Research On Key Technologies Of Part Dimension Online Visual Measurement Based On Deep Learning

Part dimension measurement is one of the most basic quality control methods in industrial production.Commonly used contact measurement methods are complicated,time consuming,and incomplete.With the development of machine vision,visual measurement is increasingly valued due to its high measurement efficiency.Applying machine vision to measure part dimension,image preprocessing and edge detection are the two most critical steps.The current measurement methods have the following two problems:(1)Due to the complicated environment of the machining site,the surface of the part is likely to adhere to irregular interference areas such as cutting fluid and chips,which affects edge detection.(2)The surface texture of the part is rich,but the traditional operator detection is poor in texture and the adjustment is difficult to determine.In order to solve the above problems,this paper researches the key technologies of Part Dimension Online Visual Measurement based on deep learning.And it aims to solve the problem that the traditional visual measurement method has high requirements on the quality of the original picture and high requirements on the detection environment.The main research contents are as follows:(1)This paper first expounds the research status of visual measurement and deep learning in the field of image processing and convolutional neural network in industrial inspection.Then a part size online vision measurement system is built.Aiming at the problem that the surface of the part is prone to interference areas such as cutting fluid and chips,FCN is used to realize the recognition of image interference area.And the interference area is processed by the directional texture repair method.Finally,the removal of interference areas such as cutting fluid and chips is realized which is used as a preprocessing method for edge detection to reduce the influence of interference area on the edge detection.(2)Aiming at the problem that the traditional operator detection edge detection method has poor in texture and difficult to adjust the parameters,a method based on HED for rough image edge detection of part image is proposed.The image edge is obtained by HED network model,and then the edge is post-processed based on the non-maximum value suppression and double threshold processing methods in the Canny operator to obtain a refined edge image.In order to further improve the accuracy of the dimension measurement,sub-pixel level edge detection accuracy is achieved by cubic spline interpolation.Finally,the method is validated by a shaft part.The experimental results show that the proposed algorithm can effectively detect the interference region and process it.The edge obtained by the edge detection algorithm based on HED can avoid the influence of surface texture.The measuring accuracy of the outer diameter of the part can reach 0.02mm,which can effectively meet the requirements of fast semi-finishing inspection.In summary,this research provides a common part dimension measurement method that reduces the constraints of the input image on the detection target.At the same time,it provides methodological guidance for the application of deep learning in industrial detection.Through the above work,the part dimension online measurement realized,which reduces the restriction of the original image on the industrial site to the detection target,improves the versatility of the processing method,and provides reference for the application of deep learning in other industrial inspection fields.