Texture Synthesis for Automated Visual Surface Inspection Planning

Automated surface inspection planning is a research area, aimed to enhance inspection in automated and custom product manufacturing. Visual surface inspection focuses on two main tasks. First, camera and illumination placement for achieving the required coverage of the inspected product. Second, the development of defect detection algorithms. Both tasks greatly benefit from realistic and automated image synthesis of the inspected object using computer graphics modeling and rendering methods. The realism of synthesized images greatly depends on object material, whose properties are greatly influenced by texture. In our work, we focus on texture synthesis and its application for visual surface inspection planning. Our research was motivated by the complexity and vastness of textures that have to be recreated as well as their consistent application on complex, free-form geometry for the purposes of automated texture synthesis. In this paper, a framework that stemmed from our research is discussed. Furthermore, we provide an overview and shortly discuss several texture synthesis models that we have developed. Presented methods enable non-computer graphics experts, such as inspection planning experts, to recreate a range of patterns often present in the machined surfaces, using only several parameters, during the synthesis of representative product image data.