Online Rebar Recognition Based on Multi-View Images

Accurate rebar recognition is a key step of automatic online rebar counting based on image processing. Current popular rebar recognition methods based on front view image usually lead to missing detection for some special rebar such as obscured rebar, oxidation dark rebar and bent rebar. So a novel rebar recognition method based on multi-view images was proposed in this paper. Firstly, rebar images were captured simultaneously from front view and top view using two cameras. Then, the centers of rebar were located respectively in top and front view image. Finally, in the image fusion step, the top recognition result was used to compensate and validate the front recognition result for reducing and eliminating above missing detection. Experiments showed that this method can efficiently recognize above missed rebar and improve rebar recognition rate.