摘要

There is a small size, lack of or less texture, transparent and reflective phenomenon on the images of convex objects. So the traditional reconstruction methods such as stereo-vision matching and active vision scanning such as laser or structure light cannot be used for such objects. In this paper, a reconstruction method based on multi-contours for miniature convex object is proposed. Firstly, multi-view silhouette images are captured, then the accurate contours are extracted;secondly coarse mesh is generated by space carving of multi-contours, and thirdly, the fragmented polygons are merged according to angle and area constriants;at last, the accurate mesh is generated by space carving of multi-polygons of the coarse mesh. The reconstruction experiments of regular and convex objects (syringe needle, diameter is about 3 mm) proved that the angular error is less than 0.7°and the time cost is less than 15 seconds and the total process is without manual intervention. This method can solve the problem of reconstruction and vision measurement of convex objects effectively.

  • 出版日期2015

全文