A Predictive Approach to Geometry Preparation for AR/VR Applications

Abstract

AR/VR applications are a valuable tool in product development and the overall product lifecycle in engineering. However, data transformation of the models from CAD systems to the AR/VR applications is labor-intensive and requires expertise. The main task in the data transformation is the tessellation of the product geometry. Depending on the product complexity and the performance of the target platform extensive optimization is needed to ensure the usability and visual quality of the AR/VR application. Current approaches to this problem use iterative and inflexible processes mostly based on tessellation and on mesh decimation that ignore the varying importance of different geometric aspects for an AR/VR application. An alternative respectively more targeted approach is proposed, that aims at predicting tessellation results and moving the optimization process before the actual tessellation. As a result, the need for iterative operations on the polygon meshes can be reduced or ideally avoided altogether. The paper presents some results of an investigation of the hypothesis that geometric complexity metrics can be used to control and enhance the choice of tessellation parameters. Several characteristics and metrics are identified and gathered from literature and subsequently evaluated with regard to the polygon count and visual quality in the geometry preparation process. Based on the evaluation, prediction models are created and implemented in a geometry preparation tool. The performance is evaluated and discussed.