Remote Operation Centres (ROCs) are safety-critical shore infrastructures enabling Maritime Autonomous Surface Ships (MASS) through remote supervision, decision-making, and human-machine cooperation. However, fragmented ROC technologies hinder a system-level understanding of their interactions, constraining reliability and safety assurance for fully autonomous maritime operations. A Large Language Model (LLM)-based patent analytics framework is developed to systematically identify and map ROC-related technologies from patent data. By integrating LLM-assisted topic modelling, complex network analysis, and temporal trend evaluation, the framework provides a data-driven perspective on both the evolutionary dynamics and structural interrelations among ROC technologies. In addition, a human-machine collaboration evaluation framework is constructed to evaluate the levels of human involvement and technological involvement for each ROC technology. This can help improve the understanding of HMC in future ROCs and provide insights into future remote operators and regulatory development.

Shipping route information provides important spatial references for understanding ship behaviour and maritime traffic organisation. Accurate route extraction can support route and speed optimisation to reduce fuel consumption and emissions, while also providing finer-grained spatial information for traffic monitoring and the identification of potentially conflict-prone areas. In structured waterways, such as Traffic Separation Schemes (TSSs) and navigational channels, ship trajectories share a common passage while following different approach and departure paths due to their different origins and destinations. This study proposes a two-stage route extraction method based on ordered turning point sequences: Change Point Detection is first used to identify turning points from course variations, and hierarchical clustering is then performed in two stages, first according to spatial similarity and then according to course differences. A case study in Øresund shows that the method can distinguish trajectories with similar turning point locations but different approach and departure patterns, providing more detailed route representations for traffic analysis and maritime traffic supervision.

Maritime traffic in narrow channels is characterised by limited navigable space, dense vessel movements and rapidly evolving multi-vessel interactions, making real-time risk identification particularly challenging. This study proposes a data-driven, dynamic multi-scale framework for assessing maritime traffic complexity using Automatic Identification System (AIS) data. Historical vessel trajectories are first used to construct category-specific empirical ship domains and identify near-miss events, allowing historical conflict patterns to be incorporated as spatial risk priors. Three physically interpretable indicators are then developed to capture traffic complexity at different interaction scales: the Conflict Factor (CF) identifies abnormal course deviations at the individual-vessel level, the Approach Factor (AF) measures dynamic interaction risk between vessel pairs, and the Group Clustering Factor (GCF) evaluates aggregation and directional inconsistency within vessel groups. The framework is validated using AIS data from the Nanjing section of the Yangtze River and a collision case in the Malacca Strait. Results show that the proposed method can better distinguish hazardous vessel interactions from normal high-density traffic and demonstrates strong sensitivity to interaction-driven risks. With an average processing time of 0.28 seconds, the framework also shows potential for real-time Vessel Traffic Service supervision and proactive maritime risk identification.