Design On Intelligent Refrigerator Information Monitoring System Based On Deep Learning

With the advent of the Industry 4.0,the country takes intelligent manufacturing as its strategic lead,aiming to promote the transformation of traditional manufacturing to intelligence through the integration of emerging technologies such as intelligent technology and information technology with traditional manufacturing.Refrigerators are an important part of home appliances,and intelligent management is one of the main development directions of the refrigerator industry in the future.Nowadays,many new "pseudo-intelligence " refrigerators are emerging one after another,which has caused great interference to consumers’ purchases.For this reason,the country has given a standard definition of intelligent refrigerators and clarified the requirements for intelligent performance and effects.Therefore,it is of great practical significance to research a complete set of intelligent management refrigerator system.Based on the Io T model architecture of intelligent home appliances,a set of intelligent refrigerator information monitoring system based on deep learing and "endcloud-end" technology model has been designed in this thesis,including three parts:monitoring terminal,cloud platform,and user end.Based on the improvement of the classic algorithm YOLOv5,a new ingredients detection algorithm P2-CBAM-YOLOv5 is proposed.In order to realize the ingredients classification and counting functions of the system,this thesis first studies the model structure of the baseline algorithm YOLOv5,makes a visual analysis of the baseline network feature map of the model,explores the problems existing in the baseline model in ingredients detection,and proposes an improvement plan.First,a feature detection head P2 is added to improve the effect of tiny target detection,and then the hybrid attention mechanism CBAM module is integrated into the model to improve the recognition ability of key details.The results show that the m AP＿50 index of the improved model P2-CBAM-YOLOv5 on the ingredients data set Ingredients301 reaches 94.99%,which is 2.11% higher than the baseline model YOLOv5.In order to complete the design of the intelligent refrigerator information monitoring terminal,according to the functional requirements of the project,the communication module,sensor module,image acquisition module and relay module are selected around the STM32 central processing unit,and the driver program of each module is written based on Keil-MDK5 software.At the same time,the data frame for communicating with the cloud platform is also designed.In order to realize the design of the remote monitoring cloud platform,the overall framework of the cloud platform data processing center is built.Among them,the Qt host computer is used to communicate with the monitoring terminal;the ingredients recognition module is used to detect images and send the recognized ingredients information to the database;the database is responsible for receiving the environmental data and ingredients data in the refrigerator.In order to realize the design of the client App,the Android system is used as the platform to complete the design of each module of the App in this thesis.As the carrier of the "endcloud-end" intelligent refrigerator information monitoring platform,users can query and update the information in the refrigerator anytime and anywhere on the mobile terminal under the Android system.Finally,the hardware and software platform is built and the joint debugging of the system is carried out.Including the test of the data collection and transmission function of the terminal,and Android remote control test.The test results show that the monitoring system can stably realize data collection,data uploading to the cloud,and data storage.At the same time,the reliability of the system in terms of ingredients detection and the real-time performance of the system in updating information also reaches the expected goal.