Research On Flange Surface Defect Detection Based On Machine Vision

Flange is a common sealing connection parts,with convenient connection,can bear the characteristics of large pressure,widely used in various industrial pipelines.However,in the flange production process,due to casting materials,technology,transmission process bump and other reasons in the flange surface often appear trachoma,scratch,scratch and other defects,not only affect the product appearance,but also cause flange leakage,serious can lead to production accidents.Therefore,how to quickly and accurately detect flange surface defects,improve and control flange product quality,has been the focus of the manufacturers.At present,most flange surface defects are still detected by manual visual inspection,but manual inspection is slow,low accuracy rate,subject to subjective influence and no accurate defect quantification.With the rapid development of machine vision,automatic detection method based on machine vision has gradually become an important means of surface defect detection.Therefore,this paper studies the flange surface defect detection system based on machine vision.The main contents of this article include:a)Design of flange surface defect detection system based on machine vision.Firstly,the overall structure of the detection system is designed and the basic workflow of the system is described.Then,according to the reflection characteristics of flange surface and surface defects,as well as the basic requirements of flange surface defects detection,a diffuse reflection lighting scheme is proposed to collect high-quality images of flange surface.Then according to the flange manufacturers to flange surface defects detection requirements,the camera,lens and other key hardware parameters calculation and selection.Finally,a visual inspection system for flange surface defects was built to complete the tasks of image acquisition and defect detection classification.b)Study on image preprocessing method of flange surface.In order to ensure the efficiency and correctness of the subsequent algorithm,gaussian filtering was used to reduce noise,and the image was processed to determine whether there was a flange and to locate the position of the flange.A template matching algorithm based on normalized cross-correlation is used to automatically construct a universal template according to the flange structure information,and an image pyramid is used to accelerate the image positioning.Experimental results show that this algorithm can accurately and effectively judge the presence of flanges in images and quickly locate flanges.c)Research on flange surface defect detection algorithm.By analyzing the types and characteristics of flange surface defects,the flange surface defects were divided into end face defects and contour defects according to the shape and distribution of defects,and the algorithm was designed to locate the defects.In order to reduce the interference of pixels outside the area to be detected on the subsequent algorithm and obtain the stable contour of the area to be detected,the region to be detected segmentation and watershed extraction based on marked watershed were carried out on the flange positioning image,which were respectively used for the subsequent positioning of candidate end face defects and candidate contour defects.For end face defects,aiming at the problem that traditional Sauvola local threshold algorithm will produce over-segmentation because of water stain noise,global gray information is introduced to effectively solve the over-segmentation problem.For contour defects,the circular fitting method based on least squares was used to fit the contour,construct a standard circle,and set threshold segmentation to obtain the selected contour defects.d)Research on feature extraction and classification of flange surface defects.The defect features of flange end face were analyzed,and the geometric features,gray features and texture features of defect images were extracted.The BP neural network was used to train and classify the defect images.The characteristics of flange contour defects were analyzed.In order to simplify the representation of the contour,the original contour defects were transformed into peaks in the polar coordinate system by polar coordinate transformation.By extracting the features of peaks,the classification method based on feature threshold was used to realize the classification of contour defects.e)Test and analysis of flange surface defects on site.According to the designed visual inspection system for flange defects,an experimental platform for flange surface defects detection was established.The collected flange images were preprocessed,the area to be detected was segmented,and the defects of the end face and contour were extracted and classified.The detection algorithm program was written and the humancomputer interaction interface was designed.Experimental verification and analysis were conducted on the accuracy and efficiency of the defect visual detection algorithm proposed in this paper.The experimental results show that the modified algorithm can meet the technical indexes of defect detection and classification recognition of flange manufacturers correctly and efficiently.