Design And Implementation Of Beverage Management System Based On Improved YOLOv4 Algorithm

In recent years,the field of object recognition has rapidly developed based on deep learning,and it has been studied and applied fully in many aspects,such as face recognition and product recognition.Beverage is one of the most common commodities,and the improvement of beverage self-service sales model by using beverage identification technology has certain research significance and application value.This thesis self-built a beverage data set.Aiming at the similarity of beverage products and the problem of occlusion when placed,the YOLOv4 algorithm is improved,and the improved YOLOv4 algorithm is used to identify beverages,and the beverage management system is designed and implemented through Python and Qt library.Main tasks as follows:(1)The beverage data set is constructed.The fisheye camera is installed directly above the beverage for data collection.During the data collection process,the position of the beverage is randomly placed.The beverage data set contains 21 types of different bottled beverages,with a total of 3098 images,each with a resolution of 640×480.(2)Provide an improved YOLOv4 beverage recognition algorithm.By adding a channel attention mechanism between the five sets of residual blocks of the basic network CSPDarknet53 of YOLOv4,the network automatically pays attention to important characteristic channel information to solve the problems of occlusion of beverage data and similar color characteristics of some beverage categories,and improve the recognition accuracy of beverage categories.Experiments show that the improved algorithm improves the effect of beverage detection and recognition.(3)Design and implement a beverage management system.The functional modules of the system design include five modules: registration and login module,beverage category information module,beverage identification module,beverage data retrieval module,and system management module.The beverage recognition module is loaded with the improved beverage recognition algorithm.The system adds a database management function,which can save the purchased beverage category information in the beverage management system database,so that the administrator can better query and manage data,and improve the efficiency of beverage management.