Research Of Detection System Of Machine Vision Steel Ball Detector

Steel ball is a critical component of bearing, whose quality has a crucial impacton the performance and life of the bearings. With the development of China ’s steelball industry, whether a steel ball testing machine with stable performance, highdetection efficiency can be invented is the key factors for steel ball industry tocontinue to develop. In this paper, basic research is done for ball revealing method,lighting solution, image processing method, and software interface design of the steelball surface defect detector based on computer vision technology, and aiming at microsteel balls with diameter of1.5mm, according to the methods proposed, a prototype ofsteel ball surface defect detecting and sorting machine is built.Ball surface revealing method in traditional steel ball detectors are usually theball rolling by meridian line or spiral line. In this paper, a surface revealing method ofone-dimensional linear motion of the steel ball with dual-CCD camera is adopted.Two CCD cameras are placed above the rail by a certain angle, so that the wholesurface of a ball can be seen while it rolls along the rail. This method circumvents thecomplex2-dimensional movement of the ball, reducing the restrictions of ball sizethat can be detected, therefore equally applicable to steel ball with diameter less that3mm, which can hardly be detected by traditional micro Ball surface revealingmethod. According to the method above, a special dome light source with dualobservation holes is designed.Aiming at several common defects of the micro steel ball，cracking, rust,scratches, and flat, image processing algorithm is designed. Using the real-timeacquired ball image frame and the template frame get the frame difference, thusextracting the ball surface defects. And program by Visual C++.Designed software system and operator interface of the ball surface defectdetecting and sorting system by MFC, and calling the library functions to control thecamera to capture images, using a data collecting card to output switch control value,realized continuous detection (monitoring mode), single-step detection (debuggingmode) function.Tested by experiment, the performance of the image processing algorithms ofsteel ball surface defect detection is stable, which reaches the detection accuracyrequirements.