Skeletonization using fuzzy distance transform for diffuse reflection structured light

Many indoor navigation robots use structured light vision measurement systems (SLVMS) for mapping, localization, and obstacle avoidances. In these systems, the extraction of light stripe skeleton directly affects the accuracy of the final measurements. One of the problems in SLVMS is that when the light is projected to some rough surface it will cause strong diffuse reflection. Thus the light stripe captured by the camera contains more or less fuzzy parts. The uncertainty of diffusion pattern and distribution seems to be a great challenge for conventional skeletionization extraction methods. In this paper we proposed a skeletonization algorithm based on fuzzy distance transform (FDT)for the tricky fuzzy light stripe, which, to some extent, overcomes the disadvantage of existing methods and leads to better skeleton representation. First we introduce the principal methods for skeletonization and the theory of FDT. Then, the image preprocessing method together with the extended FDT algorithm are elaborated. Finally, some example results are shown to demonstrate the effectiveness of our algorithm, followed by the comparison with three other skeletonization methods which are based on distance transform(DT) theory, Voronoi diagram and the principle of digital morphological erosion respectively.