Steel bars counting and splitting method based on machine vision

This paper proposes a novel on-line steel bars counting and splitting method based on machine vision which uses concave dots matching to segment, K-level fault tolerance to count and visual feedback to multiple split automatically. Firstly, it preprocesses images of bars and uses connected area analysis to obtain edge profile of adherent bars, then scans concave areas in the contour and find concave dots. Secondly, it uses concave dot matching condition to segment and counts single bar after segmentation to achieve counting purpose through movement estimation and K-level fault tolerance algorithm. Finally, visual feedback is presented, if preliminary split is wrong, redraw the line for splitting and drive the splitting mechanism again. Experiment results show that the method has a high accuracy for segmentation of adherent bars, and can split steel bars accurately. The accuracy ratio of segmentation for steel bars whose diameters are between 8mm and 20mm is more than 99.90%, which satisfies the accepted standard of enterprises.