Design And Implementation Of Intelligent Inspection System For The Appearance Of Logistics Packaging Cartons For H Company

With the rapid developing of social economics,modern logistics industry has also developed to an unprecedented level.Packaging,as one of the important guarantees in the circulation of goods,has been paid more and more attention.Carton as a kind of packaging,because of its high cost performance is widely used.H company is a famous manufacturer of electronic products.The delivered goods are packed in cartons.Now H company has decided to introduce machine vision to realize automatic warehousing.This process needs to identify important information such as external labels of cartons and ensure that the appearance of cartons is clean without damage or serious deformation.Due to the number of containers,most of the value of a single box is high,before warehousing must carry out appearance inspection;Considering the slow speed of manual inspection,it is easy to cause visual fatigue and affect the storage efficiency.In view of the above problems of H Company,this paper designed and realized an intelligent inspection system for the appearance of logistics packaging cartons of H Company,which reduced labor intensity,reduced manual input and improved warehousing efficiency.Analysising of the logistics carton appearance detection in this paper,the difficulties are summarized as follows: 1.This system is multi-faceted detection,especially the detection of the bottom surface is difficult to achieve,because the bottom surface of the carton is in close contact with the conveyor belt,and the conveyor belt is opaque and cannot see through;2.The damage of carton appearance defects is small and difficult to detect;3,the direction of the carton on the conveyor belt is irregular,and may be vertical,parallel or tilted at a certain Angle with the direction of the conveyor belt.In view of the above difficulties,this paper designed and implemented an intelligent inspection system for the appearance of logistics packaging cartons of H Company,mainly including the following research contents: 1.Automatic inspection of the appearance defects of logistics packaging cartons based on image correction processing.The appearance image of carton was captured by industrial camera,and the feature was extracted by YOLOv4 network model after correction processing.Multi-scale prediction based on k-Means ++ clustering prior frame;Based on boundary regression,the location prediction and category prediction of three kinds of defects on the appearance of logistics carton are realized.2.A comparative experiment was conducted on the core functional modules of the algorithm in this paper.After adding two modules of image correction processing and prior box estimation,the algorithm in this paper improved the detection accuracy of small size defects such as damage,and verified its effectiveness.In this paper,the system can realize the intelligent detection of carton appearance,storage and query of detection results,and analysis of detection data,to provide a certain reference for management.In this paper,the core functions were tested,and the average detection accuracy of damage defects reached 95.43%,deformation defects 98.85%,dirty defects 99.57% and 97.95%,and the number of detection frames per second reached 27,meeting H Company's recognition accuracy requirement of not less than 95%.