A Study Of Bolt Rough Detection Technology Based On Machine Vision System

In the industry fields as well as daily life,fastener connections can be seen everywhere and the bolt is the most popolar one.Bolt has a simple structure and is often overlooked.In real life lots of accidents have happened because of bolt failure which brought great losses to our society.With the advancement of science and technology,machine vision technology has made rapid development.As a new method,machine vision has lots of characteristics such as non-contact,fast,ecnomical,continuous,flexible and so on,which are suit for online shape detection of quantities of workpieces.For the above reasons,machine vision is widely used in the mordern factories.To ensure the quality of the bolts that have finished,we must first ensure the dimensional accuracy of them.For the small size bolts,the detection technology based on machine vision has already been used by many factories,the accuracy and error rate can meet the industrial demand.It is a post-test which throws away the substandard products as waste and can not monitor or adjust the process of production.To reduce the quality cost,the process control is particularly important.For the large size bolt rough which comes from the cold heading machine,successful machine vision detection exmples are hardly found due to the size and environmental factors.The research on bolt rough visual detection is very necessary.This paper deeply analyzes the research objectives and research contents.The bolt rough detection has several notable characteristics including high precision,high speed,and poor environment.Taking these characteristics into consideration,this paper designs bolt rough testing program.Daheng high-resolution digital cameras are selected to get clear pictures of the test bolts.According to different lighting ways of visual system,two measurement solutions are proposed.In terms of image processing,firstly,a multiple filter is designd,using the combination of Gaussian filter and a improved median filter to filter random noise;Then,by comparing a variety of operators,first-order Sobel operator is used to sharpen the image.Again,binarized the image through Otsr method,track counter counterclockwise and extract the boundary contour.Finally,hough transform and freeman chain code algorithms are used to extract characteristics parameters.Through the analysis and comparison of the experimental results of the two solutions,this paper chooses solution two with hough transform to measure the workpiece,measurement error is less than 0.05mm.Based on the above,the article takes VC++ as a platform and MFC as the basic framework to develop experimental detection software.The software is designed in a modular approach,user-friendly and easy to operate.The software includes functions such as image acquisition,analysis,results show,data management of information system and other functions.