Abstract

Connected Vision Systems (CVS) are transforming a wide range of applications, including autonomous vehicles, smart cities, surveillance, and human–robot interaction. These systems leverage multi-view multi-camera (MVMC) data to enhance situational awareness by integrating MVMC tracking, re-identification (Re-ID), and action understanding (AU). However, deploying CVS in dynamic real-world environments remains challenging due to issues such as occlusions, diverse viewpoints, and environmental variability. Existing surveys primarily focus on individual tasks, including tracking, Re-ID, or AU, while overlooking their integration into a unified framework. Moreover, most previous reviews emphasize single-view settings, failing to address the challenges and opportunities of multi-camera collaboration and multi-view data analysis. To the best of our knowledge, this survey is the first to provide a comprehensive and integrated review of MVMC by unifying tracking, Re-ID, and AU within a single framework. We propose a novel taxonomy that categorizes CVS into four key components: MVMC tracking, Re-ID, AU, and integrated methods. Furthermore, we systematically summarize state-of-the-art datasets, methodologies, performance comparisons, and evaluation metrics, offering a structured overview of the field’s evolution. We also discuss open research challenges and emerging directions, including lifelong learning, privacy preservation, and federated learning, that are essential for advancing future CVS technologies. Finally, we outline promising research directions to improve the robustness, efficiency, and adaptability of CVS in complex real-world scenarios. We hope this survey will serve as a valuable resource for researchers and inspire the next generation of intelligent and adaptive connected vision systems.

Additional Comments