Search

Fast and accurate ship target detection technology plays an important role in improving driving safety, rescue at sea, marine environmental protection, and sea traffic control. It is also one of the key technologies for the development of ship informatization and intelligence. However, the current ship target detection models used at different scales in multiple scenarios exhibit high complexity and slow inference speed. The trade-off between model detection speed and accuracy limits the deployment of ship target detection models on edge devices. This study proposes a lightweight multi-scale ship target detection model based on the Yolov5s model. In the proposed model, the lightweight EfficientnetV2 and C3Ghost networks are integrated into the backbone and neck networks of the Yolov5s model to compress the computational and parametric quantities of the model and improve the detection speed. The Shuffle Attention mechanism is embedded in the neck network component of the model to enhance the representation of important feature information, suppress irrelevant feature information, and improve the model's detection performance. The improved method is trained and verified on the dataset collected and labeled by the authors. Compared with the baseline model, the inference speed of the proposed model increased by 29.58%, mAP0.5 improved by 0.1%, and the parameters and floating-point operations decreased by 42.82% and 68.35%, respectively. The file size of the model is 8.02MB, which is 41.46% lower than the baseline model. Compared with other lightweight models, the method proposed in this study is more favored in edge computing. [ABSTRACT FROM AUTHOR]

This paper proposes a reinforcement learning (RL)-based traffic control strategy integrated with attention mechanism for large-scale adaptive traffic signal control (ATSC) system. The proposed attention RL integrates attention mechanism into a multiagent RL model, namely multiagent proximal policy optimization (MAPPO), so as to enable more effective, scalable, and stable learning in complex ATSC envi)ronments. In the attention RL, decentralized policies are trained using a centrally computed critic that shares an attention model, while the attention model selects relevant intersections for each agent to estimate the global critic. This framework effectively reduces the computa)tional complexity and stabilizes the training process, enhancing the ability of RL agents to control large-scale traffic networks. The proposed control strategy is tested in both a large synthetic traffic grid and a large real-world traffic network of Yangzhou city using the microscopic traffic simulation tool, SUMO. Experimental results demonstrate that the proposed approach learns stable and sustainable policies that achieve lower congestion level and faster recovery, which outperforms other state-of-art RL-based approaches, as well as a gap-based actuated controller. [ABSTRACT FROM AUTHOR]

Against the backdrop of ongoing urbanization, issues such as traffic congestion and accidents are assuming heightened prominence, necessitating urgent and practical interventions to enhance the efficiency and safety of transportation systems. A paramount challenge lies in realizing real-time vehicle monitoring, flow management, and traffic safety control within the transportation infrastructure to mitigate congestion, optimize road utilization, and curb traffic accidents. In response to this challenge, the present study leverages advanced computer vision technology for vehicle detection and tracking, employing deep learning algorithms. The resultant recognition outcomes provide the traffic management domain with actionable insights for optimizing traffic flow management and signal light control through real-time data analysis. The study demonstrates the applicability of the SE-Lightweight YOLO algorithm, as presented herein, showcasing a noteworthy 95.7% accuracy in vehicle recognition. As a prospective trajectory, this research stands poised to serve as a pivotal reference for urban traffic management, laying the groundwork for a more efficient, secure, and streamlined transportation system in the future. To solve the existing vehicle detection problems in vehicle type recognition, recognition and detection accuracy need to be improved, alongside resolving the issues of slow detection speed, and others. In this paper, we made innovative changes based on the YOLOv7 framework: we added the SE attention transfer mechanism in the backbone module, and the model achieved better results, with a 1.2% improvement compared with the original YOLOv7. Meanwhile, we replaced the SPPCSPC module with the SPPFCSPC module, which enhanced the trait extraction of the model. After that, we applied the SE-Lightweight YOLO to the field of traffic monitoring. This can assist transportation-related personnel in traffic monitoring and aid in creating big data on transportation. Therefore, this research has a good application prospect. [ABSTRACT FROM AUTHOR]

Accurate crowd flow prediction is essential for traffic guidance and traffic control. However, the high nonlinearity, temporal complexity, and spatial complexity that crowd flow data have makes this problem challenging. This research proposes a dynamic graph convolutional network model (Res-DGCN) based on the residual network structure for crowd inflow and outflow prediction in urban areas. Firstly, as the attention layer, the spatio-temporal attention module (SA) is employed for capturing the spatial relationship between the target node and the multi-order adjacent nodes by processing the features of the human flow data. Secondly, a conditional convolution module (SCondConv) is used to enhance the model's capacity for learning about the shifting characteristics of crowd flow to obtain spatial dependence. Finally, we train the model with the Huber loss function to lower the model's sensitivity to outliers and achieve optimal convergence. In two public datasets, the mean absolute error (MAE) of the proposed model is improved by 5.2% and 9.4%, respectively, compared to the baseline models, and the root mean square error (RMSE) is improved by 4.8% and 8.8%, confirming the model's usefulness for crowd flow prediction tasks. [ABSTRACT FROM AUTHOR]

Traffic flow prediction is one of the most important tasks of the Intelligent Transportation Systems (ITSs) for traffic management, and it is also a challenging task affected by many complex factors, such as weather and time. Many cities adopt efficient traffic prediction methods to control traffic congestion. However, most of the existing methods of traffic prediction focus on urban road scenarios, neglecting the complexity of multivariate auxiliary information in highways. Moreover, these methods have difficulty explaining the prediction results based only on the historical traffic flow sequence. To tackle these problems, we propose a novel traffic prediction model, namely Multi-variate and Multi-horizon prediction based on Long Short-Term Memory (MMLSTM). MMLSTM can effectively incorporate auxiliary information, such as weather and time, based on a strategy of multi-horizon time spans to improve the prediction performance. Specifically, we first exploit a multi-horizon bidirectional LSTM model for fusing the multivariate auxiliary information in different time spans. Then, we combine an attention mechanism and multi-layer perceptron to conduct the traffic prediction. Furthermore, we can use the information of multivariate (weather and time) to provide interpretability to manage the model. Comprehensive experiments are conducted on Hangst and Metr-la datasets, and MMLSTM achieves better performance than baselines on traffic prediction tasks. [ABSTRACT FROM AUTHOR]

DRL has gained wild applications in the field of urban transportation signal control. However, the existing DRL traffic signal control researches only use the traditional deep neural network, and its perception ability is limited in complex traffic scenarios. In addition, as one of the three elements of reinforcement learning, it also needs to design the traffic state carefully and manually in the existing researches. Therefore, this paper proposed a DRL traffic signal control algorithm based on attention mechanism. By introducing the attention mechanism, the neural network could automatically pay attention to the important state components to enhance the perception ability of the network, improve the signal control effect, and reduce the difficulty of state vector design. Experimental results on SUMO platform show that compared with the three classical signal control algorithms, only using a simple traffic state, the proposed algorithm has the best performance in average waiting time and travel time under the condition of low and high traffic flow at single intersections and multiple intersections. [ABSTRACT FROM AUTHOR]

• An algorithm based on heterogeneous GNNs is proposed for feature fusion. • A mutual-information loss function is proposed for training. • Based on above methods, an IHA-MDGI algorithm is proposed for traffic signal control. • Experiments show the IHA-MDGI algorithm outperforms the state-of-the-art algorithms. Adaptive traffic signal control (ATSC) facilitates alleviating traffic congestion. Multi-agent deep reinforcement learning (MDRL) is a new promising algorithm for ATSC, and Graph Neural Networks (GNNs) further promote its learning ability. However, there are some drawbacks in the state-of-the-art MDRL algorithms. (1) These algorithms cannot effectively fuse diverse heterogeneous information of traffic networks due to adopting homogeneous GNNs; (2) These algorithms cannot be effectively trained due to merely adopting MDRL loss functions. In this paper, we propose an Inductive Heterogeneous graph Attention-based Multi-agent Deep Graph Infomax (IHA-MDGI) algorithm for ATSC. The proposed IHA-MDGI algorithm conducts both feature fusion via a proposed Inductive Heterogeneous graph Attention (IHA) algorithm and training via a proposed Multi-agent Deep Graph Infomax (MDGI) framework. Specifically, (1) Unlike the algorithms which adopt homogeneous GNNs, in the IHA algorithm, heterogeneous GNNs are designed to fuse both heterogeneous structural information and heterogeneous features of traffic networks, which aims to acquire heterogeneous information embeddings of traffic networks. (2) In the MDGI framework, the acquired embeddings are used to calculate the signal-control policies and Q -value for each agent, and then a mutual-information loss function is designed, which combines with the MDRL loss function to jointly train the whole algorithm. The designed mutual-information loss function focuses on maximizing mutual information between input (i.e., heterogeneous information embeddings) and output (i.e., Q -value), which can produce cooperative signal-control policies and maximize Q -value. We conduct the experiments in both real-traffic and synthetic-traffic networks under the time-varying traffic flows, and the results demonstrate that IHA-MDGI algorithm outperforms the state-of-the-art MDRL algorithms about multiple metrics. [ABSTRACT FROM AUTHOR]

Accurate traffic flow prediction plays a significant role in urban traffic management, including traffic congestion control and public travel route planning. Recently, several approaches have been put forward to learn the patterns from historical traffic data. However, there exist some limitations resulting from the use of the static learning method to explore the dynamical characteristics of the road network. Besides, the dynamic global temporal and spatial properties are not considered in these models. These drawbacks lead to a low prediction performance and make applying to a more extensive road network challenging. To address these issues, from the inspiration of Chebyshev polynomial, we proposed an improved dynamic Chebyshev graph convolution neural network model called iDCGCN. In the proposed approach, a novel updating method for the Laplacian matrix, which approximately constructs features from different period data, is proposed based on the attention mechanism. In addition, a novel feature construction method is proposed to integrate long-short temporal and local-global spatial features for complex traffic flow representation. Experimental results have shown that iDCGCN outperforms the state-of-the-art GCN-based methods on four real-world highway traffic datasets. [ABSTRACT FROM AUTHOR]

Traffic sign detection is a research hotspot in advanced assisted driving systems, given the complex background, light transformation, and scale changes of traffic sign targets, as well as the problems of slow result acquisition and low accuracy of existing detection methods. To solve the above problems, this paper proposes a traffic sign detection method based on a lightweight multiscale feature fusion network. Since a lightweight network model is simple and has fewer parameters, it can greatly improve the detection speed of a target. To learn more target features and improve the generalization ability of the model, a multiscale feature fusion method can be used to improve recognition accuracy during training. Firstly, MobileNetV3 was selected as the backbone network, a new spatial attention mechanism was introduced, and a spatial attention branch and a channel attention branch were constructed to obtain a mixed attention weight map. Secondly, a feature-interleaving module was constructed to convert the single-scale feature map of the specified layer into a multiscale feature fusion map to realize the combined encoding of high-level semantic information and low-level semantic information. Then, a feature extraction base network for lightweight multiscale feature fusion with an attention mechanism based on the above steps was constructed. Finally, a key-point detection network was constructed to output the location information, bias information, and category probability of the center points of traffic signs to achieve the detection and recognition of traffic signs. The model was trained, validated, and tested using TT100K datasets, and the detection accuracy of 36 common categories of traffic signs reached more than 85%, among which the detection accuracy of five categories exceeded 95%. The results showed that, compared with the traditional methods of Faster R-CNN, CornerNet, and CenterNet, traffic sign detection based on a lightweight multiscale feature fusion network had obvious advantages in the speed and accuracy of recognition, significantly improved the detection performance for small targets, and achieved a better real-time performance. [ABSTRACT FROM AUTHOR]

Urban transportation destination prediction is a crucial issue in the area of intelligent transportation, such as urban traffic planning and traffic congestion control. The spatial structure of the road network has high nonlinearity and complexity, and also, the traffic flow is dynamic due to the continuous changing of the traffic environment. Thus, it is very important to model the spatial relation and temporal dependence simultaneously to simulate the true traffic conditions. Most of the existing destination prediction methods have limited ability to model large-scale spatial data that changes dynamically with time, so they cannot obtain satisfactory prediction results. This paper proposes a human-in-loop Spatial-Temporal Attention Mechanism with Graph Convolutional Network (STAGCN) model to explore the spatial-temporal dependencies for destination prediction. The main contributions of this study are as follows. First, the traffic network is represented as a graph network by grid region dividing, then the spatial-temporal correlations of the traffic network can be learned by convolution operations in time on the graph network. Second, the attention mechanism is exploited for the analysis of features with loop periodicity and enhancing the features of key nodes in the grid. Finally, the spatial and temporal features are combined as the input of the Long-Short Term Memory network (LSTM) to further capture the spatial-temporal dependences of the traffic data to reach more accurate results. Extensive experiments conducted on the large scale urban real dataset show that the proposed STAGCN model has achieved better performance in urban car-hailing destination prediction compared with the traditional baseline models. [ABSTRACT FROM AUTHOR]

Traffic flow forecasting is of great significance to urban traffic control and public safety applications. The key challenge of traffic flow forecasting is how to capture the complex correlation of different time levels and learn time dependence. Some external information is closely related to traffic flow, such as accidental traffic accidents, weather, and Point of Interests (PoI) information. This paper proposes a deep learning-based model, called AttDeepSTN+, which is used to predict the inflow and outflow of each area of the entire city. Specifically, AttDeepSTN+ uses the structure of interactive attention and convolution to model the temporal closeness, trend, and periodicity of crowd flow, in the interactive attention layer, learn the importance of closeness to periodicity and trend respectively to model the long-term dependence of time, and then use feature fusion to capture complex correlations at different levels, thereby reducing model prediction accuracy. In addition, PoI information are combined with time factors to express the influence of location attributes on crowd flow, to learn prior knowledge of crowd flow. Finally, a new fusion mechanism is used to fuse the attention layer modules and PoI information and other information together into the module to capture the complex correlation between different levels of features, to predict the final crowd flow in each area, and further improve the prediction accuracy of the model. The New York City crowd flow experiment shows that the model is better than the current state-of-the-art baseline. [ABSTRACT FROM AUTHOR]

Network traffic classification is an increasingly significant prerequisite for network management. An accurate traffic classifier can contribute to traffic engineering, traffic intrusion detection and user behavior analysis. Recently, deep learning (DL) has attracted considerable attention for traffic classification due to its excellent learning ability without the need for handcrafted feature engineering or privacy invasion. However, most DL-based traffic classification solutions mainly focus on improving the identification performance, with no regard for the computational and memory burden caused by a large model size or data redundancy. Motivated by the above gaps, we elaborate on the preprocessing of raw traffic data and investigate a lightweight traffic classifier. Specifically, we design a preprocessing approach to convert raw traffic data into available datasets for deep learning based traffic classifiers, which tailors raw traffic data as training datasets by resolving its structure and content and pruning redundant information. Thanks to its ability to focus on key information, attention mechanism is introduced to design an attention-based long short-term memory (LSTM) model for traffic classification, termed the ABL-TC. ABL-TC can effectively identify global dependencies between the input and the output with a limited number of important raw features, which can contribute to learning key information to identify various traffic types. Extensive experiments on real-world public datasets show that the ABL-TC outperforms state-of-the-art approaches in terms of various metrics for recognizing traffic categories while remaining competitive in terms of time and memory efficiency. [ABSTRACT FROM AUTHOR]

Traffic flow forecasting (TFF) is critical for constructing intelligent transportation systems and offering real-time traffic applications, and especially accurate flow forecasting based on traffic big data can drive reliable strategies for traffic management and control. To consider the weight of the influence of the spatial correlation among the road segments and capture the nonlinear characteristics of traffic flow, this paper presents a parallel Normal Distribution and Attention Mechanism Weighted Deep Bidirectional Long Short-Term Memory (NAW-DBLSTM) algorithm on Spark. Specifically, we employ the resilient distributed data set (RDD) to preprocess large-scale mobile trajectory data (e.g., taxi GPS trajectory data), and then Kalman Filter (KF) is utilized to smooth the taxi trajectory big data. Next, the parallel NAW-DBLSTM algorithm is put forward on a Spark distributed computing platform to enhance the accuracy and scalability of TFF, combined with the attention mechanism and the normal distribution, and then the time window is used for TFF. Finally, the traffic flow is forecasted successfully on Spark by our NAW-DBLSTM algorithm with the real-world GPS trajectories of taxicabs. The experimental results demonstrate that, compared with LSTM, BiLSTM, DBLSTM, DNN, SVR, KNN, SAEs, BP, CNN, GRU, and ANNs, NAW-DBLSTM can produce better performance with the MAPE value that is 85.1%, 80.1%, 85.8%, 73.1%, 78.2%, 77.9%, 78.8%, 84.6%, 96.4%, 86.2%, and 73.2% lower than that of comparable algorithms, respectively. Particularly, the MAPE value of NAW-DBLSTM is 28.3%, 20.1%, 71.1%, and 79.1% lower than that of LSTM weighted with the normal distribution and the attention mechanism (NAW-LSTM), BiLSTM weighted with the normal distribution and the attention mechanism (NAW-BiLSTM), DBLSTM weighted with the normal distribution (NW-DBLSTM), and NAW-DBLSTM weighted without the time window (NT-NAW-DBLSTM), respectively. [ABSTRACT FROM AUTHOR]

Traffic volume is an indispensable ingredient to provide fine-grained information for traffic management and control. However, due to the limited deployment of traffic sensors, obtaining full-scale volume information is far from easy. Existing work on this topic focuses primarily on improving the overall estimation accuracy of a particular method and ignores the underlying challenges of volume estimation, thereby having inferior performance in some critical tasks. This paper studies two key problems with regard to traffic volume estimation: (1) underdetermined traffic flows caused by fully undetected paths that can allow arbitrary volume values without violating the conservation law, where using local side information is insufficient to tackle, and (2) nonequilibrium traffic flows arise when traffic flows vary in density over space and time due to congestion propagation delay, which produce time-shifted volume readings to varying degrees. Here we demonstrate a graph-based deep learning method that offers a data-driven, equation-free, and correlation-adaptive approach to address the above issues and perform accurate network-wide traffic volume estimation. Particularly, in order to quantify the dynamic and nonlinear speed-volume relationships for the estimation of underdetermined flows, a speed pattern-adaptive adjacency matrix based on graph attention is developed and integrated into the graph convolution process, to capture nonlocal correlations between volumes. To measure the impacts of nonequilibrium flows, a temporal masked and clipped attention combined with a gated temporal convolution layer is customized to capture time-asynchronous correlations between upstream and downstream sensors under varying impacts of congestion delay. We then evaluate our model on a real-world highway traffic volume dataset and compare it with several benchmark models. It is demonstrated that the proposed model achieves high estimation accuracy even under 20% sensor coverage rate and outperforms other baselines significantly, especially on underdetermined and nonequilibrium flow locations. Furthermore, comprehensive quantitative model analysis is also carried out to justify the model designs. The source code and datasets are publicly available at:. • Identifies underdetermined and non-equilibrium problems for network-wide traffic volume estimation. • Tailors a spatiotemporal correlation adaptive graph convolution network (STCAGCN) for volume estimation. • Exploits global speed-volume relationship as a remedy to underdetermined flows. • Captures time-asynchronous correlations in non-equilibrium flows with masked and clipped attention. • STCAGCN achieves state-of-the-art performances on all tasks. [ABSTRACT FROM AUTHOR]