Article
Computer Science, Artificial Intelligence
Lei Huang, Jie Qin, Yi Zhou, Fan Zhu, Li Liu, Ling Shao
Summary: This paper reviews and comments on the past, present, and future of normalization methods in the context of deep neural network (DNN) training. The authors provide a unified picture of the main motivation behind different approaches from the perspective of optimization and present a taxonomy for understanding the similarities and differences between them. They also discuss the current progress in understanding normalization methods and provide a comprehensive review of the applications of normalization for particular tasks.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Article
Food Science & Technology
Yaodi Li, Jianxin Xue, Kai Wang, Mingyue Zhang, Zezhen Li
Summary: A fresh-cut cauliflower surface defect detection and classification model based on a convolutional neural network with transfer learning was proposed. The experimental results showed that the model had better capability and stronger robustness in detecting surface defects of fresh-cut cauliflower.
Article
Multidisciplinary Sciences
Sankar Ganesh Sundaram, Saleh Abdullah Aloyuni, Raed Abdullah Alharbi, Tariq Alqahtani, Mohamed Yacin Sikkandar, Chidambaram Subbiah
Summary: The emergence of COVID19 as a pandemic has led to the need for automated solutions for screening and treatment management. This paper presents a novel deep transfer learning based framework for COVID19 detection and segmentation on chest X-ray images, achieving high accuracies in binary and multi-class classifications as well as infection segmentation.
ARABIAN JOURNAL FOR SCIENCE AND ENGINEERING
(2022)
Article
Automation & Control Systems
Wenbo Zhu, Birgit Braun, Leo H. Chiang, Jose A. Romagnoli
Summary: This study explores different transfer learning approaches and implementation details for building high-performance models under the constraint of limited available training data. Various transfer learning implementations are compared, and important technical details are discussed in the research.
CHEMOMETRICS AND INTELLIGENT LABORATORY SYSTEMS
(2021)
Article
Engineering, Multidisciplinary
Jiajun Ma, Maolin Liu, Songyu Hu, Jianzhong Fu, Gui Chen, Aixi Yang
Summary: In this study, a novel transitive transfer learning convolutional neural network (CNN) ensemble framework for classifying bearing surface defects is proposed. Only small-scale datasets are needed in this framework. A transfer path and transfer method selection strategy for transitive transfer learning is then proposed to enhance the feature extraction ability of the CNN models on the basis of multiple illuminations. Experimental results show that the proposed method has an accuracy rate of 97.51% and an average detection time of 155 ms, meeting the requirements of industrial online detection.
MEASUREMENT SCIENCE AND TECHNOLOGY
(2023)
Article
Automation & Control Systems
Vignesh Sampath, Inaki Maurtua, Juan Jose Aguilar Martin, Andoni Rivera, Jorge Molina, Aitor Gutierrez
Summary: This article presents a method for improving the generalization ability of surface defect identification tasks by exploiting auxiliary information beyond the primary labels. By jointly learning features of pixel-level segmentation masks, object-level bounding boxes, and global image-level classification labels, the proposed method significantly improves the performance of state-of-the-art models. Experimental results show an overall accuracy of 97.1%, a Dice score of 0.926, and a mean average precision of 0.762 on defect classification, segmentation, and detection tasks.
IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS
(2023)
Article
Computer Science, Artificial Intelligence
Debanjan Pathak, U. S. N. Raju
Summary: In this study, an improved GN-Inception-Darknet-53 model is used to extract features for CBIR, and group normalization layer is introduced to reduce dependency on batch size. The proposed method outperforms traditional and CNN methods on seven challenging image datasets.
INTERNATIONAL JOURNAL OF MULTIMEDIA INFORMATION RETRIEVAL
(2021)
Article
Materials Science, Multidisciplinary
Yichuan Shao, Shuo Fan, Haijing Sun, Zhenyu Tan, Ying Cai, Can Zhang, Le Zhang
Summary: Defect classification is crucial in steel surface defect detection. Traditional methods using convolutional neural networks (CNNs) improve accuracy by increasing network depth and parameter count. However, this approach overlooks the memory overhead and diminishing accuracy gains. To address these issues, a multi-scale lightweight neural network model (MM) is proposed, which uses a fusion encoding module and Gaussian difference pyramid. Experimental results show that MM network achieves 98.06% accuracy in defect classification, surpassing other networks in both parameter reduction and accuracy.
Article
Computer Science, Information Systems
Fanqi Meng, Wenying Cheng, Jingdong Wang
Summary: This study proposed a semi-supervised software defect prediction model based on a tri-training algorithm, which effectively combines feature normalization, oversampling techniques, and the Tri-training algorithm to address the difficulty in software defect prediction. Simulation experiments showed that the proposed method outperformed existing supervised and semi-supervised learning methods in terms of Precision, Recall, and F-Measure values.
KSII TRANSACTIONS ON INTERNET AND INFORMATION SYSTEMS
(2021)
Article
Engineering, Electrical & Electronic
Chaofan Zhou, Meiqin Liu, Senlin Zhang, Ping Wei, Badong Chen
Summary: In the real mobile screen production line, the yield of normal products is high, making it difficult to collect defects. Additionally, emerging defects are hard to obtain in large numbers in a short time. To address this few-shot problem, we propose a method that can classify known and novel defects with limited training samples. Our method outperforms other few-shot methods and achieves high accuracy in different test scenarios. The code is available for access.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2023)
Article
Optics
Jingtian Guan, Jingjing Fei, Wei Li, Xiaoke Jiang, Liwei Wu, Yakun Liu, Juntong Xi
Summary: Automated defect inspection for specular surfaces is challenging due to their reflection property. Deflectometry has been widely used in defect detection by capturing fringe patterns. Traditional methods require hand-crafted features, but this study proposes a deep-learning-based approach using a benchmark dataset named SpecularDefect9. By combining light intensity contrast map and captured fringe pattern, a fusion network accurately classifies different defects.
OPTICS AND LASERS IN ENGINEERING
(2023)
Article
Computer Science, Artificial Intelligence
Xuejin Hu, Jing Yang, Fengling Jiang, Amir Hussain, Kia Dashtipour, Mandar Gogate
Summary: This paper proposes a novel self-supervised steel surface defect detection model that achieves excellent results by learning better embedding feature representation on large amounts of unlabeled data. To address the issue of destroying spatial structures, convolutional feature maps are preserved to enhance representation capability. The Earth Mover's Distance (EMD) metric is employed to evaluate contrastive matching similarity and eliminate the effect of random augmentations. Experimental results show superior performance compared to existing approaches and further improvement on smaller datasets.
APPLIED SOFT COMPUTING
(2023)
Article
Engineering, Multidisciplinary
Chen Li, Haoxin Yan, Xiang Qian, Shidong Zhu, Peiyuang Zhu, Chengwei Liao, Haoyang Tian, Xiu Li, Xiaohao Wang, Xinghui Li
Summary: A domain adaptation YOLOv5 model, named DAYOLOv5, is proposed in this paper for automatic surface defect inspection. This model shows better generalization in real-world industrial applications and outperforms traditional methods in the field of magnetic tile surface defect detection.
Article
Engineering, Electrical & Electronic
Deepa Natarajan, Esakkirajan Sankaralingam, Keerthiveena Balraj, Selvakumar Karuppusamy
Summary: This paper introduces a two-stage deep learning framework UNet-SNet for glaucoma detection, achieving high accuracy by segmenting fundus images and training deep learning models for glaucoma detection.
INTERNATIONAL JOURNAL OF IMAGING SYSTEMS AND TECHNOLOGY
(2022)
Article
Computer Science, Artificial Intelligence
Chun Li, Yunyun Yang, Hui Liang, Boying Wu
Summary: COVID-19 is a challenging pandemic globally, where CT images play a crucial role in disease identification. To address the issue of limited data, transfer learning can be utilized to transfer knowledge from previous tasks to the task of COVID-19 recognition.
KNOWLEDGE-BASED SYSTEMS
(2021)
Article
Food Science & Technology
Jingjing Zhang, Shijin Zhang, Linqing Wang, Wenqiang Tan, Qing Li, Zhanyong Guo
Summary: In this study, four novel derivatives of chitosan oligosaccharide (COS) were synthesized by grafting pyridine-4-aldehyde Schiff bases onto chloracetyl chitosan oligosaccharide (CACS). The structural characterization was conducted using FTIR and NMR. The COS derivatives showed better antioxidant and antibacterial activities compared to COS and CACS, which may be attributed to the positive charge of these derivatives. Among them, derivative BPCACS with the highest degree of substitution exhibited the best bioactivity with high scavenging rate and inhibition diameter.
Article
Chemistry, Physical
Ye Xu, Fang Jin, Guiying Wu, Tao Wang, Jingchao Zhang, Yanjun Chen
Summary: The borosilicate zeolite ERB-1 was modified by loading titanium and changing its polarity to improve its catalytic activity in aqueous solutions. The modified Ti-ERB-1 showed significantly improved catalytic activity in the presence of H2O2 oxidant.
Article
Biochemistry & Molecular Biology
Meimei Li, Boliao Li, Qi Yang, Yanying Li, Junxiang Wu, Xiangli Xu
Summary: In this study, we identified 50 neuropeptides including neuropeptide Y (NPY) of Mythimna separata using transcriptome sequencing. The spatial and temporal expression profile of NPY indicated its important role in feeding regulation and energy metabolism. Knockdown of NPY significantly inhibited food uptake and body weight, delayed developmental duration, and altered energy storage in M. separata larvae.
INTERNATIONAL JOURNAL OF BIOLOGICAL MACROMOLECULES
(2023)
Article
Chemistry, Physical
Yuzhao Ouyang, Decai Zhu, Chengjun Zhu, Yingbo Zhang, Jiamei Liu, Xin Jia, Jie Yu, Xinfang Li, Min Yang, Xiaowei Gao
Summary: This study reports a new Ruddlesden-Popper (R-P) structure oxide, which is widely used as the electrode material in low temperature solid oxide fuel cells (LT-SOFCs) due to its high catalytic activity and excellent oxygen transport performance. However, there are few reports on its application as the electrolyte material in LT-SOFCs. The researchers prepared a P-N heterostructure by using R-P P-type semiconductor Sm1.2Sr0.8Ni0.6Fe0.4O4+5 (SSNF) oxide material as the electrolyte and N-type semiconductor Sm0.075Nd0.075Ce0.85O2-5 (SNDC) oxide material as the cathode. The 5SSNF-5SNDC composite electrolyte exhibited a high ionic conductivity of 0.201 S.cm-1 and a remarkable fuel cell power density of 1056 mW.cm-2 at 550℃. The results indicate that P-N heterojunctions constructed from oxide materials with highly catalytically active R-P structures exhibit excellent electrolyte performance. This work provides a new perspective for developing advanced electrolytes of LT-SOFCs.
INTERNATIONAL JOURNAL OF HYDROGEN ENERGY
(2023)
Article
Chemistry, Physical
Jian Zhao, Chao Li, Xiaodong Fan, Hao Liu, Ziqi Liu, Jingchao Zhang, Zhizhong Sun, Wenyi Chu
Summary: A series of Bronsted-Lewis acidic tetraimidazolyl ionic liquids ([TeimILs]) were designed and synthesized, and [Teim-PS] [AlCl4]4 showed good catalytic performance in the conversion of glucose to 5-HMF with a biphasic system. The catalyst could be reused multiple times without significant decrease in yield. The biphasic system facilitated catalyst recycling and separation of 5-HMF. The catalytic system also exhibited applicability to the degradation of other carbohydrate compounds, and had a low activation energy of 49.98 KJ/mol. This work provides a new solution for the efficient and green conversion of glucose to 5-HMF.
APPLIED CATALYSIS A-GENERAL
(2023)
Article
Thermodynamics
Ning Qian, Fan Jiang, Marco Marengo, Jiajia Chen, Yucan Fu, Jingzhou Zhang, Jiuhua Xu
Summary: This study investigates the start-up behavior of oscillating heat pipes (OHP) under axial-rotation and highlights the importance of start-up speed on grinding efficiency and quality. It is found that OHP filled with acetone has the shortest start-up time, and the cold air pressure and centrifugal acceleration significantly affect the start-up time.
APPLIED THERMAL ENGINEERING
(2023)
Review
Materials Science, Multidisciplinary
Jingjing Zhang, Bing Zhang, Xiubo Xie, Cui Ni, Chuanxin Hou, Xueqin Sun, Xiaoyang Yang, Yuping Zhang, Hideo Kimura, Wei Du
Summary: This paper reviews recent advances in the nanoconfinement of Mg-based hydrogen storage materials by loading Mg particles on different supporting materials and discusses the prospects for designing high-performance Mg-based materials using nanoconfinement.
INTERNATIONAL JOURNAL OF MINERALS METALLURGY AND MATERIALS
(2023)
Article
Thermodynamics
Yuan-wei Lyu, Yun-duo Zhao, Jing-zhou Zhang, Jing-yang Zhang, Yong Shan
Summary: In this study, impinging heat transfer by a synthetic jet induced by a planar lobed orifice on a flat plate and a semi-cylindrical concave plate is investigated by means of Large Eddy Simulation. The results show that the temperature variation is primarily affected by the restriction of the target plate and the recirculation zone of the concave plate.
INTERNATIONAL JOURNAL OF THERMAL SCIENCES
(2023)
Article
Computer Science, Artificial Intelligence
Zewei He, Du Chen, Yanpeng Cao, Jiangxin Yang, Yanlong Cao, Xin Li, Siliang Tang, Yueting Zhuang, Zhe-ming Lu
Summary: This study proposes an efficient model, SISR-PF-OA, for single image super-resolution, achieving more accurate and faster image restoration by improving feature extraction and fusion techniques. It includes a novel Orientation-Aware feature extraction/selection Module (OAM) and an effective fusion architecture to progressively integrate multi-scale features extracted in different convolutional stages.
PATTERN RECOGNITION
(2023)
Article
Chemistry, Medicinal
Si-Wei Wang, Shu-Yu Xu, Tian Gan, Xiao-Bin Zhang, Jia-Hong Li, Xing Wang, He-Zhong Jiang
Summary: This study analyzed the chemical composition of Lindera fragrans leaf ethanol extract using UPLC-Q-Orbitrap HRMS technology. The extract was found to contain approximately 22 compounds and showed significant antidepressant activity in mice with chronic unpredictable mild stress-induced depression, improving behavioral indexes and regulating levels of CORT, 5-HT, and NE.
PHARMACEUTICAL CHEMISTRY JOURNAL
(2023)
Article
Chemistry, Multidisciplinary
Yi Ren, Jingjing Zhang, Xinxin Gao, Xin Zheng, Le Peng Zhang, Tie Jun Cui
Summary: In this paper, a general method for impedance matching in the design of miniaturized spoof plasmonic antennas based on SSPPs is proposed. A prototype of a planar spoof plasmonic dipole antenna is simulated, fabricated, and measured to verify the effectiveness of the method. The results show significant improvements compared to traditional dipole antennas.
Article
Chemistry, Multidisciplinary
Jinchai Xu, Fangfang Qu, Bihe Shen, Zhenxiong Huang, Xiaoli Li, Haiyong Weng, Dapeng Ye, Renye Wu
Summary: In this study, a portable rapid non-destructive detection device integrating visible/short-wave and long-wave near-infrared spectroscopy was developed to detect tea polyphenol content in fresh tea leaves. The device achieved better prediction performance by fusing spectral data and extracted sensitive spectral wavebands for tea polyphenols. This device provides effective technical support for tea breeding and quality control.
APPLIED SCIENCES-BASEL
(2023)
Article
Materials Science, Characterization & Testing
Yafei Dong, Bowen Zhao, Jiangxin Yang, Yanlong Cao, Yanpeng Cao
Summary: This paper analyzes the characteristics of sensor noise and background interference in lock-in thermography experiments and proposes a two-stage CNN model for removing them. Experimental results demonstrate that the proposed method outperforms existing techniques on specimens made of different materials and at various excitation frequencies. The proposed model provides high-quality input of infrared images for subsurface defect detection in lock-in thermography.
NDT & E INTERNATIONAL
(2023)
Article
Cardiac & Cardiovascular Systems
Shaohua Li, Junwen Zhang, Jingwei Ni, Jiumei Cao
Summary: This study aimed to identify specific hypoxia-related genes associated with unstable angina (UA) and myocardial infarction (MI), and to develop a predictive model for the progression from UA to MI. It was found that the 10 hypoxia-related genes identified were primarily associated with B cells and some inflammatory cells, and the CSTF2F gene was determined to be associated with atherosclerosis.
FRONTIERS IN CARDIOVASCULAR MEDICINE
(2023)
Proceedings Paper
Engineering, Electrical & Electronic
Junwen Zhang, Sizhe Xing, Guoqiang Li, Nan Chi
Summary: One of the key challenges in Coherent PON is achieving upstream coherent burst-mode reception. This article reviews the challenges, key enabling technologies, and recent progress in high-performance and robust burst reception in Coherent PON.
2023 OPTICAL FIBER COMMUNICATIONS CONFERENCE AND EXHIBITION, OFC
(2023)
Article
Optics
Yin Xiao, Lina Zhou, Wen Chen
Summary: This paper introduces a correspondence imaging approach for reconstructing high-quality objects through complex scattering media. By deriving a rectified theory and introducing temporal correction, the proposed method eliminates the effect of dynamic scaling factors. Experimental results demonstrate the advantages of the proposed method over conventional methods in complex scattering environments, and it can also be combined with other methods to further enhance the quality of reconstructed objects.
OPTICS AND LASERS IN ENGINEERING
(2024)
Article
Optics
Zengxuan Jiang, Minghao Chao, Qingsong Liu, Bo Cheng, Guofeng Song, Jietao Liu
Summary: In this paper, a multi-focal metalens with high focusing efficiency controlled by circular polarization multiplexing is demonstrated. The metalens can generate four transversely distributed focal points under normal incidence of linearly polarized light, supporting both left-circularly polarized and right-circularly polarized conversion. Furthermore, an oblique incidence metalens is designed to achieve high total focusing efficiency for terahertz waves and provides potential new applications for polarization imaging and detection.
OPTICS AND LASERS IN ENGINEERING
(2024)
Article
Optics
Yiran Wang, Yu Ji, Xuyang Zhou, Xiu Wen, Yutong Li, Zhengjun Liu, Shutian Liu
Summary: This work presents a new reconstruction framework for structured illumination microscopy (SIM), which only requires four raw images and avoids extensive iterative computation. By using checkerboard pattern illumination modulation instead of sinusoidal fringe illumination, the proposed method significantly reduces image acquisition time and achieves higher image reconstruction rate. Additionally, the reconstruction process is non-iterative and not limited by the field of view size.
OPTICS AND LASERS IN ENGINEERING
(2024)
Article
Optics
Qian He, Li Pei, Jianshuai Wang, Jingjing Zheng, Tigang Ning, Jing Li
Summary: This paper proposes a 3D refractive index profile visualization method to demonstrate mode activation and evolution in fiber fusion splicing. The method is validated through experimental results and provides support for various fiber splicing operations and mode coupling modulation.
OPTICS AND LASERS IN ENGINEERING
(2024)
Article
Optics
Qiwei Li, Qiyu Wang, Fang Lu, Yang Cao, Xu Zhao
Summary: LSHIP is a lenslet-array-based snapshot hyperspectral imaging polarimeter that combines spectral polarization modulation with integral field imaging spectrometry. It can simultaneously acquire three-dimensional spatial and spectral data-cubes for linear Stokes parameters in a single snapshot.
OPTICS AND LASERS IN ENGINEERING
(2024)
Article
Optics
Huicong Li, Bing Lv, Meng Tian, Wenzhu Huang, Wentao Zhang
Summary: This study proposes a temperature compensation scheme for unbalanced interferometers using sensing fibers with different temperature coefficients, aiming to resolve the temperature disturbance and achieve high strain resolution. The experimental results confirm the effectiveness of the proposed scheme in high-resolution, long-term, low-frequency, and static strain sensing.
OPTICS AND LASERS IN ENGINEERING
(2024)
Article
Optics
Hongxiang Chang, Rongtao Su, Yuqiu Zhang, Bowang Shu, Jinhu Long, Jinyong Leng, Pu Zhou
Summary: High-speed variable-focus optics provides new opportunities for fiber laser applications in various fields. This paper investigates a non-mechanical axial focus tuning method using coherent beam combining (CBC) technique and proposes a tilt modulation assisted method to extend the tuning range.
OPTICS AND LASERS IN ENGINEERING
(2024)
Article
Optics
Yubo Ni, Shuai Fu, Chaoyang Su, Zhaozong Meng, Nan Gao, Zonghua Zhang
Summary: This paper proposes a surface adaptive fringe pattern generation method to accurately measure specular surfaces, eliminating the out-of-focus effect and improving measurement accuracy and reliability.
OPTICS AND LASERS IN ENGINEERING
(2024)
Article
Optics
Zifan Wang, Tianfeng Zhou, Qian Yu, Zihao Zeng, Xibin Wang, Junjian Hu, Jiyong Zeng
Summary: Fast-axis collimation (FAC) lens arrays are crucial in laser systems, and their precision can be improved through the development of an optical collimation system and the use of thermal compensation to correct for non-uniform thermal expansion.
OPTICS AND LASERS IN ENGINEERING
(2024)
Article
Optics
Jincheng Chen, Qiuyu Fang, Li Huang, Xin Ye, Luhong Jin, Heng Zhang, Yinqian Luo, Min Zhu, Luhao Zhang, Baohua Ji, Xiang Tian, Yingke Xu
Summary: This study developed a novel deep learning accelerated SRRF method that enables super-resolution reconstruction with only 5 low SNR images, and allows real-time visualization of microtubule dynamics and interactions with CCPs.
OPTICS AND LASERS IN ENGINEERING
(2024)
Article
Optics
Pan Liu, Yongqiang Zhao, Ning Li, Kai Feng, Seong G. Kong, Chaolong Tang
Summary: This article presents a technique for inverse design of multilayer deep-etched gratings (MDEG) using a deep neural network with adaptive solution space. The proposed method trains a deep neural network to predict the probability distribution across the discretized space, enabling evaluation of an optimal solution. The results show improved efficiencies using only a reduced dataset and avoiding one-to-many mapping challenges.
OPTICS AND LASERS IN ENGINEERING
(2024)
Article
Optics
Evelina Bibikova, Nazar Al-wassiti, Nataliya Kundikova
Summary: Light beams possess three types of angular momentum, namely spin angular momentum, extrinsic orbital angular momentum, and intrinsic orbital angular momentum. The interaction between these momenta leads to the spin-orbit interaction of light and topological effects. This study predicts a new topological effect resulting from the influence of extrinsic orbital angular momentum on spin angular momentum in converging asymmetrical light beams. It manifests as the transformation of linear polarized light into elliptically polarized light when an asymmetrical beam passes through the left or right half of the focal plane. The measured value of the topological circular amplitude anisotropy R was found to be R = +/- (0.60 +/- 0.08) x 10(-3). This new effect contributes to our understanding of light and has potential applications in developing sensors in optics.
OPTICS AND LASERS IN ENGINEERING
(2024)
Article
Optics
Hamdy H. Wahba
Summary: This study combines multiple-beam Fizeau interference and single-shot digital holographic interferometry to study thick phase objects. By collecting optical phase at different focal planes, the angular spectrum method is used for the first time to retrieve optical phase maps through the focal depth. The proposed method proves to be effective in providing accurate numerical focusing and phase maps reconstruction.
OPTICS AND LASERS IN ENGINEERING
(2024)
Article
Optics
Mohammed A. Isa, Richard Leach, David Branson, Samanta Piano
Summary: Due to the complexity of resolving object form and pose in images, new vision algorithms prioritize identification and perception over accurate coordinate measurement. However, the use of planar targets for coordinate measurement in vision systems has several drawbacks, including calibration difficulties and limited viewing angles. On the other hand, the use of sphere targets is infrequent in vision-based coordinate metrology due to the lack of efficient multi-view vision algorithms for accurate sphere measurements.
OPTICS AND LASERS IN ENGINEERING
(2024)
Article
Optics
Ildar Rakhmatulin, Donald Risbridger, Richard M. Carter, M. J. Daniel Esser, Mustafa Suphi Erden
Summary: This paper reviews the application of machine learning in laser systems. While machine learning has been widely used in general control automation and adjustment tasks, its application in specific tasks requiring skilled workforces for high-precision equipment assembly and adjustment is still limited. The paper presents promising research directions for using machine learning in mirror positional adjustment, triangulation, and optimal laser parameter selection, based on the recommendations of PRISMA.
OPTICS AND LASERS IN ENGINEERING
(2024)