Binocular vision measurement for large-scale weakly textured ship hull plates using feature points encoding method

Abstract

The accurate forming measurement for curved hull plates is vital to ensure the quality of shipbuilding. However, the weak texture, large scale and complex shapes of curved hull plates present challenges for shape inspections. To enrich the texture of plates, checkerboard patterns are combined with the binocular vision system. Moreover, calculating the sub-pixel coordinates of corner points enhances reconstruction accuracy. A novel encoding method for the feature points on the complex shaped plates with repeatable texture is proposed to match the homologous points correctly. Finally, three dimensional coordinates are obtained by triangulation. Experimental results show that the dimensional accuracy is 300 ± 0.636 mm and the average accuracy is ±0.219 mm by registration. Inspection processes are completed within 1 min. This method is more automatic and convenient than contact measurement. Moreover, the efficiency is improved by about 2 times than multi-vision measuring method, enhancing the accuracy and efficiency of measurement for curved hull plates.