A Transductive and Inductive GNNs for Physical Moving Objects Detection in Surface Scenes for Digital Twins

Computer vision applications using static or moving cameras are often required in digital twins generation. More specifically, the detection of moving objects is essential to provide a virtual representation of an environment in order to reflect physical moving objects accurately. To this end, background subtraction (BGS) is then applied to separate the background (BG) and the foreground (FG) from videos. Numerous publications employ mathematical, machine learning, and signal processing models to be more robust to the open challenges presented in videos. Recently, many methods using graph neural networks for BGS have been reported, with very promising outcomes. This chapter provides a survey of transductive and inductive Graph Neural Networks (GNNs) for moving objects detection (MOD) comparing their architectures. After analysis of their strategies and limitations, a comparative evaluation of the large-scale CDnet2014 dataset is provided. Finally, we conclude with some potential future research directions.