Detection of Small Splits in Car-Body Manufacturing

E. de la Fuente, F.M. Trespaderne, and F. Gayubo (Spain)


Metallic Part Inspection, Markov Random Fields.


An on-line machine vision system for defect detection in large stamped car parts is presented. The acquisition system includes a CCD progressive camera and a diffuse illumination system. Camera and illumination have been mounted on a robot to be moved over the pieces proceeding from a sheet metal forming line. The acquired images are firstly restored using a Markov random field model in order to eliminate the spurious features without modifying the delicate characteristics of the small splits. Defect detection is carried out using an ultra rapid valley detection algorithm. The developed system detect the presence or absence of splits and through cracks greater than 0,2 mm in several designated zones of 10 x 10 cm on the totality of parts produced at a press line. The vision control will determine whether the part will continue in the production flow or be excluded from it. Nowadays the robustness of this machine vision system is being demonstrated on an industrial pilot line.

Important Links:

Go Back