Localizing components on printed circuit boards using 2D information

Information retrieval based re-configurable recycling is crucial for solving the environmental and social problems raised by the dramatically increasing amount of waste printed circuit boards. However, the indispensable localization of mounted components for the desired information retrieval is an unsolved problem for state-of-the-art techniques so far. In this paper, for the first time, a feasible solution addressing this problem by using general 2D information is proposed. In the form of a sophisticated modular analysis pipeline, complementary information sources are combined appropriately for capturing the sought objects. With a novel background removal algorithm and an additional candidate validation step, the proposed approach significantly outperforms state-of-the-art localization approaches in case of PCB images. Moreover, with respect to the intended trade-off between complexity and performance, it is quite straightforward to re-construct the deployed workflow accordingly by considering the knowledge obtained in a comprehensive evaluation of diverse combination possibilities.