A Research of Scraped Surface Inspection with A Machine Vision System Incorporating a CMOS Camera and a Lampshade

• Main Authors: Chun-Yen Huang, 黃峻彥
• Other Authors: Cheng-Ho Chen
• Format: Others
• Language: zh-TW
• Published: 2013
• Online Access: http://ndltd.ncl.edu.tw/handle/61894430032284654890

碩士 === 國立勤益科技大學 === 機械工程系 === 101 === Scraping is a fundamental and key technique in the machine manufacture and assembly. It is an important process for producing high-accuracy machines. Scraping is a surface work using a scraper to minutely remove the material from the surface to rectify the inaccuracy during manufacturing and assembling. When judging the bearing mask, human visual inspection and experience are not dependable. They are effected by various factors (ex. Light, view angle, subjective experience, etc.). The research in collocates a CMOS camera and a lamp shade for the inspection of the scraped surface, A CMOS camera is used to extract the surface characteristics after scraping. The background and the characteristics are separated by image processing methods, The characteristics are enhance by the morphology technique. The PPI and POP for the bearing masks are then calculated and analyzed. A line scan is used to verify the accuracy of the CMOS and lamp shade system. First, the workpiece was examined by the line scan system 34 times to get the average POP of the whole surface to be 9.59%. And then, we divided the surface of the workpiece into 9, 12, and 15 blocks and used the CMOS and lamp shade system to gather the images and analyze the data. The mean absolute percentage error (MAPE) is used to compare the results. It showed that 12 blocks provided the minimum error 5.41% an indication for high accuracy, It was used as the optimal number for sub-grids. Thereafter, we used the CMOS and lamp shade and the line scan system to inspect the same positions of the 12 blocks on the scraped workpiece. The MAPE analysis showed on 7.46% error which is an indication for high accuracy. Finally, we sampled 4, 6, 8,10 and12 positions of the 12 blocks respectively to find the optimal sampling number for CMOS inspection. A LabVIEW random number program was run to provide 30 sets of random numbers for 4,6,8,10 and 12 samples respectively. The statistical analysis showed that sampling 8 positions provided the best results, and the average error is 4.86% which is an indication for high precision. This study showed that the CMOS and lamp shade system is highly accurate, and it is able to provide lowe cost, time and labor saving, highly-mobile, consistent and stable inspection system.