Video-based detection infrastructure is crucial for promoting connected and autonomous shipping (CAS) development, which provides critical on-site traffic data for maritime participants. Ship behavior analysis, one of the fundamental tasks for fulfilling smart video-based detection infrastructure, has become an active topic in the CAS community. Previous studies focused on ship behavior analysis by exploring spatial-temporal information from automatic identification system (AIS) data, and less attention was paid to maritime surveillance videos. To bridge the gap, we proposed an ensemble you only look once (YOLO) framework for ship behavior analysis. First, we employed the convolutional neural network in the YOLO model to extract multi-scaled ship features from the input ship images. Second, the proposed framework generated many bounding boxes (i.e., potential ship positions) based on the object confidence level. Third, we suppressed the background bounding box interferences, and determined ship detection results with intersection over union (IOU) criterion, and thus obtained ship positions in each ship image. Fourth, we analyzed spatial-temporal ship behavior in consecutive maritime images based on kinematic ship information. The experimental results have shown that ships are accurately detected (i.e., both of the average recall and precision rate were higher than 90%) and the historical ship behaviors are successfully recognized. The proposed framework can be adaptively deployed in the connected and autonomous vehicle detection system in the automated terminal for the purpose of exploring the coupled interactions between traffic flow variation and heterogeneous detection infrastructures, and thus enhance terminal traffic network capacity and safety.

Reasonable bus timetable can reduce the operating costs of bus company and improve the quality of bus services. A data-driven method is proposed to optimize bus timetable in this study. Firstly, a bi-objective optimization model is constructed considering minimize the total waiting time of passengers and the departure times of bus company. Then, Global Positioning System (GPS) trajectories of buses and passenger information collected from Smart Card are fused and applied to calculate the key parameters or variables in optimization model, including time-dependent travel time, bus dwell time and passenger volume. Finally, by adopting a specific coding scheme, an improved Non-dominated Sorting Genetic Algorithm-II (NSGA-II) is designed to quickly search Pareto optimal solutions. Furthermore, an experiment is conducted in Beijing city from one bus line to validate the effectiveness of the proposed method. Comparing with empirical scheduling method and traditional single-objective optimization base on GA, the results show that the proposed model could quickly provide high-quality and reasonable timetable schemes for the administrator in urban transit system.

Ship tracking provides crucial on-site microscopic kinematic traffic information which benefits maritime traffic flow analysis, ship safety enhancement, traffic control, etc., and thus has attracted considerable research attentions in the maritime surveillance community. Conventional ship tracking methods yield satisfied results by exploring distinct visual ship features in maritime images, which may fail when the target ship is partially or fully sheltered by obstacles (e.g., ships, waves, etc.) in maritime videos. To overcome the difficulty, we propose an augmented ship tracking framework via the kernelized correlation filter (KCF) and curve fitting algorithm. First, the KCF model is introduced to track ships in the consecutive maritime images and obtain raw ship trajectory dataset. Second, the data anomaly detection and rectification procedure are implemented to rectify the contaminated ship positions. For the purpose of performance evaluation, we implement the proposed framework and another three popular ship tracking models on the four typical ship occlusion videos. The experimental results show that our proposed framework successfully tracks ships in maritime video clips with high accuracy (i.e., the average root mean square error (RMSE), root mean square percentage error (RMSPE), mean absolute deviation (MAD) and mean absolute percentage error (MAPE) are less than 10), which significantly outperforms the other popular ship trackers.

The accurate forecasting of traffic states is an essential application of intelligent transportation system. Due to the periodic signal control at intersections, the traffic flow in an urban road network is often disturbed and expresses intermittent features. This study proposes a forecasting framework named the spatiotemporal gated graph attention network (STGGAT) model to achieve accurate predictions for network-scale traffic flows on urban roads. Based on license plate recognition (LPR) records, the average travel times and volume transition relationships are estimated to construct weighted directed graphs. The proposed STGGAT model integrates a gated recurrent unit layer, a graph attention network layer with edge features, a gated mechanism based on the bidirectional long short-term memory and a residual structure to extract the spatiotemporal dependencies of the approach- and lane-level traffic volumes. Validated on the LPR system in Changsha, China, STGGAT demonstrates superior accuracy and stability to those of the baselines and reveals its inductive learning and fault tolerance capabilities.

Most previous research has handled the task of ship type recognition by exploring hand-craft ship features, which may fail to distinguish ships with similar visual appearances. This situation motivates us to propose a novel deep learning based ship type recognition framework which we have named coarse-to-fine cascaded convolution neural network (CFCCNN). First, the proposed CFCCNN framework formats the input training ship images and data, and provides trainable input data for the hidden layers of the CFCCNN. Second, the coarse and fine steps are run in a nesting manner to explore discriminative features for different ship types. More specifically, the coarse step is trained in a similar manner to the traditional convolution neural network, while the fine step introduces regularisation mechanisms to extract more intrinsic ship features, and fine tunes parameter settings to obtain better recognition performance. Finally, we evaluate the performance of the CFCCNN model for recognising the most common types of merchant ship (oil tanker, container, LNG tanker, chemical carrier, general cargo, bulk carrier, etc.). The experimental results show that the proposed framework obtains better recognition performance than the conventional methods of ship type recognition.