Article
Transportation Science & Technology
Jinlong Li, Zhigang Xu, Lan Fu, Xuesong Zhou, Hongkai Yu
Summary: The article explores how to utilize daytime images to assist in nighttime vehicle detection, proposing a situation-sensitive method based on Faster R-CNN and domain adaptation. Experimental results using new datasets demonstrate the accuracy and effectiveness of the proposed method.
TRANSPORTATION RESEARCH PART C-EMERGING TECHNOLOGIES
(2021)
Article
Physics, Multidisciplinary
Haojia Lin, Zhilu Yuan, Biao He, Xi Kuai, Xiaoming Li, Renzhong Guo
Summary: This paper applies deep learning for vehicle counting in traffic videos. A method based on transfer learning is proposed to solve the problem of lacking annotated data. A vehicle counting method based on fusing the virtual detection area and vehicle tracking is proposed. Suppression modules are designed to improve the accuracy of vehicle counting.
FRONTIERS IN PHYSICS
(2022)
Article
Computer Science, Information Systems
Asfak Ali, Ram Sarkar, Debesh Kumar Das
Summary: One of the challenging tasks in computer vision is the classification and detection of vehicles. Researchers worldwide are working on autonomous vehicle detection systems, which have various practical applications. The current trend in AVD is deep learning techniques, although many Indian vehicles are not included in the existing detection datasets. In this research, a dataset for still-image-based vehicle detection is presented, including one class of pedestrians and 13 different types of vehicles commonly seen on Indian roads. A baseline result is provided using state-of-the-art deep learning models, and an ensemble-based object detection and classification model is proposed to further improve accuracy. The dataset consists of 4K images and 14.3K bounding boxes, providing researchers with annotated rectangular boxes for future use.
MULTIMEDIA TOOLS AND APPLICATIONS
(2023)
Article
Engineering, Civil
Alexandre Lombard, Ahmed Noubli, Abdeljalil Abbas-Turki, Nicolas Gaud, Stephane Galland
Summary: By utilizing inter-vehicular communication, individual right-of-way can be given to each vehicle to increase the throughput of intersections. This paper proposes a Deep Reinforcement Learning (DRL) approach to efficiently distribute the right-of-way, showing benefits such as increased flow and reduced CO2 emissions compared to traditional traffic lights and other cooperative scheduling policies.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Ying Gao, Jinlong Li, Zhigang Xu, Zhangqi Liu, Xiangmo Zhao, Jianhua Chen
Summary: This study proposes a new image-based traffic congestion estimation method, which first defines the traffic congestion accurately and integrates a traffic parameter layer into a CNN model. By training and testing with a large dataset of traffic images, the proposed method shows better efficiency and stability in various traffic conditions and weather scenarios.
EXPERT SYSTEMS WITH APPLICATIONS
(2021)
Article
Computer Science, Artificial Intelligence
Ming Jin, Chuanxia Sun, Yinglei Hu
Summary: Intelligent transportation systems play a significant role in alleviating traffic congestion, and real-time data collection and decision support systems can improve the accuracy of vehicle detection and recognition, increasing traffic efficiency.
Article
Computer Science, Artificial Intelligence
Prashant Deshmukh, G. S. R. Satyanarayana, Sudhan Majhi, Upendra Kumar Sahoo, Santos Kumar Das
Summary: Intelligent vehicle detection (IVD) plays a crucial role in intelligent traffic management systems, especially in undisciplined traffic environments. This paper proposes a swin transformer-based vehicle detection (STVD) framework that effectively addresses the multi-scale feature extraction problem and achieves high detection accuracy on various datasets.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Article
Engineering, Electrical & Electronic
Duo Li, Joan Lasenby
Summary: Real-time traffic monitoring is crucial for transportation management. This research proposes a method for estimating space occupancy of a single road segment using partially observed trajectories, specifically commercial fleet trajectories. The method formulates the traffic estimation as a video computing problem and utilizes trajectory data to generate video-like data. By embedding the video input using a specific strategy, a Revised Video Vision Transformer (RViViT) is employed for traffic state estimation. Experimental results on a public dataset demonstrate the effectiveness of the proposed method.
IEEE SENSORS JOURNAL
(2022)
Article
Green & Sustainable Science & Technology
Bo Peng, Hanbo Zhang, Ni Yang, Jiming Xie
Summary: This study proposed a method combining morphological detection and deep convolutional networks for locating and identifying vehicle models from UAV videos. The improved AlexNet* model achieved superior performance in vehicle recognition, demonstrating effective identification of UAV video targets.
Article
Computer Science, Theory & Methods
Jian Hou, Fangai Liu, Hui Lu, Zhiyuan Tan, Xuqiang Zhuang, Zhihong Tian
Summary: Malicious traffic detection is crucial for cyber security, and using flow as the detection object is deemed effective. This paper proposes a novel approach for detecting malicious traffic by utilizing only the packet header fields in the raw traffic to generate characteristic representations and employing a two-layer attention network to generate flow vectors. Experimental results demonstrate high accuracy rates and AUC-ROC values for binary and multi-class classification tasks.
JOURNAL OF PARALLEL AND DISTRIBUTED COMPUTING
(2022)
Article
Computer Science, Artificial Intelligence
Hesamodin Mohammadian, Ali A. Ghorbani, Arash Habibi Lashkari
Summary: Intrusion detection systems play a crucial role in defending networks against security threats. Deep neural networks have shown excellent performance in intrusion detection, but they are vulnerable to adversarial attacks. This paper proposes a new approach using Jacobian Saliency Map to generate adversarial examples for deep learning-based malicious network activity classification. The experiments demonstrate that the proposed method achieves better performance with fewer features compared to other attacks.
APPLIED SOFT COMPUTING
(2023)
Article
Environmental Sciences
Xiaohe Li, Jianping Wu
Summary: This study proposes a framework for extracting vehicle motion data from UAV videos captured under different weather conditions, improving YOLOv5 and introducing a new vehicle-tracking algorithm called SORT++. A new dataset of traffic images under various weather conditions is established to evaluate the proposed method for vehicle orientation detection.
Article
Engineering, Civil
Xiaohe Li, Jianping Wu
Summary: Unmanned aerial vehicles (UAVs) have been extensively used for collecting traffic data due to their flexibility, stability, and ease of operation. However, traditional horizontal detectors and rotated detectors are inefficient and less accurate for detecting vehicles in UAV videos. To address this issue, a framework based on YOLOv5-OBB object detection and DeepSORT-OBB tracking algorithms was proposed to extract highly accurate traffic data from UAV videos. The framework was tested using aerial videos recorded by a UAV-mounted high-definition camera and evaluated using reference data collected from an onboard high-precision sensor. The extracted traffic data, including trajectory, yaw angle, speed, and heading of vehicles, achieved an overall extraction accuracy of 98.5%, indicating the reliability of the proposed framework in extracting highly accurate traffic data.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2023)
Article
Engineering, Civil
Bing Liu, Yu Tang, Yuxiong Ji, Yu Shen, Yuchuan Du
Summary: This study proposes a deep reinforcement learning method to optimize ramp metering control using traffic video data, which results in lower travel times in the mainline, shorter vehicle queues at the on-ramp, and higher traffic flows downstream of the merging area compared to a state-of-the-practice method.
JOURNAL OF ADVANCED TRANSPORTATION
(2021)
Article
Engineering, Multidisciplinary
Liang Zhao, Menglin Li, Zili He, Shihao Ye, Hongliang Qin, Xiaoliang Zhu, Zhicheng Dai
Summary: Fatigue can lead to low efficiency and accidents. Detecting fatigue in the field of education can improve learning efficiency. This study develops a multimodal learning fatigue detection system using ECG and video signals to classify a learner's state into alert, normal, and fatigued. Experimental results show that the system outperforms other methods, achieving detection accuracies of 99.6% and 91.8% on two datasets.
Article
Physics, Multidisciplinary
Li He, Haifei Zhu, Tao Zhang, Honghong Yang, Yisheng Guan
Article
Computer Science, Artificial Intelligence
Li He, Hong Zhang
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2018)
Article
Automation & Control Systems
Li He, Nilanjan Ray, Yisheng Guan, Hong Zhang
IEEE TRANSACTIONS ON CYBERNETICS
(2019)
Article
Engineering, Electrical & Electronic
Jinming Wen, Li He, Fumin Zhu
IEEE COMMUNICATIONS MAGAZINE
(2018)
Article
Engineering, Aerospace
Tao Zhang, Xilun Ding, Kun Xu, Shuting Liu, Li He, Haifei Zhu, Yisheng Guan
Article
Chemistry, Multidisciplinary
Weinan Chen, Lei Zhu, Li He, Yisheng Guan, Hong Zhang
APPLIED SCIENCES-BASEL
(2019)
Article
Automation & Control Systems
Honghong Yang, Jinming Wen, Xiaojun Wu, Li He, Shahid Mumtaz
IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS
(2019)
Article
Automation & Control Systems
Jian Li, Haifei Zhu, Tao Zhang, Li He, Yisheng Guan, Hong Zhang
INTERNATIONAL JOURNAL OF ADVANCED MANUFACTURING TECHNOLOGY
(2020)
Article
Automation & Control Systems
Yisheng Guan, Zhaoheng Zeng, Daye Chen, Tao Zhang, Haifei Zhu, Li He
Summary: The article introduces the Essboard, a unique two-wheeled vehicle with a special structure and locomotion mode, focusing on its kinematics. The geometric model and kinematic equations of the system are established, and the accuracy of the model is verified through simulation and experimentation. The developed kinematics are general and can be applied to similar monorail vehicles such as bicycles and motorcycles.
IEEE-ASME TRANSACTIONS ON MECHATRONICS
(2021)
Article
Automation & Control Systems
Weinan Chen, Lei Zhu, Xubin Lin, Li He, Yisheng Guan, Hong Zhang
Summary: This article introduces a dynamic keyframe selection strategy that dynamically adjusts the threshold for keyframe selection based on the view change between camera observation and keyframes in the built map. The proposed method improves the precision of visual tracking compared with existing studies.
IEEE-ASME TRANSACTIONS ON MECHATRONICS
(2022)
Article
Automation & Control Systems
Xubin Lin, Jiahao Ruan, Yirui Yang, Li He, Yisheng Guan, Hong Zhang
Summary: In this paper, a 2D motion inference method based on local projective warping consistency is proposed, along with an object association method called HOA that integrates deep appearance feature and semantic information. The proposed methods enhance the accuracy and robustness of object association under detection deficiency.
IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING
(2023)
Proceedings Paper
Computer Science, Artificial Intelligence
Guangcheng Chen, Li He, Yisheng Guan, Hong Zhang
Summary: Current polarimetric 3D reconstruction methods developed under the orthographic projection assumption may result in significant errors when applied to a large field of view. To address this problem, a perspective phase angle (PPA) model is proposed in this study. The PPA model accurately describes the relationship between polarization phase angle and surface normal under perspective projection, and enables surface normal estimation using only one single-view phase angle map without suffering from the p-ambiguity problem. Experimental results demonstrate that the PPA model is more accurate for surface normal estimation with a perspective camera than the orthographic model.
COMPUTER VISION - ECCV 2022, PT II
(2022)
Article
Engineering, Electrical & Electronic
Lei Zhu, Weinan Chen, Xubin Lin, Li He, Yisheng Guan
Summary: The research proposes a curvature variation based sampling method for point cloud classification and segmentation, which is motivated by the observation that points with high curvature variation can depict object outlines effectively. By combining this method with existing sampling techniques, a higher accuracy and mean IoU can be achieved, demonstrating the advantage of considering curvature variation in classification and segmentation tasks.
IEEE SIGNAL PROCESSING LETTERS
(2022)
Article
Engineering, Electrical & Electronic
Weinan Chen, Lei Zhu, Chaoqun Wang, Li He, Max Q. -H. Meng
Summary: This study introduces a new method to predict localization errors by considering both the spatial distribution and uncertainty of visual landmarks, and further improve navigation accuracy by additional mapping. Experimental results demonstrate a strong correlation between predicted and actual errors, showing significant enhancement in localization precision with the proposed approach.
IEEE SENSORS JOURNAL
(2021)
Article
Computer Science, Information Systems
Chengcai Leng, Hai Zhang, Bo Li, Guorong Cai, Zhao Pei, Li He