Research On The Automatic Magnetic Particle Detection System Of Tank Cracks By Image Processing Method

At present,the periodic inspection of large-scale pressure-bearing special equipment such as storage tanks adopts manual inspection,and the detection of magnetic particle cracks has many problems,such as heavy workload,high labor intensity,bad environment,etc.It is urgent to develop intelligent testing equipment which can replace manual operation,including the detection part and the walking part of the mechanism.This paper studies the visual inspection system for tank cracks on automatic inspection equipment,and proposes a visual automatic identification system for tank cracks based on magnetic particle detection.The system uses machine vision technology to detect tank cracks,and replaces the human eye recognition function,including image acquisition,image processing,image crack length calibration,crack feature extraction and crack length.calculation of degree and so on.According to the characteristics of Tank Crack Image and detection accuracy,the type and parameters of system hardware are selected,including camera,lens,light source,lighting mode,magnetic powder and magnetization type.MATLAB software is selected as the image processing software of the system.Aiming at the image noise such as grinding and stain in Tank Crack image,adaptive median filter is used to remove the noise interference.In view of the low contrast between crack information and background information,the crack image is enhanced by the combination of open operation and linear transformation,which improves the contrast between crack information and background information.The image region containing crack information is processed by image block processing method.The circularity of crack,the ratio of length to width of crack and the average width of crack are introduced as the conditions of crack identification to distinguish the crack and interference information,so as to eliminate the interference information such as stains and pits.According to the characteristics of Tank Crack shape,two methods of calculating crack pixel length,linear fitting method and cumulative calculation method are proposed.The results of calculation are compared,and the cumulative calculation method which is closer to the actual length is selected.The proportional relationship between the crack pixel length and the actual crack length is defined by camera calibration.On this basis,the actual crack length of the tank is calculated.Using VB6.0 software,the automatic detection system of magnetic powder for Tank Crack Based on image processing is developed.The acquisition,processing,storage and detection of weld image are realized.Through field experiments,the actual measured crack length is 9.5 mm,the accumulative measured length is 8.9 mm,and the error is 6.4% within the allowable range.The feasibility and accuracy of the detection system are verified,which can provide reliable technical support for the development of tank crack automatic detection.Figure 49;Table 5;Reference 50.