Abstract: In the research, image processing and machine learning techniques are employed on the surface defect detection
of metal injection molded parts. There are two sections in this manuscript; one is image capturing system, and another
one is the machine vision inspection system. For image capturing system, the ring light with diffuser is used, and it could
project uniform light on the metal surface. CCD cameras are used to capture the images of metal injection molded parts.
To obtain the best quality image, the inspection system uses Sobel operator searching for the focused position where
the sharpness measure is maximized.
For the machine vision inspection system, local feature information of the image is extracted by Histogram of oriented
gradient, Gary level co-occurrence matrix and Local binary patterns. The selected features are imported into the support
vector machines classification image to classify the crack defects.
