Multi-level Detection Of Injection Molding Products Defects Based On Hybrid Features

Intelligent manufacturing refers to the realization of intelligent manufacturing in the manufacturing industry.It applies artificial intelligence to the manufacturing industry,and intelligent systems perform inference,analysis,judgment,and decisionmaking in the manufacturing process.Injection molding is one of the main methods of plastic manufacturing.Injection molding is a complex production process,which is affected by different parameters and complex external factors at each injection stage.Therefore,various types of injection-molded defects often occur on plastic products,which directly affects the quality of the products.Applying machine vision technology to the defect detection of injection-molded products can improve the detection accuracy and efficiency,which is an important part of achieving intelligent injection molding.At present,the intelligent detection technology of injection-molded product defects is not effective enough,which is difficult to adapt to changes of the injection molding production line and detect various types of injection-molded product defects.The purpose of this paper is to realize the real-time intelligent detection,classification and identification of defects in injection-molded products during the injection molding process.The defect area extraction,defect feature extraction,defect classification and identification of injection-molded products are studied.The first chapter introduces the injection molding production and the defect detection of injection-molded products.It also analyzes the recent research of product defect identification,injection molding process and equipment and defect identification of injection-molded products.Finally,it explains the meaning,main content and organizational structure of this article.In the second chapter,a two-dimensional neighborhood search algorithm for defect area extraction of injection-molded products is introduced in detail.The twodimensional neighborhood search algorithm solve the problem of micro vibration in actual production of injection molding.Through the comparative analysis of different threshold segmentation methods,the Otsu's threshold method is selected to binarize the image to segment defect regions.In the third chapter,the progressive analysis based feature of defects in injection molded products extraction method is proposed.The common injection-molded product defects are introduced and their characteristics are analyzed.Based on progressive analysis of defect category characteristics,three types of features are extracted.The mixed location feature of defects is extracted based on center matching.The local gray feature is extracted based on morphology.The shape features of defects is based on trajectory fitting.The fourth chapter introduces the process and method of classification and identification of defects in injection-molded products.A multi-level tree classifier based on SVM is proposed for defect classification of injection-molded products.The decision tree classifier and support vector machine classifier are briefly introduced.The model structure of the SVM multi-level tree classifier is described.According to the formation mechanism of several types of product defects,a quantitative description of the size evaluation of each type of defect is proposed,and the calculation process is explained in detail.The fifth chapter introduces the hardware configuration of the defect recognition system for injection-molded products and validates the method in this paper.Based on the methods and techniques proposed in this article,a software system for defect detection for injection-molded products is developed.The functions and application examples of the system are introduced.The validity of the method is verified by experiments.The sixth chapter summarizes the full text,and gives further research directions and work prospects.