Research On Application Of Machine Vision Technology In Mold Inspection And Positioning

The industrial automation of our country is gradually improving at present,but many manufacturing enterprises still have a large number of non-automated equipment,which are the key factors affecting the overall production efficiency of enterprises.Machine vision technology is mainly used to improve the flexibility and automation of production,but there are many restrictions on the layout and use of machine vision systems in these traditional machine tool equipment.In this paper,the research work is carried out from the two aspects of automatic detection of mold hole position and the design of the automatic function module of mold processing equipment.The purpose of mold intelligent processing is to provide a research direction for the automatic transformation of traditional equipment.This article firstly conducted an in-depth investigation of the application status of machine vision technology in industrial production,learned about some of the commonly used target detection algorithms in the mold production process,and then used the Hough transform algorithm for specific two-dimensional shape detection to mold The detection of the hole position was tested.Experimental analysis shows that this type of algorithm can not effectively complete the mold hole detection task in the absence of a strict scene light source,so this paper proposes to use a deep convolutional neural network target detection algorithm to detect the mold hole position.Because the deep learning model needs to use a large amount of data for model training,this paper created a data set of mold hole positions by manually labeling.In the process of data set production,data enhancement technology is used to improve the quantity and quality of image data.This paper completes the engineering code related to model building,model training,and actual model detection based on the CenterNet original model,and then improves the design of the model for the original model with high complexity and excessive resource consumption.The optimized mold hole position The parameters of the detection model are only 3.2%of the original model.Aiming at the problem that the optimized model suffers from a decrease in detection performance during training,this paper introduces the training idea of knowledge distillation and uses the original model to guide the optimized model for training,which improves the performance of the optimized model.The machine vision system is composed of algorithm software and system hardware.The mold detection algorithm designed in this paper cannot be directly deployed to traditional equipment,and some necessary auxiliary function modules need to be added.In order to realize the intelligent processing of mold,three main functional modules are designed in this paper,including machine vision module,clamping module and positioning calibration module,which are used to cooperate with the detection algorithm to realize the automation of mold product processing and production.