A Systemic Point-cloud De-noising and Smoothing Method for 3D Shape Reuse

3D shape reuse, as an effective way to carry out innovative design, requires a digital model database where the entities are accurate and sufficient representations of objects in the real world. 3D scanning is a prevailing tool to quickly convert physical models into virtual ones. However, the scanned models without post-processing could not be used directly due to environment noise and accuracy limitation in terms of discrete sampling property in scanning. This paper introduces a systemic point-cloud de-noising and mesh smoothing method to handle this issue. The model de-noising and regularity is based on k-means clustering, and mesh smoothing module is an improved mean approach which processes the discrete data in the regular order. Case study will be given to verify the smoothing effectiveness. The proposed method could facilitate the construction of model database for design reuse, and could be output to downstream applications such as shape adaptive deformation, and shape searching.