Research On The Application Of Machine Vision In On-line Measurement Of Workpieces

With the birth of intelligent manufacturing and industrial 4.0,not only placed higher requirement the machining accuracy of workpieces but also brought great innovation to workpieces detection.Based on the traditional measurement methods,due to the inconvenient of the testing tools,there are defects such as high working intensity,low accuracy and low efficiency;while the high precision measurement machines can meet the requirement of accuracy,but they are not suitable for mass production measurement.Therefore,the method of workpieces detection based on machine vision which overcomes these problems,It is a hot spot for the current research.In this paper,a non-contact online workpieces measurement system based on machine vision is developed to meet the requirements of workshop industrial automation,achieved the measurement of key parameters of small shaft workpieces.This system overcomes many shortcomings of traditional contact measurement and improves detection efficiency.The main tasks of the thesis are as follows:Firstly,according to the processing and testing conditions of the shaft workpieces,the overall structure of the measurement system is designed,and finished the selection of the hardware equipment and designed the software workflow planning based on the system requirements.At the same time,this provides hardware support to ensure the quality of image.Secondly,studying the filtering and eliminate noise methods,the threshold segmentation method and the edge extraction algorithm of workpieces image based on the conventional image processing method in machine vision.Ultimately,using the median filtering method to eliminate noise of the image,selecting the SIFT feature point algorithm stitching images and segmenting the binary image with the adaptive threshold Otsu method(Otsu method).At the same time,the traditional Canny edge detection algorithm is improved to improve the pixel edge quality.Thirdly,the concept of sub-pixel edge segment is introduced to solve the problem of low accuracy of pixel level edge detection algorithm.The principle of sub-pixel edge detection is discussed,solved and reasoned the several common sub-pixel edge detection methods,the Zemike moment method,polynomial fitting method and polynomial interpolation methods were experimented to analyze the measurement accuracy,efficiency and anti-noise resistance.Finally,Zernike moment method is chosen.Fourthly,according to the structural characteristics of rotary parts,the parts are divided into external screwthread,step shaft,arc curve shaft and cone shaft,and studied the corresponding visual measurement methods respectively.Finally,making the design and experiment of some programs,including camera calibration,workpieces image import,deletion,image preprocessing,edge detection and dimension parameter detection.Through the measurement of a typical part had verified the above detection algorithm.The results show that the measurement accuracy is within the range of 0.018mm,which meets the measurement accuracy of 0.02mm.It is proved that the system is feasible and practical by analyzing the inaccuracy of measurement results.