Abstract
In order to automate the human visual inspection process in the diode production factories, we construct an experimental computer system with an image digitizing circuit and a microscope. Based on the statistical decision theory, we formulate six mathematical functions for the six major defects frequently seen on diode pellets.
On one thousand good samples we calculate various statistical parameters, with which another 800 samples are tested. The correct answer ratio is 94 percent, if the human decisions are assumed correct. Detailed investigation indicates that about one-half the split decisions comes from the human errors, whereas the other half comes from computer error. The final hit ratio increases to about 97 percent. We conclude with a proposal for a practical inspection system.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
References
Daele JV, Oosterlinck A, Van den Berghe H (1980) Automataic visual inspection of reed switches. Optical Engineering 19(2):240–244
Fletcher E (1979) MIDAS-high speed automatic surface inspection of steel billets, IPC Business Press 12(4): 163–166
Jarvis JF (1980) A method for automating the visual inspection of printer wiring boards. IEEE Transactions on PAMI 2(1):77–82
Okawa Y (1984) Automatic inspection of the surface defects of cast metals. Computer Vision, Graphics and Image Processing 25:82–112
Poter, GB, Mundy JL (1980) Visual inspection systems design. IEEE Computer (May):40–48
Vanderbrug GJ, Nagel RN (1974) Image pattern recognition in industrial inspection. NBSIR 79-1764
Author information
Authors and Affiliations
Rights and permissions
About this article
Cite this article
Okawa, Y., Mizuno, S. Automatic inspection of diode pellets. Machine Vis. Apps. 4, 131–133 (1991). https://doi.org/10.1007/BF01230196
Issue Date:
DOI: https://doi.org/10.1007/BF01230196