Fruit And Vegetable Identification Method Of Supermarket Self-checkout Based On Inception Structure And Residual Module

The rapid development of information technology and the maturity of the logistics and transportation industry have made online shopping more convenient,and more and more consumers have become accustomed to buying daily necessities through the Internet.However,fresh fruits and vegetables are not easy to buy online due to their perishable and fragile characteristics,so consumers mostly choose to go to supermarkets to buy them.At present,the checkout system of supermarkets can only identify products by scanning barcodes.For fresh fruits and vegetables,the supermarket staff can only weigh,pack,and post barcodes,and then scan the codes for settlement,which leads to the need for merchants to invest a lot of manpower and material resources.Resources also bring a lot of inconvenience to consumers.Therefore,how to realize automatic identification of fruits and vegetables is an urgent problem facing the current supermarket self-checkout system.The rapid development of machine vision provides new technical means to solve the self-service settlement of fruits and vegetables,but it is difficult to deploy the existing high-performance convolutional neural network in the self-service settlement system due to the constraints of computing resources and cost.To solve this problem,this research improves the classic neural network Alex Net based on the Inception structure and residual ideas,and proposes a fruit and vegetable image recognition network,named RI Net(Res-Inception Net),and builds it on the self-built fruit and vegetable data set Test,and finally build a hardware platform to realize the functions of the self-service settlement system.The main work of this paper is as follow:(1)Determine the system network topology.According to the functional requirements of the fruit and vegetable self-service settlement system,the self-service settlement system is divided into two parts: an upper computer and a terminal.The upper computer is mainly responsible for information storage and management,and the terminal is mainly responsible for completing product settlement and uploading sales information.(2)Create a supermarket fruit and vegetable image data set.Aiming at the actual application scenarios of supermarket self-checking,we collected images of fruits and vegetables commonly found in supermarkets under different lighting conditions,used image filtering,segmentation,and morphological processing to extract target images,and expanded the data scale by rotating,cropping,and flipping.,Established a fruit and vegetable image data set.(3)Design and improvement of fruit and vegetable recognition algorithm.In view of the complicated supermarket environment,various types of fruits and vegetables,and limited equipment computing resources,Alex Net is used as the backbone network to study the characteristics of the Inception structure and residual network,introduce convolution kernels of different scales,and increase the network width and use"shortcuts" "The connection method maps low-level features to the deep network to improve the structure of Alex Net and build a lightweight fruit and vegetable recognition network RI Net.(4)The realization of fruit and vegetable self-service settlement system.According to the system function,the hardware equipment is selected and the experimental platform is completed.The management of supermarket operation information is realized on the upper computer;the identification of fruit and vegetable commodities and the display of settlement information are realized on the settlement terminal;finally the system is comprehensively performed on the experimental platform Functional testing to verify the feasibility and scalability of hardware systems and algorithms.