Virtual reality generation of synthetic images of SMDS for automatic optical inspection algorithms

In this paper we introduce a generation method of synthetic images designed to produce reference patterns in an Automatic Optical Inspection System of Surface Mounted Devices, SMD, on Printed Circuit Boards, PCB. The images are created numerically emulating a frontal view applying basic reflection laws to a Virtual Reality model of the SMD under known illumination sources. The synthetic images are processed and used as reference patterns in the classification method to determine the state of an inspected device, thus avoiding the use of reference samples to train the system. We describe the main features of the Visual System used in this work and we introduce the method we developed to create and characterize b/w images. Next, we introduce the method created to generate numerically 2D b/w images reproducing the distribution of light intensity produced by the coplanar illumination system on the surfaces of the SMD in a frontal view, and, we illustrate how the synthetic images can be used in classification methods describing a Bayes Classifier implemented to determine the state ¿present/absent¿ of an inspected SMD. Finally, we discuss about the possibilities of using synthetic images to innovate the classification methods on inspection systems.