Research On Medicine Box Detection And Drug Inventory Management System Based On Machine Vision

With the rapid development of society,the public has higher demands on the quality and efficiency of medical and health services.A hospital cannot be short of pharmacy,it's very major for its automation,information and intelligent upgrading.Automated pharmacy in developed countries early development,product maturity and high popularity.In recent years,more big hospital buy automated pharmacy equipment,domestic pharmacy automation system research and development and application is developing rapidly.In this paper,not only analyzes the development process of automated pharmacy at home and abroad,but also analyzes the present situation of robot vision technology,and understands the development trend of automated pharmacy,the key technologies involved,and the domestic market demand.So this paper makes theoretical analysis and experimental study on the layout plan of automated pharmacy,vision system of medicine picking robot,drug discern and stock control.First of all,according to the characteristics of large movement range,fast operation speed and rotary operation,the design plan of Cartesian coordinate robot is determined.The robot has merit of long travel,high-accuracy,fast velocity,easy expansion and mend.By analyzing the tenet of binocular vision,the robot vision uses "eye in hand" parallel binocular vision,and decides the visual system parameters.Secondly,the linear imaging model,distortion model and the filiation between picture dots and object dots are built.Based on opencv3.4.3,the program design of Zhang's calibration way is done by Python,and the binocular vision calibration is fulfilled.Then,the actual kit was tested.By using the self-designed image processing program,the 3D reconstruction is carried out with the centroid point of the medicine box as the feature point to realize the detection and positioning of the medicine box,and gained good results,verified the design of the binocular system can run and use.Then,the methods of image preprocessing,text detection,drug name region determination and drug name character recognition are analyzed theoretically and experimentally.Gray scale and skew correction are used to preprocess the image,and the text detection is based on the craft algorithm.According to the characteristics of the drug name area on the surface of the drug box,an accurate method to detect the drug name area is proposed;then,the aspect ratio of the oblong domain is sorted to avoid the disturbance of the pattern on the medicine box.After the drug name region was determined,Tesseract,CRNN and Baidu OCR were used to carry out the drug name recognition experiment,and the Py Torch framework was used to build the CRNN network structure to conduct deep learning training for the synthetic Chinese character set,and the preliminary results were obtained.Finally,in the windows system,combining with Python,MYSQL and pyqt5,the pharmacy drug inventory management system is preliminarily designed,the personnel management,drug management and operation record management functions are realized,and the system debugging and function verification are completed.