Development of auto defect classification system on porosity powder metallurgy products

Automatic optical inspection (AOI) has been applied to many manufacturing fields for defect inspection of mass production parts, such as PCB and TFT-LCD, but it has never been applied to the production line of porous powder metallurgy. By its nature, the powder-formed part has inherent non-uniform porosity pattern on the metal substrate. The defect’s images are not easily separated from the substrate surface using the conventional binarization technique. This study develops a new image processing methodology and employs optical system design to build up an on-line surface defect inspection system for powder metallurgy parts. An analysis algorithm is also developed for the auto defect classification technique. It removes the noise signals from the porous image, detects the object edge and uses the hybrid-based method to sort out defects on the surface, such as crevice, scratch, broken corner and dent. Experimental tests show the maximum miss rate can be controlled to less than 5.65%.