Synthetic Image Generation Using the Finite Element Method and Blender Graphics Program for Modeling of Vision-Based Measurement Systems

Abstract

Computer vision is a frequently used approach in static and dynamic measurements of various mechanical structures. Sometimes, however, conducting a large number of experiments is time-consuming and may require significant financial and human resources. On the contrary, the authors propose a simulation approach for performing experiments to synthetically generate vision data. Synthetic images of mechanical structures subjected to loads are generated in the following way. The finite element method is adopted to compute deformations of the studied structure, and next, the Blender graphics program is used to render images presenting that structure. As a result of the proposed approach, it is possible to obtain synthetic images that reliably reflect static and dynamic experiments. This paper presents the results of the application of the proposed approach in the analysis of a complex-shaped structure for which experimental validation was carried out. In addition, the second example of the process of 3D reconstruction of the examined structure (in a multicamera system) is provided. The results for the structure with damage (cantilever beam) are also presented. The obtained results allow concluding that the proposed approach reliably imitates the images captured during real experiments. In addition, the method can become a tool supporting the vision system configuration process before conducting final experimental research.

Keywords

• image-based measurement
• vision sensor modeling
• vision system simulation
• image-based reconstruction
• finite element method
• physics-based computer graphics