Intensity-based registration with voxel-based computer-aided manufacturing for adaptive machining

Abstract

This paper presents a novel application of intensity-based volume registration to manufacturing using voxel-based computer-aided manufacturing (CAM) models. The introduced techniques are presented in the context of machining irregularly shaped materials by integrating volumetric imaging feedback to computer numerical control (CNC) machine tools. This requires a comparison and alignment to be performed in the software to geometrically "fit" the source design model inside a rendered starting material. The requirements differ from many typical registration applications in that the workpiece will necessarily be larger (i.e., greater in volume) than the desired final computer-aided design (CAD) file. Therefore, models need to be aligned for toolpath generation to workpiece counterparts that have been either volumetrically offset or contain additional material/volume. Intensity-based registrations are unique in that they consider only the voxel values over the entire volume. Although advancements in medical imaging have produced efficient, robust voxel registration algorithms, these techniques have not yet been applied to manufacturing. This research introduces the use of maximization of mutual information (MMI) for voxel-based CAM to drive an alignment registration for systems integrating imaging technology. A simple but novel method, which the authors have named minimization of distance variance (MDV), is also introduced. This minimizes the variance between voxel intensities to demonstrate the design of a similarity metric for a simple case in machining rough castings.