Article
Computer Science, Artificial Intelligence
Jingwen Su, Boyan Xu, Hujun Yin
Summary: This paper presents an extensive review of deep learning methods for image restoration tasks. Deep learning techniques, particularly convolutional neural networks, have been widely used in image processing, but image restoration remains a challenging topic. This paper compares deep learning techniques for image denoising, deblurring, dehazing, and super-resolution, and summarizes the principles and methods involved.
Article
Mechanics
Yiqian Mao, Shan Zhong, Hujun Yin
Summary: This study introduces a hybrid DRL method with a Markov decision process (MDP) with time delays and a first-order autoregressive policy (ARP) to control the vortex-shedding process of a two-dimensional circular cylinder. Compared to the standard DRL method, this method achieves a more stable and effective reduction of force fluctuations in the vortex-shedding process.
Article
Multidisciplinary Sciences
Yao Peng, Mary M. Dallas, Jose T. Ascencio-Ibanez, J. Steen Hoyer, James Legg, Linda Hanley-Bowdoin, Bruce Grieve, Hujun Yin
Summary: Cassava brown streak disease (CBSD) is an emerging viral disease that can greatly reduce cassava productivity. Current detection techniques are labor-intensive and inaccurate. We have developed a handheld active multispectral imaging (A-MSI) device combined with machine learning for real-time early detection of CBSD. The technique offers improved spectral signal-to-noise ratio and temporal repeatability, and can reliably distinguish healthy cassava from infected plants.
SCIENTIFIC REPORTS
(2022)
Article
Computer Science, Artificial Intelligence
Boyan Xu, Hujun Yin
Summary: This paper investigates the balance between the depth and complexity of convolutional neural networks, specifically in image restoration tasks. By increasing the network depth and decreasing the width of certain feature maps, performance can be improved while maintaining the same level of computational costs.
Article
Plant Sciences
Xueren Cao, Qun Zhang, Yongxiang He, Haiyan Che, Yating Lin, Daquan Luo, Jonathan S. West, Xiangming Xu
Summary: This study analyzed the genotypes of 112 isolates of Colletotrichum siamense and found significant genetic differentiation among three clusters, regardless of host, location, and year. This suggests that C. siamense infecting rubber tree, areca palm, and coffee in Hainan can be considered as one disease and should be controlled simultaneously.
Article
Microbiology
Kevin M. King, Gail Canning, Kang Zhou, Zekuan Liu, Mingde Wu, Jonathan S. West
Summary: This study indirectly investigates the potential role of sexual reproduction in the pathogenic fungus Plenodomus biglobosus, which causes blackleg disease in oilseed rape. The mating types of P. biglobosus in China are unbalanced, indicating predominantly asexual reproduction. These findings are important for improving the understanding of blackleg disease, particularly in China.
Article
Engineering, Electrical & Electronic
Songtao Xie, Junyan Hu, Zhengtao Ding, Farshad Arvin
Summary: Cooperative adaptive cruise control (CACC) is considered a potential solution for reducing traffic congestion, increasing road capacity, reducing fossil fuel consumption and improving traffic safety. This paper innovatively applies a spring damping energy model to construct a robust autonomous vehicle platoon system in order to overcome the negative influence of unreliable communication on CACC. Based on this model, a distributed control protocol utilizing only local information from neighbors is proposed, guaranteeing the connectivity and control input bounds of the vehicle platoon system. Finally, the proposed CCAC strategy is validated through multiple simulation experiments in Unreal Engine.
IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY
(2023)
Article
Mechanics
Anthony Man, Mohammad Jadidi, Amir Keshmiri, Hujun Yin, Yasser Mahmoudi
Summary: This paper introduces a novel zonal machine learning approach for Reynolds-averaged Navier-Stokes turbulence modeling based on the divide-and-conquer technique. The approach involves training and testing ML models in different flow physics regions called zones. The results show that the zonal models outperform the non-zonal models in predicting the Reynolds stress anisotropy and turbulent kinetic energy.
Article
Agronomy
Xueren Cao, Ying Xiao, Qiaohui Han, Jonathan S. West, Luo Daquan
Summary: Rubber tree powdery mildew, caused by Erysiphe quercicola, poses a significant threat to rubber tree plantations worldwide. This study investigated the population structure of E. quercicola at different stages of disease epidemic in Hainan, China. The results showed significant differences in population structure among the different epidemic stages, with late epidemic populations exhibiting higher genetic diversity. Analysis also revealed four distinct clusters of E. quercicola samples, suggesting the potential spread of the disease through inoculum drifting or host jumps.
Article
Agronomy
Samara Nunes Campos Vicentini, Nichola J. Hawkins, Kevin M. King, Silvino Intra Moreira, Adriano Augusto de Paiva Custodio, Rui Pereira Leite Junior, Diego Portalanza, Felipe Rafael Garces-Fiallos, Loane Dantas Krug, Jonathan S. West, Bart A. Fraaije, Waldir Cintra De Jesus Junior, Paulo Cezar Ceresini
Summary: We developed a new monitoring tool for managing wheat blast, which provides quantitative measurement of pathogen inoculum levels and detection of fungicide resistance alleles. The results showed significant variations in pathogen inoculum levels and widespread distribution of fungicide resistance alleles in aerosol populations. However, the impact of weather variables on pathogen dynamics was not significant.
Article
Agronomy
Joanna Kaczmarek, Jonathan S. West, Kevin M. King, Gail G. M. Canning, Akinwunmi O. Latunde-Dada, Yong-Ju Huang, Bruce D. L. Fitt, Malgorzata Jedryczka
Summary: The study used SYBR-Green qPCR technology to successfully detect the abundance of airborne ascospores. Different primer pairs showed variations in specificity and sensitivity.
PEST MANAGEMENT SCIENCE
(2023)
Proceedings Paper
Automation & Control Systems
Fatemeh Rekabi-Bana, Martin Stefanec, Jiri Ulrich, Erhan E. Keyvan, Tomas Roucek, George Broughton, Bilal Y. Gundeger, Omer Sahin, Ali E. Turgut, Erol Sahin, Tomas Krajnik, Thomas Schmickl, Farshad Arvin
Summary: This paper presents a robotic system that collaborates with social insects inside their hive. The robot consists of a micro- and macro-manipulator, as well as a tracking system. The micro-manipulator interacts with individual specimens using bio-mimetic agents, while the macro-manipulator positions and keeps the micro-manipulator's base around the individual. The system was verified in a honeybee observation hive, where it flawlessly tracked the honeybee queen and extracted behaviors of honeybee workers.
2023 IEEE INTERNATIONAL CONFERENCE ON MECHATRONICS, ICM
(2023)
Proceedings Paper
Computer Science, Interdisciplinary Applications
Jiri Ulrich, Ahmad Alsayed, Farshad Arvin, Tomas Krajnik
Summary: This paper proposes a new method for the full 6 degrees of freedom pose estimation of a circular fiducial marker. The method achieves real-time detection and accurate localization of the marker using a vision localization system. Experimental results show that the method achieves three times the accuracy of the original marker.
37TH ANNUAL ACM SYMPOSIUM ON APPLIED COMPUTING
(2022)
Proceedings Paper
Automation & Control Systems
Yifeng He, Barry Lennox, Farshad Arvin
Summary: This paper investigates the application of swarm robotics in monitoring nuclear waste storage facilities. By utilizing the active elastic sheet model and low-cost autonomous micro-surface robots, we implemented the task of exploring unknown areas. Although the accuracy of path planning was not very high, we demonstrated the feasibility of using a low-cost robotic platform for this new application.
TOWARDS AUTONOMOUS ROBOTIC SYSTEMS, TAROS 2022
(2022)
Article
Computer Science, Artificial Intelligence
Runqi Chai, Hanlin Niu, Joaquin Carrasco, Farshad Arvin, Hujun Yin, Barry Lennox
Summary: This article proposes a deep learning-based control framework for planning optimal maneuver trajectories and guiding mobile robots in uncertain environments. The framework consists of a motion planning layer and a waypoint tracking layer, using recurrent deep neural network algorithms and deep reinforcement learning algorithms. Experimental results demonstrate that the control framework performs well in achieving autonomous exploration missions.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2022)