Research On Surface Defect Detection Method Of Special Steel Bar Based On Machine Vision

As one of the raw materials of important industrial equipment parts,special steel bar is an important product in metallurgical industry.During production and processing,various external defects will inevitably occur on the surface of special steel bar due to factors such as rolling equipment and processing technology,which will affect the appearance of products and change the properties of special steel,resulting in toughness,Bending strength and wear resistance decreased.As the market’s requirements for products increase,the surface quality of raw materials has also become critical.In order to effectively control the delivery quality and improve the production process of special steel bars,study the causes of surface defects,it is particularly important to detect the surface defects of special steel bars accurately and efficiently.At the same time,conducting relevant research studies is great beneficial to reduc production costs,improve working environment,and improve detection accuracy and production efficiency.Compared with traditional automatic detection methods such as ultrasonic,magnetic flux leakage,eddy current,etc.Machine vision inspection is widely used in surface defect inspection due to it has many advantages,such as non-contact,accuracy,continuity,repeatability,intelligence and automation.Based on the analysis,research and summary of the research status and achievements in related fields,the research objectives and contents of this thesis are formulated by combining the research on surface defect characteristics of special steel bar with the subject requirements,as follows:Defect detection test bench designing: Deeply studying on the composition and working principle of the existing machine vision system.According to the surface defects characteristics of special steel bars,the image acquisition environment and the technical indicators of defect detection,the camera,lens and other image acquisition units were designed and selected.Through analyzing the morphological characteristics of the special steel bar,the cause of the serious reflection was determined.In order to weaken this effect and improve the original image quality,a large number of different types,colors,sizes lights and different lighting arrangements were compared.Choosing the most suitable light source and lighting method for special steel bars;designing a machine vision inspection system that is easier to form a clear surface image of special steel bars.Image preprocessing algorithm designing: Since the shape of the special steel bar is arc and the surface is curved,the uneven illumination in the original image acquisition process will cause the image to be brighter in the middle,darker on both sides,and the image grayscale is uneven,and due to the acquisition and transmission of the original image During the image process,various noises will be introduced,which seriously interferes with the observable information of the image and adversely affects the image detection.Based on this,a preprocessing algorithm for the special steel bars is proposed.First,the original image is segmented linear gray scale transformation to achieve the purpose of uniform image gray;second,the special steel bar image is subjected to homomorphic filtering to achieve the purpose of filtering and denoising;finally,morphological operations and edge detection are used for the bar image to achieve the purpose of extracting defects.Surface defect detection algorithm designing: Aiming at the characteristics of complex surface texture of special steel bars,uneven illumination,uneven image gray level,etc.,this thesis proposes a visual detection algorithm for metal bar surface scratch defects based on the combination of non-subsampled shear wave transform(NSST)and morphological theory.First,using NSST to decompose the collected original image,then,using anisotropic diffusion and improved adaptive gamma correction to filter and enhance the decomposed high-frequency components,the low-frequency components use the scale space to subtract the bottom scale space and the scale space of each layer and then perform weighting and normalization;then using the NSST inverse transform to reconstruct each component to obtain a uniform background and less noise image;finally,combining the insensitivity of morphological processing to noise and the accuracy of Sobel operator extraction and retention of edge features,using perform morphological opening operation,Sobel operator edge detection and morphological closing operation on the reconstructed image;Finally,the Hough line detection is used to complete the detection of the number,size,and position coordinates of the surface defects.Surface defect detection system designing: In order to simplify the user’s operation process,a set of surface defect detection system with intuitive and concise interface,complete functions and easy to operate is designed based on Matlab.The system includes image display module,data management module,defect detection control module and defects the data feedback module,this can simplify the inspection process to a great extent and realize the online intelligent inspection of special steel bars.