Design And Implementation Of Embedded Printing Character Inspection System Based On Qt And Arm NN

Industrial product packaging often contains important printing character information such as product number,serial number,production and deadline,and is affected by uncertainties such as mechanical failures,environmental changes,and transportation wear during product packaging.The printing effect of the printing character may change.As the product package is printed,the printing character needs to be inspected to ensure that the printing character is correctly readable.Products with incorrect packaging information will have a series of adverse effects on the company's production efficiency,logistics warehousing,brand image and other aspects.The traditional printing character inspection method uses manual identification of the quality of the printing character.Due to its disadvantages in terms of efficiency,error rate and labor cost,it is not suitable for the large-scale packaging production line of the enterprise.The existing printing character inspection systems are mostly based on industrial PCs.Compared with embedded ARM platforms that meet the same computing power requirements,they have certain limitations in terms of price,size,power consumption and flexibility.Therefore,this paper designs a set of embedded printing character inspection system based on Qt and Arm NN in the context of practical engineering.The specific work is as follows:(1)The production of the reagent packaging carton,the problem of the production process,the analysis of the requirement analysis of the printing character inspection system are introduced.The overall design of the embedded printing character inspection system is proposed,including the system design overview and the specific content of the embedded hardware and software platform.(2)A position detection algorithm is designed for position detection of moving cartons.After studying the commonly used area location algorithm,a printing character area localization algorithm is designed to locate and extract the printing character area in the system,and the barcode extraction sub-string extracted by the algorithm is barcode-identified.The extracted character substrings are subjected to tilt correction,character division,and character size normalization.(3)The commonly used character recognition algorithms are studied,and the advantages of convolutional neural network in image recognition field,such as feature extraction and recognition rate,etc.compared with the traditional algorithm are analyzed.The classical convolutional neural network lenet-5 is adjusted adaptively and applied to the character recognition of printing code.(4)The Arm NN sdk(Software Development Kit)was researched,the Arm NN sdk environment was constructed and configured,and the trained printing character recognition model was introduced into the ARM end by Arm NN to realize the printing character recognition.(5)In terms of system hardware,the assembly and debugging of each hardware module and the docking with the production line equipment were completed.In terms of system software,the multi-thread structure design,Qt image interface design,detection algorithm design,specific code writing and debugging work were completed.After the system is implemented,the system function test and system real-time test are completed.The multi-model product test has a positive test rate of over 98%,and the complete single-shot test time is less than 1.8s,which basically meets the needs in production.