The Research And Application Of TEHCM Surface Defect Detecting Using Machine Vision

By more production capacity and higher quality demand, the operator eyes detect defect is not enough for company and customer. Machine vision have higher efficiency and scientific than operator's in defect detecting and was used in many plants. The application in our production line can increase the detecting efficiency and decrease the effect form operator's subjectivity.We first learned the machine vision element, researched its development and application at home and abroad. Mainly finished four aspects contents in design machine vision: Confirm how to work; select hardware model and debugging the machine; detecting software program and debugging; verify the machine performance.On how to work, according to the demand from capacity and quality, we analyzed vision station decide to use multi-camera system on the machine; considering the reliability and safety, we designed detecting position and preventing error operation function.On hardware construction design, we using electrical, EMC, machinery, production knowledge. By refer to machine vision in other production line, we designed right camera parameter, light source and optoelectronic sensor.On software design, we used In-sight to process the image after learning digital image processing element. The surface defect detecting including connector, thermistor, grommet, screws and filter area. By analysis of surface character, we select the right function to execute image matching, filter and so on, then get the position, area, distance of AOI(area of interesting). Final it can output “Pass” or “Fail”.To verify the machine can distinguish defect, we used a standard “Pass” and a “Fail” module for test, result shows the machine can get the parameter of AOI and output pass or fail result; to verify its reliability, we did Gage R&R evaluation using 2 operators and 20 pcs modules from production line, result shows machine's reproductively and repeatability can reach to 100%, so machine have enough capability to distinguish module surface is good or not. Using 20 pcs module to verify the detecting time, the average detecting time is 5.82 second that can meet the 6 second demand. All test demonstrated the vision machine can meet the desired target, so it can be used in line. Benefit evaluation shows the vision machine used in production line can save RMB 708,000 every year.