An Approach to Curve Reconstruction from Multiple Images
Zhang Weizhong, Zhang Chi, Pan Zhen-kuan, Wang Guanglei · 2012
In this paper, a method for recovering 3D curves of the free-from object from multiple images with partial occlusion is put forward. To ensure reliability of feature recognition and to heighten the precision of image processing, a set of marked object curves are applied. A method to extract the curves in the images by skeleton extracting and fitting is proposed. The optimization of curve matching between the image pairs is established. The objective function is optimized to reconstruct 3D curves with using the triangulation method and by means of Bundle adjustment. Experimental results demonstrated that the approach is practical and reliable for 3D modeling of free-form objects in the application of industry production.