Context dependent automatic view planning: the inspection of mechanical components

Literature shows how 3D CAD product models can be used as reference in order to manage the verification process by 3D optical scanning systems and the computation of the optimal viewpoints. However, in the mechanical field, a variety of inspection tasks is experienced by engineers involved in the quality control process: GD&T verification, production phases control such as sheet metal cutting, evaluation of aesthetic appearance of parts, global shape deformation measurement and specific point deviations assessment. This leads to the necessity of flexible view planning approaches which adapt to the specificity of the required inspection task. The present work targets the development of a comprehensive view planning approach in which several algorithmic options are triggered by the product features to be inspected. Algorithms have been implemented in a prototypal software system which has been experimented as an off-line application to provide inputs to a multi-axis Degree of Freedom (DoF) robot arm mounting an optical 3D scanner. Two test cases from die casting and automotive fields are presented. They show the computation of acquisition poses in a suitable sequence and efficiency in the obtained results.