Research Of Arm Vein Detection Module Based On Deep Learning

In order to realize the full automation of vein puncture about human arm,this paper proposes an automatic detection algorithm module based on deep learning.The module contains two sub modules,which are responsible for the detection of veins and the location of the puncture point respectively.First,in this paper,the Faster-RCNN object detection module based on deep learning is used to detect the vein area for several kinds of vein images(raised,ordinary and blurred veins)which are difficult for traditional object detection algorithm to deal with at the same time.Then through the puncture point positioning module,the vein structure in the vein area is refined,and the location of the puncture point and the puncture direction are determined by the method of image enhancement and linear fitting.The experimental results of 1200 images in the test set show that the classification accuracy of raised vein area,ordinary vein area and blurred vein region is 0.998,0.993 and 0.960 in turn,and the average time of testing one picture is 0.335 seconds.It shows that this method meets the clinical requirements for the accuracy and real-time performance of human arm vein,and can be used as an auxiliary method for automatic puncture of veins.The research work carried out in this paper can be summarized as follows:(1)We investigated the current research status of vein puncture assisted technology and vein detection technology at home and abroad.The shortage of traditional vein detection methods are summarized,they are complex algorithm structure,low degree of automation,and t weak generalization ability of the model.(2)The relevant theory of object detection is deeply studied.The representative feature model and classification method in the traditional object detection method are summarized and analyzed.SIFT and HOG characteristics as well as the classification methods such as k-NN,SVM and AdaBoost are emphatically studied.Then,the framework of deep learning is introduced,and the theoretical basis of the object detection algorithm based on depth learning is deeply studied.The Faster-RCNN object detection algorithm improves the method of feature extraction and regional recommendation in the traditional object detection algorithm,realizes the full automation of the object detection,and shows great advantages in the accuracy,robustness and generalization ability of the detection.(3)A vein automatic detection module based on deep learning is proposed.The module contains two sub modules.The first sub module detects all kinds of vein areas through the Faster-RCNN algorithm.Second sub module extracts the vein structure from the vein area through the traditional image algorithm,and finally determine the best puncture point and direction of the puncture.(4)The performance of automatic vein detection module is tested.The test performance of the module is evaluated by accuracy,recall and running time.The test results show that the automatic vein detection module proposed in this paper has achieved the technical requirements of the industrialization application in the degree of automation,accuracy,efficiency and robustness.