Study On Visual Detection System Of Surface Defects Inside The Pressure Vessel

It can be said that the pressure vessel is one of the most important equipments in the industrial era.During the service of the pressure vessel,the environment in which it is located is often accompanied by high temperature,high pressure,high corrosive medium and the vessel are faced with the potential risk of damage to pressure vessels due to cracks,deformation,corrosion holes,etc.If timely and effective testing is not carried out,it will lead the company to stop production or even to explosion,which will bring serious dangers to people’s lives and property.The traditional non-destructive testing method is not only inefficient,but also can not eliminate the safety hazard of detecting.Therefore,we study the feasibility of several image processing algorithms in the detect system of the surface defects inside the pressure vessel based on the id-depth uderstanding of machine vision system.Its purpose is to design an effective visual detection system of surface defects inside the pressure vessel so that the production safety of the enterprise and the personal safety of the staff can be guaranteed.Compared to traditional pressure vessel non-destructive testing technology,machine vision inspection methods have unparalleled advantages in terms of safety and speed.Combined with the application research of the existing machine vision system in the field of defect detection,a visual inspection system for surface defect inside the pressure vessels capable of real-time online detection is proposed.Based on the in-depth understanding of the machine vision detection system,innovatively proposed a design scheme for visual inspection system of surface defects inside the pressure vessels based on multi-lens image acquisition device.What’s more,combined with the actual working conditions,the idea about using fisheye lens instead of the traditional lens for image acquisition is proposed,and the appropriate fisheye lens and USB modules are selected to form the image acquisition device.At the same time,according to the conditions inside the pressure vessel,the strip LED light was uesd as the system light source for forward illumination.In order to effectively detect defects,this paper also designed a series of lighting experiments in terms of illuminance and irradiation angle which proved that better results can be obtained when using proper light intensity and using low illumination angles.Finally,the software workflow of the visual inspection system is proposed,including image acquisition and stitching,data comparison and point distance measurement,which can complete the qualitative and quantitative detection of defects.Image preprocessing and fisheye image processing are necessary preparation for the next image stitching step.In this subject,the image preprocessing algorithm mainly includes image filtering algorithm and gradation transformation algorithm.In the fisheye image processing experiment,the results show that fast and effective processing results can be obtained when scan-line filling method be used to extract the effective area of the fisheye image and the latitude-longitude projection used to correct the distortion.Due to the poor lighting conditions inside the pressure vessel,it is easy to produce salt and pepper noise during the image acquisition.Therefore the system compares the performance of mean filter,median filter and adaptive median filter in different noise environments throng experiments which shows that the adaptive median filtering algorithm has better processing effect and is not affected by noise density.Due to the influence of structure and illumination,the contrast of the collected image area is low and the defects are not obvious.Histogram equalization is performed on such images to obtain a strong contrast image and high recognition defects which prove the necessity of grayscale transform algorithm in image preprocessing.The research of image stitching algorithm is the key to the design of visual detection system for surfuce defects inside the pressure vessel.Based on the in-depth study of the principle of image stitching algorithm and the consideration of real time,a image registration algorithm based on SURF feature extraction has been proposed to achieve fast registration of defect images.In order to achieve the effect of exact matching,the nearest neighbor and the next nearest neighbor method are used to eliminate the excessive mismatching point.At the same time,the Ransac algorithm is introduced to further filter and calculate the optimal transformation model of the image.Due to the difference between the hardware deviation of the acquisition device and the illumination light,the spliced image may be caused by ghosting,stitching,etc.,the weighted averaging method is used to make the stitching area smoothly transition and obtain a good stitching result.Then,choose the right material for the algorithm implementation and finally obtain a satisfied stitching result.Finally,the image of the metal wall with a girth weld which is similar to the surface inside the pressure vessel was used as the experimental object to verify the feasibility of SURF algorithm applied to the image stitching of the surface inside the pressure vessel.According to the system design,select the appropriate fisheye lens,CMOS camera and LED light source to build an experimental platform.The system software is implemented in the Visual Studio environment using C++ language,and the functions including image acquisition and processing,data comparison and point distance measurement are respectively completed.After the system software and hardware are connected,the experimental platform is placed inside the pressure vessel to collected,processed and analyzed images through the system operation interface.The function of image acquisition and processing achieves the expected effect and a good panoramic image of the surface inside the pressure vessel is obtained.But the functions of data comparison and point distance measurement still need to be improved.This system is the application research of machine vision in the field of defect detection of surface defects inside the pressure vessel.Fisheye camera is used to obtain the images inside the pressure vessel with appropriate lighting scheme and the fisheye image correction algorithm,image preprocessing algorithm,image stitching algorithm based on SURF are used to process the obtained images to achieve the panoramic images of the pressure vessel inner surface.System achieve the function of safty and efficient detection of surface defects inside the pressure vessel.