Article
Computer Science, Information Systems
Yantao Yu, Heng Li, Jiannong Cao, Xiaochun Luo
Summary: This article proposed a monocular-camera-based 3-D estimation method suitable for industrial working poses, utilizing a residual artificial neural network (RANN) with flexible complexity and weighted training loss. By establishing a 3-D pose data set containing diverse working poses in worksites, the network's performance in complex scenarios was evaluated. Compared to previous 3-D pose capture methods, the mean per joint position error was reduced by 31.42%, with a latency of 0.24 s. Hence, the proposed monocular-camera-based method shows great potential in industrial application scenarios.
IEEE INTERNET OF THINGS JOURNAL
(2021)
Article
Computer Science, Hardware & Architecture
Chao Yang, Xuyu Wang, Shiwen Mao
Summary: A real-time 3-D human pose estimation system based on RFID tags and deep learning is proposed in this study, achieving high accuracy pose tracking through calibrating RFID data and completing missing data. Experimental results demonstrate that the system is capable of efficiently reconstructing human pose in realtime.
IEEE TRANSACTIONS ON RELIABILITY
(2021)
Article
Computer Science, Artificial Intelligence
Jiahao Xia, Haimin Zhang, Shiping Wen, Shuo Yang, Min Xu
Summary: The study introduces an efficient multitask neural network ATPN for face alignment, face tracking, and head pose estimation. ATPN improves performance in face alignment with shortcut connections and enhances other tasks with heatmap fusion, while also saving time in face detection and increasing real-time capability for video tasks.
EXPERT SYSTEMS WITH APPLICATIONS
(2022)
Article
Construction & Building Technology
Gang Peng, Hangqi Duan, Zejie Tan, Yicheng Zhou, Jianfeng Li, Bin Hu, Cheng Zhou
Summary: The rapid development of automatic control technology and AI has sparked interest in driverless systems. While research on driverless systems is mainly focused on the automotive field, it also has significant potential in engineering machinery. This article specifically focuses on the field of unmanned bulldozers, where path tracking and pose estimation are crucial for construction quality and safety. The article presents a construction path tracking method and improved pose estimation methods that outperform general SLAM systems in terms of precision and dynamic performance.
AUTOMATION IN CONSTRUCTION
(2023)
Article
Automation & Control Systems
Tomohide Takeyama, Hideyuki O-Tani, Satoru Oishi, Muneo Hori, Atsushi Iizuka
Summary: In this study, a program was developed to create mediated data for constructing analytical models with a common data structure for numerical analysis. A grid model of the 3-D ground surface was built based on borehole data, enabling efficient updates and cost-effective changes in the analytical model for disaster response and project planning.
IEEE-ASME TRANSACTIONS ON MECHATRONICS
(2021)
Article
Chemistry, Multidisciplinary
Haijian Wang, Qingxuan Shi, Beiguang Shan
Summary: Three-dimensional human pose estimation is a hot research topic in computer vision. In this study, a new method called STFormer is proposed to address the issues of self-occlusion and depth ambiguity. The framework consists of two stages: feature extraction from temporal and spatial domains, and modeling the communication of information across domains. Experimental results on Human3.6 dataset show that STFormer outperforms recent methods and reduces MPJPE by 2.1% compared to PoseFormer. Ablation studies are also conducted to analyze the validity of the constituent modules of STFormer.
APPLIED SCIENCES-BASEL
(2023)
Article
Geography, Physical
Christoph Blut, Joerg Blankenbach
Summary: Augmented reality systems, particularly in mobile AR, have application possibilities beyond entertainment, such as in civil engineering and city planning. Combining AR with CityGML building models can simplify planning processes and optimize communication among decision-makers. Research shows that CityGML and low-cost mobile devices can provide accurate pose tracking for AR in civil engineering and city planning.
INTERNATIONAL JOURNAL OF DIGITAL EARTH
(2021)
Article
Chemistry, Multidisciplinary
Alireza Ansaripour, Milad Heydariaan, Kyungki Kim, Omprakash Gnawali, Hafiz Oyediran
Summary: Pose estimation is crucial for real-time safety monitoring in road construction sites. UWB radios are considered as promising sensing technology for accurate object localization. However, the performance of UWB radios declines in road construction environments due to blockages, resulting in Non-Line of Sight (NLOS) situations. To address this issue, a real-time pose estimating system called ViPER+ is proposed, which overcomes NLOS situations and accurately determines the boundary of heavy construction equipment using multiple UWB tags. Evaluation results show that embedding NLOS detection technique in UWB-based pose estimation improves location accuracy by 40% and update rate by 25% compared to the previous implementation (ViPER).
APPLIED SCIENCES-BASEL
(2023)
Article
Biochemical Research Methods
Jan Stenum, Cristina Rossi, Ryan T. Roemmich
Summary: The author suggests that novel pose estimation algorithms can automatically track human movement for analyzing human gait, moving away from the dependence on laboratories. Their research shows that their workflow can accurately estimate gait parameters, and discusses the impact of camera viewpoint on the results.
PLOS COMPUTATIONAL BIOLOGY
(2021)
Review
Construction & Building Technology
Guangxu Liu, Qingfeng Wang, Tao Wang, Bingcheng Li, Xiangshuo Xi
Summary: This paper explores the applications and future trends of vision-based manipulator pose measurement in hydraulic excavators through literature review and quantitative analysis, emphasizing the relevant challenges involved.
AUTOMATION IN CONSTRUCTION
(2024)
Article
Engineering, Electrical & Electronic
Asma Boudrioua, Abdelouahab Aloui, Basel Solaiman, Larbi Asli, Douraied Ben Salem, Souhil Tliba
Summary: Brain tumor segmentation is crucial for diagnostic radiology. The proposed approach utilizes variational level set method for efficient and accurate tumor volume estimation. The method achieves high accuracy and fast execution speed in experiments, outperforming 18 state-of-the-art methods.
INTERNATIONAL JOURNAL OF IMAGING SYSTEMS AND TECHNOLOGY
(2021)
Article
Chemistry, Analytical
George Retsinas, Niki Efthymiou, Dafni Anagnostopoulou, Petros Maragos
Summary: Agricultural robotics focuses on developing robotic systems that can efficiently handle various agricultural tasks. This paper specifically discusses the case of mushroom collection in industrial farms, where a well-performing perception module is necessary for accurately detecting the 3D pose of mushrooms. The authors propose a vision module using multiple RealSense cameras for 3D pose estimation of mushrooms from multi-view point clouds. They address the challenge of limited annotation data by developing a pipeline for mushroom instance segmentation and template matching using only a 3D model of a mushroom. The proposed approach is quantitatively evaluated on a synthetic dataset and qualitatively validated on real data collected under different vision settings.
Article
Computer Science, Information Systems
Yeonggwang Kim, Giwon Ku, Chulseung Yang, Jeonggi Lee, Jinsul Kim
Summary: In this study, a novel transformer-based model with independent tokens is proposed for estimating 3D human pose and shape from monocular videos, with a focus on rehabilitation therapy. The model incorporates joint rotation and camera tokens to enhance performance and utilizes a temporal model to reduce jitter. The results show comparable performance with current best-performing models.
Article
Engineering, Mechanical
Chujin Sun, Donglian Gu, Xinzheng Lu
Summary: This study proposes a 3D structural displacement measurement method using monocular vision and deep learning based pose estimation. The method synthesizes a training set using virtual rendering, trains a deep learning model called DPOD to estimate the poses of the target object, and finally measures the 3D translation of the structures based on the original and destination poses or the original pose and keypoint matching.
MECHANICAL SYSTEMS AND SIGNAL PROCESSING
(2023)
Article
Construction & Building Technology
Jingyuan Tang, Han Luo, Weiwei Chen, Peter Kok-Yiu Wong, Jack C. P. Cheng
Summary: This study investigates a method for IMU-based full-body pose estimation of construction machines, and proposes a kinematics-based trajectory generation method and an optimal installation scheme for IMU sensors.
AUTOMATION IN CONSTRUCTION
(2022)
Article
Computer Science, Information Systems
Oscar Perez-Gil, Rafael Barea, Elena Lopez-Guillen, Luis M. Bergasa, Carlos Gomez-Huelamo, Rodrigo Gutierrez, Alejandro Diaz-Diaz
Summary: This paper proposes the use of deep learning algorithms in the control layer of autonomous vehicles and compares the performance of DQN and DDPG. The results show that DDPG outperforms DQN in terms of trajectory and accuracy.
MULTIMEDIA TOOLS AND APPLICATIONS
(2022)
Article
Computer Science, Information Systems
Carlos Gomez-Huelamo, Javier Del Egido, Luis Miguel Bergasa, Rafael Barea, Elena Lopez-Guillen, Javier Araluce, Miguel Antunes
Summary: This paper presents a real-time and power-efficient method for 3D Multi-Object Detection and Tracking in autonomous vehicles. The method allows the vehicle to track surrounding objects and predict trajectories to prevent collisions. The proposed pipelines utilize lightweight Linux containers and the Robot Operating System for system isolation and flexibility. The method is validated using the KITTI-3DMOT evaluation tool and tested in both simulation and real-world environments.
MULTIMEDIA TOOLS AND APPLICATIONS
(2022)
Article
Engineering, Mechanical
Peijun Fang, Yingfeng Cai, Long Chen, Hai Wang, Yicheng Li, Miguel Angel Sotelo, Zhixiong Li
Summary: In this paper, a new vehicle dynamics model based on a neural network is proposed, which can predict the dynamics of a vehicle more accurately. The model is established using data-driven methods and has long-term memory cells, which can adapt to different road conditions. A trajectory tracking control algorithm is designed based on this model, and the simulation analysis shows the effectiveness of the algorithm.
PROCEEDINGS OF THE INSTITUTION OF MECHANICAL ENGINEERS PART D-JOURNAL OF AUTOMOBILE ENGINEERING
(2023)
Article
Computer Science, Artificial Intelligence
Ruben Izquierdo, Alvaro Quintanar, David Fernandez Llorca, Ivan Garcia Daza, Noelia Hernandez, Ignacio Parra, Miguel Angel Sotelo
Summary: This work presents a novel method for predicting vehicle trajectories in highway scenarios using efficient bird's eye view representations and convolutional neural networks. The study found that a U-net model with 6 depth levels using a linear terminal layer and a Gaussian representation of the vehicles is the best performing configuration. The prediction error is up to 50% lower compared to the baseline method.
APPLIED INTELLIGENCE
(2023)
Article
Engineering, Civil
Qishui Zhong, Jin Yang, Kaibo Shi, Shouming Zhong, Zhixiong Li, Miguel Angel Sotelo
Summary: This article presents a new event-triggered $H_{infinity}$ load frequency control approach with dynamic triggered algorithm and non-fragile proportional integral control strategy, aiming to reduce the communication bandwidth usage and data computation burden in multi-area nonlinear power systems.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2022)
Article
Engineering, Civil
Guibing Zhu, Yong Ma, Zhixiong Li, Reza Malekian, M. Sotelo
Summary: This study investigates the tracking control problem of marine surface vessels (MSVs) in the presence of uncertain dynamics and external disturbances, considering undesirable faults and input saturation of actuators. A novel control scheme is proposed using a saturation function, event-triggered mechanism, and neural network technique, which is robust, adaptive, tolerant, and guarantees stable tracking of MSVs without prior knowledge of dynamics or faults. Simulation results demonstrate the effectiveness of the proposed scheme.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2022)
Article
Acoustics
Bo Zhu, Qishui Zhong, Yinsheng Chen, Shaohui Liao, Zhixiong Li, Kaibo Shi, Miguel Angel Sotelo
Summary: This article proposes a novel two-step reconstruction method for precise temperature distribution measurement, which provides high-resolution images and maintains high accuracy. By utilizing equilibrium optimizer and Gaussian process regression techniques, it effectively improves the resolution of temperature distribution images and reduces reconstruction errors.
IEEE TRANSACTIONS ON ULTRASONICS FERROELECTRICS AND FREQUENCY CONTROL
(2022)
Article
Automation & Control Systems
Dezheng Hua, Xinhua Liu, He Lu, Shuaishuai Sun, Miguel Angel Sotelo, Zhixiong Li, Weihua Li
Summary: This article proposes a ferrofluid soft capsule robot (FSCR) with magnetic actuation for diagnostic and therapeutic medical applications inside the human stomach. The FSCR features a composite shell with improved magnetic circuit design, controllable locomotion, and advanced functions such as drug-releasing.
IEEE-ASME TRANSACTIONS ON MECHATRONICS
(2022)
Article
Computer Science, Artificial Intelligence
Yicheng Li, Yingfeng Cai, Zhixiong Li, Shizhe Feng, Hai Wang, Miguel Angel Sotelo
Summary: This study proposes a map-based localization method using bi-sensor data fusion. A multi-scale map is generated using a camera, LRF, differential GPS, and inertial navigation system. The method includes coarse, node-level, and metric localizations for accurate positioning. Experimental results show that the proposed method achieves low localization errors.
EXPERT SYSTEMS WITH APPLICATIONS
(2022)
Article
Engineering, Civil
Yong Ma, Yujiao Zhao, Zhixiong Li, Huaxiong Bi, Jing Wang, Reza Malekian, Miguel Angel Sotelo
Summary: This work focuses on collaborative coverage path planning for unmanned surface mapping vehicles (USMVs) and proposes an improved algorithm called CCIBA*. The algorithm achieves coverage path planning for a single vehicle through task decomposition and level map updating, and designs a collaborative behavior strategy for multiple USMVs, resulting in high efficiency and high quality coverage paths.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2022)
Article
Engineering, Civil
Guibing Zhu, Yong Ma, Zhixiong Li, Reza Malekian, M. Sotelo
Summary: This paper proposes a novel control solution for marine surface vehicles (MSVs) without velocity information and subject to internal and external uncertainties. By utilizing the advantages of adaptive neural network and disturbance observer, a classification reconstruction idea is developed and integrated into the control design. The proposed solution uses a vector-backstepping design framework, a serial-parallel estimation model, and a dynamic event triggering protocol to improve control performance and reduce mechanical wear of actuator.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2023)
Article
Engineering, Civil
Hai Wang, Yanyan Chen, Yingfeng Cai, Long Chen, Yicheng Li, Miguel Angel Sotelo, Zhixiong Li
Summary: Considerable progress has been made in semantic segmentation of images in favorable environments in recent years, but the environmental perception of autonomous driving under adverse weather conditions remains challenging. This paper aims to explore image segmentation in low-light scenarios to expand the application range of autonomous vehicles. We propose a novel nighttime segmentation framework and demonstrate its effectiveness through experiments.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2022)
Article
Engineering, Electrical & Electronic
Xin Hu, Guibing Zhu, Yong Ma, Zhixiong Li, Reza Malekian, Miguel Angel Sotelo
Summary: This article investigates dynamic event-triggered adaptive neural coordinated disturbance rejection for marine vehicles, addressing unknown sinusoidal disturbances in terms of their frequencies, amplitudes, and phases. The article transforms the mathematical models of vehicle movements into parameterized expressions using neural networks for nonlinear dynamics approximation. It exploits parametric exogenous systems to express external disturbances, which are then converted into linear canonical models with coordinated changes. The proposed approach involves adaptive techniques and disturbance filters to estimate and reject disturbances. Through vectorial backstepping, a dynamic event-triggered adaptive neural coordinated disturbance rejection controller is derived, incorporating dynamic event-triggering conditions to reduce the execution frequency of the vehicle's propulsion systems. The closed-loop semi-global stability enables coordinated formation control. The proposed scheme achieves disturbance estimation and cancellation without requiring a priori knowledge of the marine vehicle's model dynamics, as validated by illustrative simulations and comparisons.
IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY
(2023)
Article
Engineering, Civil
Hai Wang, Le Tao, Yingfeng Cai, Long Chen, Yicheng Li, Miguel Angel Sotelo, Zhixiong Li
Summary: In this paper, a single-stage anchor-free 3-D object detector algorithm called CenterPoint-SE is proposed, which enhances the spatial perception ability of anchor-free detection networks based on CenterPoints. An efficient 3-D backbone network is constructed to extract fine-grained spatial geometric features, and a powerful spatial semantic feature fusion module called EF-Fusion is designed. Various improvements, including a lightweight IoU prediction branch and a foreground point segmentation auxiliary training branch, are added to enhance the algorithm's performance.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Sandra Carrasco Limeros, Sylwia Majchrowska, Joakim Johnander, Christoffer Petersson, Miguel Angel Sotelo, David Fernandez Llorca
Summary: Predicting the motion of other road agents is crucial for safe and efficient path planning of autonomous vehicles. Existing multi-modal motion prediction approaches rely on complex machine learning systems with limited interpretability. This study proposes new evaluation criteria, robustness assessment, and an intent prediction layer for enhancing the design of trustworthy motion prediction systems.
CAAI TRANSACTIONS ON INTELLIGENCE TECHNOLOGY
(2023)
