检索结果-内蒙古大学图书馆

作者： Lai, Zhibing Chen, Lihui Jeon, Gwanggil Liu, Zitao Zhong, Rui Yang, Xiaomin College of Electronics and Information Engineering Sichuan University ChengduSichuan610064 China School of Electronic Engineering Xidian University Xi’an710071 China Department of Embedded Systems Engineering Incheon National University Incheon22012 Korea Republic of TAL Education Group Beijing China Institute of Nuclear Science and Technology Sichuan University Chengdu610064 China School of Aeronautics & Astronautics Sichuan University Chengdu610064 China

real-time monitoring and surveillance play an important role in the field of remote sensing, where multi-spectral (MS) images with high spatial resolution are widely desired for better analysis. However, high-resolution MS images cannot be directly obtained due to the limitations of sensors and bandwidth. As an essential way to alleviate this problem, pan-sharpening aims at fusing the complementary information of a low-resolution MS image and a high-resolution panchromatic (PAN) image to reconstruct a high-resolution MS image. Most previous deep-learning based methods can meet the real-time requirements with the help of graphics processing unit (GPU). However, they don’t fully exploit the favorable hierarchical information, sparing huge room for performance improvement. In this paper, to meet the requirement of real-time implementation and achieve more effective performance simultaneously, we propose a multi-scale fusion network (MSFN) to make full use of hierarchical complementary features of PAN and MS images. Specifically, we introduce an encoder–decoder structure and coarse-to-fine strategy to effectively extract multi-scale features of PAN and MS images, separately. Meanwhile, an information pool is adopted to preserve primitive information. Then a multi-scale feature fusion module is applied to fuse multi-scale features from the decoder and information pool. Finally, the fused features are utilized to reconstruct the high-resolution MS image. Extensive experiments demonstrate that our proposed method achieves favorable performance against other methods in terms of quantitative metrics and visual quality. Besides, the results on running time indicate that our method can achieve real-time performance. © 2021, The Author(s), under exclusive licence to Springer-Verlag GmbH, DE part of Springer Nature.

关键词： Graphics processing unit

来源：评论

学校读者我要写书评

暂无评论

Traffic prediction in SDN for explainable QoS using deep learning approach

引用

SCIENTIFIC REPORTS 2023年第1期13卷 1-15页

作者： Wassie, Getahun Ding, Jianguo Wondie, Yihenew Addis Ababa Univ IP Networking & Mobile Internet Addis Ababa Ethiopia Blekinge Inst Technol BTH Dept Comp Sci S-37179 Karlskrona Sweden Addis Ababa Univ Dept Elect & Comp Engn Addis Ababa Ethiopia

The radical increase of multimedia applications such as voice over Internet protocol (VOIP), image processing, and video-based applications require better quality of service (QoS). Therefore, traffic Predicting and explaining the prediction models is essential. However, elephant flows from those applications still needs to be improved to satisfy Internet users. Elephant flows lead to network congestion, resulting in packet loss, delay and inadequate QoS delivery. Recently, deep learning models become a good alternative for real-time traffic management. This research aims to design a traffic predicting model that can identify elephant flows to prevent network congestion in advance. Thus, we are motivated to develop elephant flow prediction models and explain those models explicitly for network administrators' use in the SDN network. H2O, deep Autoencoder, and autoML predicting algorithms, including XGBoost, GBM and GDF, were employed to develop the proposed model. The performance of Elephant flow prediction models scored 99.97%, 99.99%, and 100% in validation accuracy of under construction error of 0.0003952, 0.001697, and 0.00000408 using XGBoost, GBM, and GDF algorithms respectively. The models were also explicitly explained using Explainable Artificial Intelligence. Accordingly, packet size and byte size attributes need much attention to detect elephant flows.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Research on P300 Brain-Computer Interface Classification Algorithm Combining Channel Selection and deep learning 16

Research on P300 Brain-Computer Interface Classification Alg...

引用

16th ICME International Conference on Complex Medical Engineering, CME 2022

作者： Li, Qi Sun, Meiqi Song, Yu Sun, Jiajun Zhongshan Institute of Changchun University of Science and Technology School of Computer Science and Technology Changchun University of Science and Technology Zhongshan China Changchun University of Science and Technology School of Computer Science and Technology Changchun China

ISBN: (纸本)9781665496995

The P300-based brain-computer interface system can realize the control of external devices by decoding the P300 signal. Most of the current processing algorithms for P300 signals analyze the full channel. If the data contains bad channels or invalid electrodes with poor contact, then the feature extraction of the full channel is detrimental to the subsequent classification. Meanwhile, since different brain regions have different sensitivity to P300 signals, retaining channel information with better accuracy and eliminating poorer channels can reduce the data dimension and the number of parameters, thus reducing the complexity of the model and its execution time and improving the classification accuracy. Although we can perform manual bad channel rejection in advance, manual rejection is not applicable in online experiments such as real-time control of brain-computer interfaces, which requires an automatic channel selection method. In this paper, we use the Hilbert-Huang Transform to extract features in the time-frequency domain for a single channel, use RBFSVM for initial classification, select the top five channels for simple feature splicing based on their respective classification accuracy, and then input the spliced feature vectors into the deep learning neural network SENet to identify P300 signals. The results show that the accuracy of the method proposed in this paper is significantly improved compared with other comparative methods, with an average accuracy of 86.37%, which has significant research implications for P300 brain-computer interface control of external devices and other rehabilitation-related applications. © 2022 IEEE.

关键词： Brain computer interface

来源：评论

学校读者我要写书评

暂无评论

Designing and Searching for Lightweight Monocular Depth Network 28th

Designing and Searching for Lightweight Monocular Depth Netw...

引用

28th International Conference on Neural Information processing

作者： Liu, Jinfeng Kong, Lingtong Yang, Jie Shanghai Jiao Tong Univ Inst Image Proc & Pattern Recognit Dept Automat Shanghai Peoples R China

ISBN: (纸本)9783030922726;9783030922733

Depth sensing is extremely notable in some tasks of robot and autonomous driving. Nowadays, monocular depth estimation based on deep learning becomes a research focus in computer vision. However, most of the current work is seeking for more complex models to get higher accuracy, which can not achieve real-time inference on mobile or embedded systems. Therefore, we aim to design a lightweight model in this paper. At first, we improve the state-of-the-art model, FastDepth, producing FastDepthV2, which has higher accuracy and lower latency on the NYU Depth v2 dataset. Besides, since designing artificial networks takes time and effort, we make it automatic to design lightweight models in monocular depth estimation, using neural architecture search (NAS). Further, inspired by the architecture of MobileNetV2, a factorized hierarchical search space for encoder is used in this paper. In the meanwhile, we incorporate the accuracy and multiply-add operations of a model together into the searching objective, and use gradient-based reinforcement learning algorithm in the searching iterations. The controller in the reinforcement learning framework has converged after more than 1000 searching iterations, and three network architectures with the best performance are obtained. Under the same conditions of training and testing, two of them perform better than FastDepthV2.

关键词： deep learning Monocular depth estimation Lightweight network Neural architecture search

来源：评论

学校读者我要写书评

暂无评论

real-time Air Quality Index Detection through Regression-Based Convolutional Neural Network Model on Captured images

引用

ENVIRONMENTAL QUALITY MANAGEMENT 2024年第1期34卷 e22276-e22276页

作者： Sarkar, Pritisha Saha, Duranta Durbaar Vishal Saha, Mousumi Natl Inst Technol Durgapur Dept CSE Durgapur W Bengal India

Our research addresses the imperative need for an efficient air quality monitoring and forecasting system to mitigate the significant health risks of air pollution. Departing from conventional binary data collection methods, we employ image-based techniques to overcome inherent limitations. A pioneering aspect of our work involves the development of a novel model capable of predicting six air pollutants (PM2.5, PM10, O3, CO, SO2, NO2) along with Air Quality Index for Bengaluru, Delhi, and Tamil Nadu, achieving a commendable mean absolute error of 0.1432 on the test set. The efficacy of our approach is validated through a meticulously curated dataset comprising approximately 5455 images. We emphasize the significance of normalization by presenting outputs before and after, shedding light on the impact of parameters with varying ranges and strategies employed to mitigate such discrepancies. A detailed analysis of our model's best and worst outputs provides valuable insights into its strengths and limitations. To enhance user accessibility, we introduce an innovative image-based, real-time, user-friendly dashboard that allows users to conveniently assess a location's air pollution levels by uploading an image. This holistic approach offers a promising avenue for accurate air quality prediction and real-time monitoring.

关键词： air pollution air quality prediction deep learning image processing regression-based convolutional neural networks user interface

来源：评论

学校读者我要写书评

暂无评论

A Power Grid Equipment Fault Prediction Model Based on Faster RCNN and Video Streaming 8th

A Power Grid Equipment Fault Prediction Model Based on Faste...

引用

8th International Conference on Artificial Intelligence and Security (ICAIS)

作者： Feng, Youjun Luo, Jinman Wen, Zhaocong Yao, Zirui Guangdong Power Grid Corp Dongguan Power Supply Bur Dongguan 523000 Peoples R China

ISBN: (纸本)9783031067884;9783031067877

With the increasing proportion of power equipment in human life and production, the safe and stable operation of power equipment is related to the people's livelihood and national economy, and it is also very important to the sustainable development of the power industry. Grid equipment in long-term operation condition, the equipment of continuous monitoring and forecasting can guarantee the health of the whole grid system is running, the traditional way of artificial detection disadvantages such as low efficiency, high cost and difficult, gradually are replaced by the method of artificial intelligence. At present, power equipment detections mainly include two categories: image detection and video detection. Considering the real-time and image integrity of video stream detection, this article is based on the Faster RCNN model, using VGG16 network and ResNet50 network to detect video streams respectively, and comparing the efficiency and accuracy between them. According to the test results of the collected power grid video dataset, the performance of Resnet50 network is better than VGG16 network in terms of accuracy and recall rate, but VGG16 has more advantages in terms of processing speed.

关键词： deep learning Faster RCNN Feature extraction Bounding Box Regression Power grid equipment fault prediction Video streaming

来源：评论

学校读者我要写书评

暂无评论

Efficient and accurate object detection for 3D point clouds in intelligent visual internet of things

引用

MULtimeDIA TOOLS AND APPLICATIONS 2021年第20期80卷 31297-31334页

作者： Li, Hui Wang, Junyin Xu, Lingwei Zhang, Shujun Tao, Ye Qingdao Univ Sci & Technol Sch Informat Sci & Technol Qingdao 266061 Peoples R China South Cent Univ Nationalities Hubei Key Lab Intelligent Wireless Commun Wuhan 430074 Peoples R China Lanzhou Jiaotong Univ Minist Educ Key Lab Optotechnol & Intelligent Control Lanzhou 730070 Peoples R China

Visual perception is a key technology in the Intelligent Visual Internet of Things. The research of object detection methods is of great significance for improving the safety and efficiency of unmanned driving technology and intelligent visual Internet of Things. 3D point clouds object detection of deep learning can not only use the deep network to automatically learn characteristics of the multi-layer abstract structure, improve calculation efficiency and detection accuracy of the model, but also have better performance in dealing with object occlusion, absence and data sparsity with obtained high-dimensional point clouds information. However, the current review of object detection methods for 3D point clouds based on deep learning is scarce. In order to provide a more comprehensive understanding and understanding of the security and efficiency development of driverless technology, this paper is divided into the monocular camera, RGB-D image and LiDAR point cloud, according to the main data of the network model, and further subdivides according to the different use methods of the model. Analyze the performance of various model detection methods. This article also summarizes current commonly used 3D point clouds datasets of object detection, organizes and describes detection metrics of commonly used 3D point clouds, and discusses research challenges and development trends. The real-time performance of 3D point cloud object detection under the intelligent vision Internet of Things needs to be improved.

关键词： Internet of things Object detection 3D point clouds Monocular camera RGB-D image LiDAR

来源：评论

学校读者我要写书评

暂无评论

Multi-view Bistatic SAR Automatic Recognition Study based on Ground Small Target SAR Dataset

Multi-view Bistatic SAR Automatic Recognition Study based on...

引用

Signal, Information and Data processing (ICSIDP), IEEE International Conference on

作者： Jiming Lv Daiyin Zhu College of Electronics and Information Engineering Nanjing University of Aeronautics and Astronautics Nanjing China Faculty of Engineering Niigata University Niigata Japan Shenzhen Research Institute Nanjing University of Aeronautics and Astronautics Nanjing China

ISBN: (数字)9798331515669

ISBN: (纸本)9798331515676

MiniSAR systems possess unique advantages in ground target detection compared to traditional spaceborne and airborne SAR systems, mainly in terms of lower cost, real-time performance, and portability. Although numerous studies have focused on target recognition based on SAR systems with unmanned aerial vehicle, most are limited to monostatic SAR systems. Bistatic SAR systems offer stronger stealth capabilities compared to monostatic systems, as they can receive electromagnetic reflections from any angle. However, due to the lower intensity of received electromagnetic signals in bistatic systems, targets in SAR images tend to experience deformation and attenuation. In this study, based on independently developed monostatic and bistatic SAR systems, we study the integration of multi-view and large-scale deep learning models to aiming to address the issue of hard to recognized small targets. Additionally, the paper validates our proposed method on the publicly available MSTAR dataset and the self-developed FAST-Vehicle dataset. Experimental results demonstrate that the proposed multi-view target recognition method effectively identifies SAR bistatic deformed targets.

关键词： Training image recognition Accuracy Target recognition Imaging Object detection Transformers Radar polarimetry real-time systems Electromagnetics

来源：评论

学校读者我要写书评

暂无评论

Abnormal Recognition Technology of 3D Virtual Scene image Based on Wireless Network Sensor

Abnormal Recognition Technology of 3D Virtual Scene Image Ba...

引用

Ambient Intelligence, Knowledge Informatics and Industrial Electronics (AIKIIE), International Conference on

作者： Guangwei Li HEXI University Zhangye Gansu China

The performance and reliability requirements of wireless network sensors are getting higher and higher. In the complex 3D virtualization scene, abnormal recognition of sensor data becomes a challenging task. Therefore, this study aims to explore an effective technique for identifying anomalies in wireless network sensor images in 3D virtualized scenarios. In this study, deep learning techniques were used, combined with image processing and pattern recognition methods, to achieve accurate identification of anomalies in wireless network sensor images. Through training and testing on a large number of 3D virtualized scene images, the experimental results show that the proposed method has high accuracy in identifying anomalies in wireless network sensor images. The recall rate of the algorithm in this paper is 77%-85%, compared with the traditional method, the method in this paper can more accurately detect and classify various types of abnormal conditions, including signal interference, equipment failure and abnormal data. In addition, the method studied in this paper also shows good real-time performance, which can detect and identify anomalies in real time when processing large-scale sensor data.

关键词：

来源：评论

学校读者我要写书评

暂无评论

JOINT SUPER-RESOLUTION AND image RESTORATION FOR PLEIADES NEO imageRY 24

JOINT SUPER-RESOLUTION AND IMAGE RESTORATION FOR PLEIADES NE...

引用

24th ISPRS Congress on Imaging Today, Foreseeing Tomorrow

作者： Chouteau, F. Gabet, L. Fraisse, R. Bonfort, T. Harnoufi, B. Greiner, V Le Goff, M. Ortner, M. Paveau, V Airbus Def & Space Upstream Ground Syst 31 Rue Cosmonautes F-31400 Toulouse France Airbus Def & Space Geo Intelligence 15 Rue Satellites F-31030 Toulouse France

Modern Earth Observation optical satellite systems, such as Airbus's Pleiades Neo (PNeo) push the boundaries of high spatial resolution by providing commercial imagery products with up to 30cm ground sampling distance (GSD). To further enhance the quality of the images, the in-space imaging system is usually complemented by on-ground image restoration processing, such as deconvolution and denoising (Latry et al., 2012). Recent advances leverage Convolutional Neural Networks (CNNs) to improve the image restoration quality (K. Zhang et al., 2021a). Single image Super-Resolution (SISR), or Zoom, the process of obtaining a higher resolution (HR) image from a lower resolved (LR) source, has recently gained traction for both medium resolution sensors such as Sentinel 2 (Lanaras et al., 2018) and high resolution such as Pleiades and GeoEye-1 (Zhu et al., 2020). This process further enhances the resolution of the image to improve downstream applications such as mapping (L. Zhang et al., 2021) and small objects recognition (Shermeyer and Van Etten, 2019). While SISR for remote sensing has been successfully tackled using CNNs (Rohith and Kumar, 2021) the main challenge for reaching acceptable image quality performance lies in the generation of realistic LR/HR training pairs (K. Zhang et al., 2021b). In this paper, we propose: a dedicated simulation chain leveraging extremely-high-resolution (EHR) aerial imagery to generate realistic 30cm Pleiades Neo images and their corresponding fully restored HR equivalent at 15cm GSD A residual-based CNN architecture which we train to jointly restore and zoom the images All contributions are assessed on real PNEO images. We deployed the trained models in a production context, to enhance the full Pleiades Neo products - with a swath of 47k pixels - in an efficient and scalable manner.

关键词： Single-image Super-Resolution Pleiades Neo Remote Sensing deep learning Convolutional Neural Networks

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：