A Review of Scene Understanding in Smart Manufacturing Environments

Scene understanding is the process of analysing and interpreting various perceptual information in the environment to understand and reason about people, objects, events, and situations in the environment. The basic concepts and research contents of scene understanding are outlined, and deep learning-based scene understanding methods are classified and summarised, including data-driven models, end-to-end models, cross-modal models and other methods. The application of scene understanding in the field of smart factories and its practical application are introduced. The challenges and future directions of scene understanding are summarised, including how to improve the efficiency, accuracy and robustness of scene understanding, and how to apply scene understanding to more realistic scenes.