Abnormality Detection Of Mold Surface In Iniection Molding Process Based On Hybrid Feature Fusion

With the wide application of plastic products,plastic processing and molding technology has received widespread attention.Injection molding machine is the most important plastic products manufacturing equipment,when the injection products such as the failure to fall and other production anomalies,the mold is very easy to damage,thus seriously affecting the production.Traditional injection mold anomaly detection method accuracy or efficiency is not enough,cannot meet the actual production needs.Mold protection system based on machine vision with a high degree of automation become more and more widely to be used in actual production.In addition,there is always obvious vibration in the process of injection molding machine.Most of the researches based on the machine vision injection mold anomaly detection have not considered the detection effect of the system in the vibration environment.When it is directly applied in the vibration environment,the anomaly detection accuracy is low.In this paper,the purpose is improving the accuracy of anomaly detection of injection mold under high amplitude environment and the research on image registration and mold anomaly detection.The first chapter analyzes the object and target of the anomaly detection task of the injection mold in this paper,puts forward the difficulty of the task of this paper,and uses the technology of image registration and mold anomaly detection to accomplish visual inspection tasks.The significance of the research on anomaly detection of injection mold is introduced.The research status of image registration and anomaly detection of injection mold is analyzed.And finally the main content of the composition and organizational structure of the framework is introduced.In the second chapter,the content of SIFT descriptor in corner feature extraction and image registration based on SIFT feature are introduced in detail.The problem of high amplitude vibration of mold production is solved by using the image registration of the surface of injection mold.Analyzing the feature-based method in image registration,the method based on feature points is selected as the basis of the image registration of the surface of the mold.In the third chapter,aiming at solving the problem of low accuracy of local registration in high-amplitude injection mold surface detection,a hybrid feature-based image registration is proposed.The global feature GRGC descriptor sub-model and the global vector GRGC-SIFT construction are proposed.Further,the parameters of the GRGC-SIFT descriptor and the results of the registration of the mold surface image based on the GRGC-SIFT vector are discussed.The fourth chapter presents the method based on image registration difference injection mold anomaly detection.The method extracts the abnormal area of the surface image of the injection mold through the use of transmission transformation and template difference.Firstly,the eight-point method is used to calculate the transformation matrix between the template image and the real-time image.Then,the real-time image is converted by the transformation matrix to be differentiated from the template image.Binarization is used to extract the difference between the template image and the live production image after difference processing.Finally use the open operation to remove the noise to get the final difference image.The fifth chapter introduces the hardware configuration and performance evaluation of the injection mold anomaly detection system designed in this paper,and realizes the detection method of the high amplitude injection mold anomaly based on the mixed feature fusion proposed-in this paper.Finally,the effectiveness of this method is verified.The sixth chapter summarizes the full text and puts forward constructive prospects for the follow-up research.