检索结果-内蒙古大学图书馆

Applied imagery pattern recognition for photovoltaic modules' inspection: A review on methods, challenges and future development

引用

SUSTAINABLE ENERGY TECHNOLOGIES AND ASSESSMENTS 2022年第PartA期52卷

作者： Yahya, Zefri Imane, Sebari Hicham, Hajji Ghassane, Aniba Safia, El Bouchini-Idrissi IAV Hassan II Sch Geomat & Surveying Engn Photogrammetry Cartog Dept Rabat 10112 Morocco Mohammed V Univ Rabat Mohammadia Sch Engineers Elect Engn Dept Rabat Morocco

We present a literature review of Applied imagery Pattern Recognition (AIPR) for the inspection of photovoltaic (PV) modules under the main used spectra: (1) true-color RGB, (2) long-wave infrared (LWIR), and (3) electroluminescence-based short-wave infrared (SWIR). Three sequentially linked building blocks underpin this work. The first overviews reference guidelines of image acquisition and the main detectable defect patterns under each spectrum. It also provides key insights regarding the implementation of Unmanned Aerial Vehicles (UAVs) to acquire imagery, especially from a photogrammetric perspective. The second block presents various image pre-processing steps used to prepare inspection-ready datasets. These comprise radiometric correction, segmentation and edge extraction, geometric correction and cell clipping. The third surveys defect detection and classification through digital image processing and machine/deep learning techniques. We elaborate an in-depth topic discussion that, in parallel: highlights the main related challenges, provides core guidelines for AIPR-based PV inspection workflows, and suggests key research avenues for future studies. This review synthesizes the recent advances of the body knowledge. It also constitutes an insightful reference for professionals and academics within the PV operations and maintenance field who are considering the possibilities that digital imagery can offer.

关键词： Photovoltaics imagery Digital image processing Machine learning Deep learning

来源：评论

学校读者我要写书评

暂无评论

Focus on Hard Categories and Hard Examples: Remote Sensing image Scene Classification via Expert Model and Hard Example Mining

引用

IEEE GEOSCIENCE AND REMOTE SENSING LETTERS 2022年 19卷 1页

作者： Xiong, Yunsheng Zhang, Peng Dou, Yong Xu, Kele Niu, Xin Natl Univ Def Technol Dept Comp Changsha 410005 Peoples R China Key Lab Parallel & Distributed Proc Changsha 410005 Peoples R China

Deep learning has seen dramatic improvements in remote-sensing image scene classification. However, hard categories and hard examples widely exist in the data sets, due to the intraclass diversity and interclass similarity. In this letter, we propose a novel framework to address these issues. Specifically, our method first trains a general model to obtain the confusion matrix and select the hard categories. Then a sampling strategy is proposed to restructure the training set and an expert model is trained to focus on the hard categories. Finally, the knowledge of the expert model is distilled into the student model through a novel loss function, which encourages the student model to predict the hard label provided by manual annotation. Thus, the model can match the soft label provided by the expert model and pay more attention to the hard examples simultaneously. With this method, the student model cannot only deal with hard categories but also hard examples existing in other easy categories. To empirically demonstrate the effectiveness of the proposed method, we comprehensively evaluate the method on three publicly available benchmark data sets, the obtained results show that the proposed method outperforms the existing baseline methods and achieves superior results on all three data sets.

关键词： Training Remote sensing Predictive models Data models Loss measurement Task analysis Feature extraction Expert model hard categories hard examples knowledge distillation (KD) remote sensing scene classification

来源：评论

学校读者我要写书评

暂无评论

Federated Learning for Coronary Artery Plaque Detection in Atherosclerosis Using IVUS Imaging: A Multi-Hospital Collaboration

arXiv

引用

arXiv 2024年

作者： Hsiao, Chiu-Han Chen, Kai Peng, Tsung-Yu Huang, Wei-Chieh Research Center for Information Technology Innovation Academia Sinica Taipei City Taiwan Department of Information Management National Taiwan University Taipei City Taiwan Division of Cardiology Department of Internal Medicine Taipei Veterans General Hospital School of Medicine National Yang Ming Chiao Tung University Taiwan Department of Biomedical Engineering National Taiwan University Taipei Taiwan

The traditional interpretation of Intravascular Ultrasound (IVUS) images during Percutaneous Coronary Intervention (PCI) is a time-consuming process that often lacks consistency due to its reliance on the expertise and subjective judgment of physicians. Additionally, the collection and integration of IVUS data across multiple hospital systems are significantly constrained by regulatory restrictions and privacy concerns, posing challenges to collaborative analysis and model development. To establish a consistent and efficient method for IVUS image segmentation and enhance plaque detection during PCI procedures, a parallel 2D U-Net model with a multi-stage segmentation architecture has been developed. The model leverages a federated learning algorithm, enabling collaborative data analysis across multiple healthcare institutions while maintaining data privacy. Plaque segmentation is achieved by identifying and subtracting the External Elastic Membrane (EEM) and lumen areas. To optimize processing efficiency, polar coordinates are generated from Cartesian coordinates during the preprocessing stage, ensuring precise segmentation and improving computational performance in distributed medical imaging workflows. The model demonstrates effective plaque identification, achieving a Dice Similarity Coefficient (DSC) of 0.706, and accurately detects circular boundaries in real-time. Through collaboration with domain experts, the IVUS segmentation system provides enhanced plaque burden interpretation via precise quantitative parameter measurements. Future improvements may include integrating advanced federated learning methods to further enhance model performance. This technology is highly adaptable and suitable for environments where sensitive and distributed data are prevalent. Expanding the number of participating institutions and datasets could further optimize outcomes, enabling hospitals to achieve greater success in medical imaging and intervention. © 2024, CC BY-NC-ND.

关键词： image segmentation

来源：评论

学校读者我要写书评

暂无评论

Single Node Acceleration of Generative Adversarial Networks using HPC for image Analytics 22

Single Node Acceleration of Generative Adversarial Networks ...

引用

Proceedings of the 2022 5th International Conference on Computational Intelligence and Intelligent Systems

作者： Aswathy Ravikumar Harini Sriraman School of Computer Science and Engineering Vellore Institute of Technology India

ISBN: (纸本)9781450397612

Generative Adversarial Networks (GAN) are approaches that are utilized for data augmentation, which facilitates the development of more accurate detection models for unusual or unbalanced datasets. Computer-assisted diagnostic methods may be made more reliable by using synthetic pictures generated by GAN. Generative adversarial networks are challenging to train because too unpredictable training dynamics may occur throughout the learning process, such as model collapse and vanishing gradients. For accurate and faster results the GAN network need to trained in parallel and distributed manner. We enhance the speed and precision of the Deep Convolutional Generative Adversarial Networks (DCGAN) architecture by using its parallelism and executing it on High-Performance Computing platforms. The effective analysis of a DCGAN in Graphic processing Unit and Tensor processing Unit platforms in which each layer execution pattern is analyzed. The bottleneck is identified for the GAN structure for each execution platforms. The Central processing Unit is capable of processing neural network models, but it requires a great deal of time to do it. Graphic processing Unit in contrast, side, are a hundred times quicker than CPUs for Neural Networks, however, they are prohibitively expensive compared to CPUs. Using the systolic array structure, TPU performs well on neural networks with high batch sizes but in GAN the shift between CPU and TPU is huge so it does not perform well.

关键词： Generative Adversarial Network High-Performance Computing Graphic processing Unit Tensor processing Unit

来源：评论

学校读者我要写书评

暂无评论

Scalable Deep Learning on distributed Infrastructures: Challenges, Techniques, and Tools

引用

ACM COMPUTING SURVEYS 2020年第1期53卷 3-3页

作者： Mayer, Ruben Jacobsen, Hans-Arno Tech Univ Munich Boltzmannstr 3 D-85748 Garching Germany

Deep Learning (DL) has had an immense success in the recent past, leading to state-of-the-art results in various domains, such as image recognition and natural language processing. One of the reasons for this success is the increasing size of DL models and the proliferation of vast amounts of training data being available. To keep on improving the performance of DL, increasing the scalability of DL systems is necessary. In this survey, we perform a broad and thorough investigation on challenges, techniques and tools for scalable DL on distributed infrastructures. This incorporates infrastructures for DL, methods for parallel DL training, multi-tenant resource scheduling, and the management of training and model data. Further, we analyze and compare 11 current open-source DL frameworks and tools and investigate which of the techniques are commonly implemented in practice. Finally, we highlight future research trends in DL systems that deserve further research.

关键词： Deep-learning systems

来源：评论

学校读者我要写书评

暂无评论

Transformer in UAV image-Based Weed Mapping

引用

REMOTE SENSING 2023年第21期15卷 5165-5165页

作者： Zhao, Jiangsan Berge, Therese With Geipel, Jakob Norwegian Inst Bioecon Res NIBIO Dept Agr Technol POB 115 NO-1431 As Norway Norwegian Inst Bioecon Res NIBIO Dept Invertebrate Pests & Weeds Forestry Agr & Hort POB 115 NO-1431 As Norway

Weeds affect crop yield and quality due to competition for resources. In order to reduce the risk of yield losses due to weeds, herbicides or non-chemical measures are applied. Weeds, especially creeping perennial species, are generally distributed in patches within arable fields. Hence, instead of applying control measures uniformly, precision weeding or site-specific weed management (SSWM) is highly recommended. Unmanned aerial vehicle (UAV) imaging is known for wide area coverage and flexible operation frequency, making it a potential solution to generate weed maps at a reasonable cost. Efficient weed mapping algorithms need to be developed together with UAV imagery to facilitate SSWM. Different machine learning (ML) approaches have been developed for image-based weed mapping, either classical ML models or the more up-to-date deep learning (DL) models taking full advantage of parallel computation on a GPU (graphics processing unit). Attention-based transformer DL models, which have seen a recent boom, are expected to overtake classical convolutional neural network (CNN) DL models. This inspired us to develop a transformer DL model for segmenting weeds, cereal crops, and 'other' in low-resolution RGB UAV imagery (about 33 mm ground sampling distance, g.s.d.) captured after the cereal crop had turned yellow. images were acquired during three years in 15 fields with three cereal species (Triticum aestivum, Hordeum vulgare, and Avena sativa) and various weed flora dominated by creeping perennials (mainly Cirsium arvense and Elymus repens). The performance of our transformer model, 1Dtransformer, was evaluated through comparison with a classical DL model, 1DCNN, and two classical ML methods, i.e., random forest (RF) and k-nearest neighbor (KNN). The transformer model showed the best performance with an overall accuracy of 98.694% on pixels set aside for validation. It also agreed best and relatively well with ground reference data on total weed coverage, R2 = 0.598. I

关键词： 1Dtransformer perennial weed mapping low-resolution UAV image

来源：评论

学校读者我要写书评

暂无评论

Efficient Volume Estimation for Dynamic Environments using Deep Learning on the Edge

Efficient Volume Estimation for Dynamic Environments using D...

引用

IEEE International Symposium on parallel and distributed processing Workshops and Phd Forum (IPDPSW)

作者： Chandan Kumar Yamini Mathur Ali Jannesari Department of Computer Science Iowa State University Ames IA USA

ISBN: (数字)9781665497473

ISBN: (纸本)9781665497480

The utility of edge devices has increased in volume estimation of uneven terrains. Existing techniques utilize several geo-tagged images of the landscape, captured in-flight by an edge device mounted over a UAV, to generate 3D models and perform volume estimation through manual boundary marking. These methods, although accurate, require significant time, human effort and are heavily dependent on GPS. We present an efficient deep learning framework that detects the object of interest and automatically determines the volume (independent of GPS) of the detected object on-the-fly. Our method employs a stereo camera for depth sensing of the object and overlays a unit mesh grid over the object's boundary to perform volume estimation. We explore the accuracy vs computational complexity trade-off on variations of our technique. Experiments indicate that our method reduces the time for volume estimation by several orders of magnitude in contrast to existing methods and is independent of GPS as well. Also, to the best of our knowledge, this is the first method that can perform volume analysis in a dynamic environment.

关键词： Deep learning Solid modeling Three-dimensional displays Error analysis Volume measurement image edge detection Heuristic algorithms

来源：评论

学校读者我要写书评

暂无评论

A Multistrategy Evolutionary Multiobjective Optimization Method for Hyperspectral Endmember Extraction

引用

IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING 2023年 61卷

作者： Ye, Chuanlong He, Fazhi Luo, Jinkun Tong, Lyuyang Gao, Xiaoxin Si, Tongzhen Fan, Linkun Wuhan Univ Sch Comp Sci Wuhan 430072 Peoples R China Univ Jinan Sch Informat Sci & Engn Jinan 250022 Peoples R China

Hyperspectral endmember extraction (HEE) is an essential part of remote-sensing image processing. There have been recent attempts to model the HEE as a multiobjective optimization problem and apply multiobjective evolutionary algorithms to solve the problem. However, because of the large HEE search space, it is difficult for the current algorithms to achieve exploration-exploitation balance, and they easily stall prematurely. To address these issues, this article proposes a multistrategy evolutionary multiobjective method based on roulette wheel selection and the genetic algorithm (RWS-GA) for endmember extraction. This method designs two parallel algorithms corresponding to global exploration and local exploitation. In the RWS-GA, an improved NSGA-ii method, adopting a novel method to sort individuals on the same front instead of the crowding distance, is proposed to divide individuals into superior and inferior subpopulations. Thereafter, different modified population update strategies are utilized for subpopulations based on characteristics. Pixels that appear more frequently in the population are considered to perform better to have a higher probability of forming an endmember set with other pixels. In addition, excellent individuals often exhibit a higher probability of including endmembers compared with inferior individuals. Considering the abovementioned opinions, roulette wheel selection is performed on the inferior subpopulation for global search. Meanwhile, the superior subpopulation is responsible for local search based on the genetic algorithm (GA). Furthermore, an offspring complement mechanism (OCM) is presented to prevent duplicate individuals from appearing in historical archives. Numerous comparative experiments show that the proposed method is superior to other endmember extraction methods in three real-world datasets.

关键词： Evolutionary algorithms hyperspectral endmember extraction (HEE) multiobjective optimization NSGA-ii

来源：评论

学校读者我要写书评

暂无评论

A Fast Adversarial Sample Detection Approach for Industrial Internet-of-Things Applications

A Fast Adversarial Sample Detection Approach for Industrial ...

引用

International Workshop on Quality of Service

作者： Shigeng Zhang Yudong Li Shuxin Chen Xuan Li Jian Zhang School of Computer Science and Engineering Central South University College of Computer Science and Electronic Engineering Hunan University The Science and Technology on Parallel and Distributed Processing Laboratory (PDL) China

Adversarial attacks reveal the inherent vulnerability of deep neural networks, which face serious security issues for their security. Among them, the attack against the Deep Neural Network (DNN) application used in the Industrial Internet of Things (iioT) is a key area in adversarial attacks. Adversarial examples generated by attackers by adding human-undetectable interference to legitimate examples may cause models to make wrong decision results, resulting in serious accidents. Many detection technologies have been proposed to mitigate the harm of adversarial examples to neural networks, among which the methods based on the difference of feature attribution between normal examples and adversarial examples show state-of-the-art detection performance, but they suffer from detection efficiency. In this work, we focus on improving the detection efficiency of the feature-attribution-based detection methods. We observe that there is still a significant difference in the feature attribution distribution of a normal image and an adversarial image even only some pixels in the image are processed, which can be verified by utilizing the Kolmogorov-Smirnov test. Based on this observation, we first adopt a variety of strategies to sample partial pixels in an image and then utilize the selected pixels to train a feature-attribution-based detector for detecting adversarial examples. Extensive experiments conducted on four datasets (MNIST, CIFAR-10, SVHN, CIFAR-100) against various attacks proved that the detection efficiency of the accelerated detection method is improved (for example, the average execution time was increased by 8.7 times on CIFAR-10) while the detection performance maintains state-of-the-art.

关键词：

来源：评论

学校读者我要写书评

暂无评论

FAST 3D VOLUMETRIC image RECONSTRUCTION FROM 2D MRI SLICES BY parallel processing

arXiv

引用

arXiv 2023年

作者： Ghoshal, Somoballi Goswami, Shremoyee Chakrabarti, Amlan Sur-Kolay, Susmita A. K. Choudhury School of Information Technology University of Calcutta India Advanced Computing and Microelectronics Unit Indian Statistical Institute India

Magnetic Resonance Imaging (MRI) is a technology for non-invasive imaging of anatomical features in detail. It can help in functional analysis of organs of a specimen but it is very costly. In this work, methods for (i) virtual three-dimensional (3D) reconstruction from a single sequence of two-dimensional (2D) slices of MR images of a human spine and brain along a single axis, and (ii) generation of missing inter-slice data are proposed. Our approach helps in preserving the edges, shape, size, as well as the internal tissue structures of the object being captured. The sequence of original 2D slices along a single axis is divided into smaller equal sub-parts which are then reconstructed using edge preserved kriging interpolation to predict the missing slice information. In order to speed up the process of interpolation, we have used multiprocessing by carrying out the initial interpolation on parallel cores. From the 3D matrix thus formed, shearlet transform is applied to estimate the edges considering the 2D blocks along the Z axis, and to minimize the blurring effect using a proposed mean-median logic. Finally, for visualization, the sub-matrices are merged into a final 3D matrix. Next, the newly formed 3D matrix is split up into voxels and marching cubes method is applied to get the approximate 3D image for viewing. To the best of our knowledge it is a first of its kind approach based on kriging interpolation and multiprocessing for 3D reconstruction from 2D slices, and approximately 98.89% accuracy is achieved with respect to similarity metrics for image comparison. The time required for reconstruction has also been reduced by approximately 70% with multiprocessing even for a large input data set compared to that with single core processing. © 2023, CC BY.

关键词： Magnetic resonance imaging

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：