检索结果-内蒙古大学图书馆

A Novel Approach for Estimating Vehicle Speed in Nighttime Traffic Accidents from Daytime Video Information

学校读者我要写书评

暂无评论

A Novel Approach for Estimating Vehicle Speed in Nighttime T...

International Conference on Intelligent Transportation

作者： Youngsoo Choi Yongmun Yun Woo-Jeong Jeon Seung-Hyun Kong Robotics Program Korea Advanced Institute of Science and Technology Engineering Division Daejeon Institute National Forensic Service Republic of Korea Traffic Accident Analysis Division National Forensic Service Republic of Korea Graduate School of Mobility Korea Advanced Institute of Science and Technology Republic of Korea

In vehicle accident investigations, the pre-accident speed of the vehicle is an important factor that affects the reliability of accident reconstruction results. However, it is difficult to analyze vehicle speed in nighttime videos where the vehicle and its surroundings are not clearly visible. This study proposes a novel approach to estimate the speed of an accident vehicle in nighttime videos using daytime videos. The proposed method first estimates some missing coordinates from the wheel center coordinate data extracted from all daytime video sequences. Then, the vehicle speed in the nighttime video is estimated with a calibration curve that transforms the inter-pixel distance in the 2D video to the distance in the real 3D space using the cross-ratio. Experimental results show that the proposed Kalman filter-based method is effective in estimating the occluded coordinates and is useful in calculating the cross-ratio. Additionally, the vehicle speeds in the nighttime video estimated by the proposed method are similar to the simulated and measured actual vehicle speeds, as well as the real accident analysis case. These results show that the proposed method can be used to complement and effectively cross-validate existing methods to improve reliability in the field of vehicle accident investigations.

关键词：

Comparative Study on Fall Detection using Machine Learning Approaches 19

学校读者我要写书评

暂无评论

Comparative Study on Fall Detection using Machine Learning A...

19th International Conference on Electrical Engineering/Electronics, Computer, Telecommunications and Information Technology, ECTI-CON 2022

作者： Tongskulroongruang, Thienrawit Wiphunawat, Paphatchaya Jutharee, Wisanu Kaewmahanin, Watchara Rassameecharoenchai, Teerameth Jennawasin, Tanagorn Kaewkamnerdpong, Boonserm King Mongkut's University of Technology Thonburi Department of Control Systems and Instrumentation Engineering Bangkok Thailand Institute of Field Robotics and Automation King Mongkut's University of Technology Thonburi Bangkok Thailand Naresuan University Faculty of Allied Health Sciences Department of Cardiothoracic Technology Phitsanulok Thailand King Mongkut's University of Technology Thonburi Biological Engineering Program Faculty of Engineering Bangkok Thailand

ISBN: (纸本)9781665485845

Falls are one of the most dangerous problems for the elderly. A reliable fall detection system can aid in reducing the harmful repercussions of an unintentional fall. The focus of this paper is on a dataset that includes signals from portable devices. We propose a simple feature selection approach to reduce the number of effective input attributes based on a voting strategy. The dataset is classified into three categories, i.e., pre-fall, fall and post-fall using various machine learning approaches. The computational time is significantly reduced for most of the classification algorithms, and the relative reduction reaches to 66.67% with decision tree algorithm. Classification accuracy can reach as high as 95.17% when using neural network, while the relative reductions on the classification accuracy due to the feature selection do not exceed 1.50%. © 2022 IEEE.

关键词： Decision trees

RelMobNet: End-to-end relative camera pose estimation using a robust two-stage training

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Rajendran, Praveen Kumar Mishra, Sumit Vecchietti, Luiz Felipe Har, Dongsoo Division of Future Vehicle KAIST Korea Republic of The Robotics Program KAIST Korea Republic of Data Science Group Institute for Basic Science Korea Republic of The CCS Graduate School of Mobility KAIST Korea Republic of

Relative camera pose estimation, i.e. estimating the translation and rotation vectors using a pair of images taken in different locations, is an important part of systems in augmented reality and robotics. In this paper, we present an end-to-end relative camera pose estimation network using a siamese architecture that is independent of camera parameters. The network is trained using the Cambridge Landmarks data with four individual scene datasets and a dataset combining the four scenes. To improve generalization, we propose a novel two-stage training that alleviates the need of a hyperparameter to balance the translation and rotation loss scale. The proposed method is compared with one-stage training CNN-based methods such as RPNet and RCPNet and demonstrate that the proposed model improves translation vector estimation by 16.11%, 28.88%, and 52.27% on the Kings College, Old Hospital, and St Marys Church scenes, respectively. For proving texture invariance, we investigate the generalization of the proposed method augmenting the datasets to different scene styles, as ablation studies, using generative adversarial networks. Also, we present a qualitative assessment of epipolar lines of our network predictions and ground truth poses. © 2022, CC BY.

关键词： Cameras

Automatic Facial Asymmetry Analysis for Elderly Stroke Detection by using Cosine Similarity 19

学校读者我要写书评

暂无评论

Automatic Facial Asymmetry Analysis for Elderly Stroke Detec...

19th International Conference on Electrical Engineering/Electronics, Computer, Telecommunications and Information Technology, ECTI-CON 2022

作者： Kaewmahanin, Watchara Rassameecharoenchai, Teerameth Jutharee, Wisanu Tongskulroongruang, Thienrawit Wiphunawat, Paphatchaya Jennawasin, Tanagorn Kaewkamnerdpong, Boonserm Naresuan University Faculty of Allied Health Sciences Department of Cardiothoracic Technology Phitsanulok Thailand King Mongkut's University of Technology Thonburi Institute of Field Robotics and Automation Bangkok Thailand Mongkut's University of Technology Thonburi Department of Control Systems and Instrumentation Engineering Bangkok Thailand King Mongkut's University of Technology Thonburi Biological Engineering Program Faculty of Engineering Bangkok Thailand

ISBN: (纸本)9781665485845

For acute stroke patients, time to recognize the stroke symptom onset is crucial for the lifesaving treatment. The automatic detection of stroke signs has been increasingly developed for practical use. The better timely classification results could allow better detection of the symptoms and avoid morbidity and mortality. In this study, we proposed using cosine similarity between the left and right sides of the face as a part of features for classifying the stroke elderly patients from healthy elderly participants. We employed our method on images from the Toronto NeuroFace dataset. The results showed high classification accuracy (the average accuracy of 97.8998% from 6 classification methods). These results indicate that the cosine similarity features can be useful to detect facial asymmetry and pose very promising potential for elderly stroke detection. © 2022 IEEE.

关键词： Patient treatment

InstaGraM: Instance-level Graph Modeling for Vectorized HD Map Learning

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Shin, Juyeb Jeong, Hyeonjun Rameau, Francois Kum, Dongsuk Robotics Program Korea Advanced Institute of Science and Technology Daejeon34051 Korea Republic of Graduate School of Mobility Korea Advanced Institute of Science and Technology Daejeon34051 Korea Republic of Computer Science department State University of New York - Korea Songdo21985 Korea Republic of

For scalable autonomous driving, a robust map-based localization system, independent of GPS, is fundamental. To achieve such map-based localization, online high-definition (HD) map construction plays a significant role in accurate estimation of the pose. Although recent advancements in online HD map construction have predominantly investigated on vectorized representation due to its effectiveness, they suffer from computational cost and fixed parametric model, which limit scalability. To alleviate these limitations, we propose a novel HD map learning framework that leverages graph modeling. This framework is designed to learn the construction of diverse geometric shapes, thereby enhancing the scalability of HD map construction. Our approach involves representing the map elements as an instance-level graph by decomposing them into vertices and edges to facilitate accurate and efficient end-to-end vectorized HD map learning. Furthermore, we introduce an association strategy using a Graph Neural Network to efficiently handle the complex geometry of various map elements, while maintaining scalability. Comprehensive experiments on public open dataset show that our proposed network outperforms state-of-the-art model by 1.6 mAP. We further showcase the superior scalability of our approach compared to state-of-the-art methods, achieving a 4.8 mAP improvement in long range configuration. Our code is available at https://***/juyebshin/InstaGraM. Copyright © 2023, The Authors. All rights reserved.

关键词： Graph neural networks

Enabling Human-Robot Collaboration via Holistic Human Perception and Partner-Aware Control

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Yeshasvi Tirupachuri, Venkata Sai University of Genova PhD Program in Bioengineering and Robotics

Dual Quaternion-Based Control for Dynamic Robot Formations

学校读者我要写书评

暂无评论

Dual Quaternion-Based Control for Dynamic Robot Formations

International Conference on Unmanned Aircraft Systems (ICUAS)

作者： Juan I. Giribet Harrison Neves Marciano Ignacio Mas Alejandro S. Ghersin Daniel Khede Dourado Villa Mário Sarcinelli-Filho Artificial Intelligence and Robotics Lab. (LINAR) Universidad de San Andrés Argentina CONICET - Consejo Nacional de Investigaciones Científicas y Técnicas Argentina Graduate Program on Electrical Engineering Universidade Federal do Espírito Santo Vitória ES Brazil Instituto Tecnológico de Buenos Aires Argentina

ISBN: (数字)9798331513283

ISBN: (纸本)9798331513290

This paper evaluates a dual quaternion-based control algorithm designed to manage dynamic changes in the number of vehicles in robot formations. By defining a virtual structure, the algorithm coordinates the formation's position, orientation, and shape parameters, ensuring seamless transitions when the number of vehicles changes. The approach enables a low-level controller to calculate commands for individual robots while maintaining the overall formation integrity. The strategy's performance is analyzed through simulations and experimental results, demonstrating its effectiveness in handling variations in the number of vehicles of robotic formations.

关键词： Geometry Analytical models Shape Robot kinematics Heuristic algorithms Quaternions Transportation Autonomous aerial vehicles Stability analysis Vehicle dynamics

Learning Terminal State of the Trajectory Planner: Application for Collision Scenarios of Autonomous Vehicles

学校读者我要写书评

暂无评论

Learning Terminal State of the Trajectory Planner: Applicati...

IEEE International Conference on robotics and Automation (ICRA)

作者： Joonhee Lim Kibeom Lee Jangho Shin Dongsuk Kum Robotics Program Korea Advanced Institute of Science and Technology (KAIST) Daejeon Republic of Korea Department of Mechanical Engineering Gachon University Seongnam Republic of Korea Research & Development Division Hyundai Motor Company Hwaseong Republic of Korea The Cho Chun Shik Graduate School of Mobility Korea Advanced Institute of Science and Technology (KAIST) Daejeon Republic of Korea

ISBN: (数字)9798350384574

ISBN: (纸本)9798350384581

Collision Avoidance/Mitigation System (CAMS) for autonomous vehicles is a crucial technology that ensures the safety and reliability of autonomous driving systems. Conventional collision avoidance approaches struggle in complex and various scenarios by avoiding collisions based on rules for specific collision scenarios. This has led to learning-based methods using neural networks for adaptive collision avoidance. However, the approaches directly outputting control inputs through neural networks have drawbacks in interpretability and stability. To address these limitations, we propose a trajectory planning method for CAMS that combines deep reinforcement learning (DRL) and quintic polynomial (QP) trajectory planning. The proposed method determines the terminal state and confidence of the trajectory using DRL and plans a QP trajectory based on them. By utilizing the terminal state and confidence of the trajectory rather than direct control inputs as the output of the neural network, it generates a more realistic and continuous path. Moreover, this approach considers collision avoidance and mitigation in an integrated manner through the reward function of RL. Our experimental results demonstrate that the proposed method not only improves interpretability and stability compared to existing learning-based methods but also upholds performance in complex and various collision scenarios.

关键词： Learning systems Trajectory planning Neural networks Trajectory Cams Safety Reliability

Tactile Perception in Upper Limb Prostheses: Mechanical Characterization, Human Experiments, and Computational Findings

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Ivani, Alessia S. Catalano, Manuel G. Grioli, Giorgio Bianchi, Matteo Visell, Yon Bicchi, Antonio Soft Robotics for Human Cooperation and Rehabilitation Istituto Italiano di Tecnologia Genova16163 Italy Centro di ricerca E. Piaggio University of Pisa Pisa56122 Italy Department of Information Engineering University of Pisa Pisa56122 Italy Department of Mechanical Engineering and Bioengineering Program University of California Santa Barbara United States

Tactile feedback is essential for upper-limb prostheses functionality and embodiment, yet its practical implementation presents challenges. Users must adapt to non-physiological signals, increasing cognitive load. However, some prosthetic devices transmit tactile information through socket vibrations, even to untrained individuals. Our experiments validated this observation, demonstrating a user's surprising ability to identify contacted fingers with a purely passive, cosmetic hand. Further experiments with advanced soft articulated hands revealed decreased performance in tactile information relayed by socket vibrations as hand complexity increased. To understand the underlying mechanisms, we conducted numerical and mechanical vibration tests on four prostheses of varying complexity. Additionally, a machine-learning classifier identified the contacted finger based on measured socket signals. Quantitative results confirmed that rigid hands facilitated contact discrimination, achieving 83% accuracy in distinguishing index finger contacts from others. While human discrimination decreased with advanced hands, machine learning surpassed human performance. These findings suggest that rigid prostheses provide natural vibration transmission, potentially reducing the need for tactile feedback devices, which advanced hands may require. Nonetheless, the possibility of machine learning algorithms outperforming human discrimination indicates potential to enhance socket vibrations through active sensing and actuation, bridging the gap in vibration-transmitted tactile discrimination between rigid and advanced hands. © 2024, CC BY-NC-ND.

关键词： Knee prostheses