Machine Vision Inspection Of Surface Defect Of Brake Master Cylinder Pistons

Becouse of the advantages of space saving, large travel and long life, theplunger type brake master cylinder piston is the most widely used in thecurrent automobile industry. The piston is of high surface quality requirement,while the defects are of tiny size without rules, and the surface is stronglyreflective. Defects detection is of high difficulty, high cost and low efficiency.And the detection can only rely on manual working for so long. The articlefocuses on the key technical issues of the surface defect vision inspection.The machine vision detection system of surface defect of brake mastercylinder pistons is designed and optimized. Tridimensional combinationsystem of multiple light sources is designed, which is suitable for thedetection of defects of different directions. Soft light is designed to reduce thenoise in the detection image. Mathematical model of light intensity loss in theprocess of light emission-reflection-imaging is set up. Polarization filteringmethod is proposed and used on the vertical light source and the camera lens.By choosing the polarizing direction of polarizing film, the brightness of areawithout defect and the width of reflective band can be reduced, and the imagecontrast can be enhanced.Several key technical problems are deeply studied. Mathematical modelof Camera-light source-work piece is established, which helps to analyzethe influence of light source posture and position on imaging quality. Imagecalibration and the measurement of geometric parameters of the piston are completed. Image preprocessing includes Image Mask, Color Operators andColor Plane Extraction. Image processing includes Image segmentation,Advanced Morphology, Reject Border and Particle Filter. BackgroundCorrection algorithm of local threshold is used to do image segmentation.Combining algorithm principle and histogram contrast experiment study, therule to choose different local threshold window sizes is discussed. Geometrytemplates matching technology and area measuring technology are used todesign two different defect detection modes. The former is of strongerstability, while the latter is of faster running velocity.Friendly software system is designed. And a series of experiments arecarried out. Features of the scratch defects, bump defects, bruise defects,wood grain defects, spiral cut defects and sand wounded defects are detailed.Random sampling experiment with300samples shows that the visioninspection has no leak detection. The false positive rate and stabilization biasrate are both not more than2%. Each inspection costs2.5seconds, whichgreatly improves efficiency and reduces cost. Combining with practicalexperience, system error is analysed, which provides the feasible measuresfor system improving.