Enhanced detection of diverse defects by developing lighting strategies using multiple light sources based on reinforcement learning

Traditional inspection systems with a single light source are not efficient at detecting a few particular defects with a single inspection. Unlike before, the multi-light source inspection environment allows us to extract more different defects in a piece of work, depending on what we are working on with varying sources of light. We proposed the formulation of the multi-lights source lighting strategy to improve the inspection capability of Automated Optical Inspection (AOI). The process of developing this study not only utilizes the ubiquitous image processing to extract defects but also imports the design of generalized defect sample and reinforcement learning, dealing with diverse defects under in-depth inspection by cascading both light and camera parameters. As a result, the AOI system emphasized that the inspection parameters can be intelligently adjusted to appropriate values based on various defects, maximizing the detection of diverse defects. From the perspective of intelligent AOI results, there are two outstanding outcomes for a multi-light source lighting strategy. One is an efficient learning process, which facilitates us to obtain the strategy needed in 40 to 50 min, depending on the reward function designed. The other is an advanced inspection function that can extract 37% more defects than conventional methods.