Image analysis in surface inspection is very dependable on the image acquisition step. Fine control and optimization of the acquisition process, along with understanding the light dependent material behavior, will result in higher image analysis efficiency. Aim of the research is to develop methods for to be used in a standalone system, capable of inspecting a complete product regardless of its geometrical complexity and light response properties, based only on the provided zero-defect samples of the product. Such system would be made possible through interconnection of computer graphics, image processing, machine learning and robotics, where the first two may be considered the most important, as their combination is to produce a new field of research named virtual image processing.
Use of prior knowledge of products characteristic makes it possible to adapt the inspection to each new product, as every material has its inherent characteristics such as light dependent behavior and texture. By modeling the characteristics relevant for further inspection, system controls the overall conditions and allows less possibility of error due to change in characteristics consistency.