross-infection. Automatic puncture blood collection robots have the advantages of high automation and accurate puncture. However, the existing automatic puncture blood collection robots for vein recognition have low control accuracy, large volume, and modalities. Single and other issues. Based on the difficulties faced by traditional venipuncture methods, an automatic venipuncture robot based on deep learning was developed to improve the success rate of one-time venipuncture. The YOLO-v5 target detection module is used to locate the blood vessel coarsely, while the U-Net segmentation module accurately divides the blood vessel area to determine the location of the puncture point and its direction. To obtain the depth and area information of blood vessels, we use the algorithm of OpenCV. The stepper motor and servo motor in the actuator are closed-loop controlled by the bus motion controller, and the puncture training arm model is used to perform puncture tests to verify the performance of the robot. Tests show that the robot has high puncture repeatability and accuracy, which meets the application requirements.