An adaptable vision system for the automatic inspection of surface defects in automotive headlamp lenses

As a great part of the industrial inspection problems require specific, expensive, and highly complex solutions, there is the need to build flexible systems capable to be easily adapted to different models of a part. In this context, this paper proposes the use of a component-based methodology for designing and building flexible inspection systems for automotive headlamp lenses. Through computer vision, and taking into account the different elements that can be utilized in an industrial environment, we propose the use of a component based architecture for easily designing and building industrial inspection systems for this kind of parts. Within this architecture, each software, or hardware element, is encapsulated in a component with a predefined input/ouput layout and, later, it can be used considering the task inspection requirements. The main advantages of the proposed working philosophy are its flexibility, that the resulting machine vision can be easily adapted to any kind of headlamp lens, and that it is fully open to any new component (lighting system, process algorithm, industrial conveyor, etc.). This proposition has been successfully validated building an industrial prototype that was tested in both, the lab, and in a production line.