Surface shape recovery from grayscale images and its application in process control

Layered Manufacturing is a novel rapid prototyping technology which additively builds free-form parts by depositing material layer by layer. Such approach imposes higher requirements on process control, since it is impossible to fix the process error in the layer after next layer is deposited on top of it. Therefore, adaptive process control mechanism must be embedded into the fabrication process and must include a surface quality feedback upon deposition of each layer. Providing fast and reliable surface quality information based on optical Machine Vision has been the primary motivation for this work.; In this work, we approach this problem at two different levels. The first level provides information about the surface pattern by analyzing it using regular texture analysis. Process Signature serves as a texture element. The signature mass distribution is first defined. Inertia Tensor is then used to compute features and texture segmentation is performed. Regions with different textures identify surface areas with different process flow. Two mass distributions are considered and inertia tensor features for them are discussed. The second level carries out thorough analysis by recovering the shape of the cross-section of the surface, called profile, which is imaged by the Process Signature. Snake is used to define the deformable model of the profile. Snake elements act as control points for the profile and iteratively migrate towards minimizing the difference between the profile's image and the Process Signature. We develop several techniques based on snake methodology and carry out comparisons with the Tsai-Shah shape-from-shading technique. We propose the surface quality monitoring algorithm based on the combination of both first and second level techniques. We attest the applicability of the proposed quality monitoring scheme as a feedback tool in adaptive process control using Matlab simulation of Layered Manufacturing deposition. Proposed image segmentation and surface recovery techniques are generic and may be used in number of various application areas.