检索结果-内蒙古大学图书馆

Adaptive Feature Aggregation Centric Enhance Network for Accurate and Fast Monocular 3-D Object Detection

IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT 2024年 73卷 1页

作者： Lin, Peng-Wei Hsu, Chih-Ming Natl Taipei Univ Technol Inst Mech & Elect Engn Taipei 10608 Taiwan Natl Taipei Univ Technol Dept Mech Engn Taipei 10608 Taiwan

Three-dimensional object detection is crucial in autonomous driving. Monocular 3-D object detection has become a popular area of research in autonomous driving because of its ease of deployment and cost-effectiveness. In real-world autonomous driving, detectors should be both real time and accurate. These features can be achieved using deep learning (DL). A one-stage center-based object detector is suitable for real-world applications. However, in center-based object detectors, object-centric estimation plays an important role because it significantly influences detection results. To address this issue, we propose a real-time monocular 3-D object detection neural network called the adaptive feature aggregate centric enhance network. The proposed model is an anchor-free and center-based method. To improve accuracy while maintaining inference speed, we propose an adaptive feature aggregation network that aggregates multiscale features by weighting. Furthermore, we propose a centric enhanced module for heatmap prediction to improve the accuracy of object localization and classification. Our model can achieve 34.48 fps using an Nvidia RTX3070 graphy processing unit (GPU). Extensive experiments on the KITTI benchmark demonstrate that our method achieves good average precision (AP) for cars and pedestrians.

关键词： Three-dimensional displays Heating systems Feature extraction Accuracy Detectors Head Location awareness Convolution Autonomous vehicles Aggregates 3-D object detection attention mechanism autonomous driving deep learning (DL) monocular image real time

来源：评论

学校读者我要写书评

暂无评论

Enhancing passive gamma emission tomography data with deep learning

引用

ANNALS OF NUCLEAR ENERGY 2024年 204卷

作者： Sanchez-Belenguer, Carlos Casado-Coscolla, Alvaro Wolfart, Erik Joint Res Ctr European Commiss Via Enrico Fermi 2749 I-21027 Ispra VA Italy Seidor Italy SRL Contract European Commiss Ispra Italy

In this paper, we address the problem of generating and enhancing Passive Gamma Emission Tomography (PGET) data from a deep learning perspective. The PGET instrument has been developed for the verification of spent nuclear fuel and relies on image reconstruction and analysis algorithms to detect missing or substituted fuel pins. Such techniques are sensitive to the quality of the input data: noisy or incomplete sinograms yield to poor reconstructions and, consequently, to low-confidence results. The development and validation of these algorithms is based on complex Monte Carlo simulations that are time-consuming and computationally-demanding. We propose the use of Convolutional Neural Networks (CNNs) for enhancing PGET data. Our technique learns the mapping between incomplete or noisy sinograms and their corresponding full representation. It effectively exploits the high degree of redundancy of the measurements, i.e . the contribution of a single pin can be observed from many different directions, to learn the underlying model of the data and to make informed predictions. The two main applications of our approach are: (1) accelerating Monte Carlo simulations and (2) pre-processing real measurements to enhance them before running the standard image reconstruction and analysis techniques. The experimental evaluation was performed with both, simulated and real measurements. Results show how effectively CNNs can learn and exploit the structure of the data. For the two use cases evaluated, denoising sinograms and inpainting incomplete ones, our technique achieved state-of-the-art performance with execution times in the order of milliseconds.

关键词： Passive gamma emission tomography deep learning Monte Carlo simulations

来源：评论

学校读者我要写书评

暂无评论

An Efficient and Fast image Mosaic Approach for Highway Panoramic UAV images

引用

IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING 2024年 17卷 10454-10467页

作者： Zheng, Haoxin Chang, Zhanqiang Li, Yakai Zhu, Jie Wang, Wei Yang, Qing Xie, Chou Zhang, Jingfa Liu, Jiaxi Capital Normal Univ Coll Resource Environm & Tourism Beijing 100048 Peoples R China China Earthquake Networks Ctr Beijing 100045 Peoples R China Henan Inst Remote Sensing & Surveying Zhengzhou 450000 Peoples R China China Siwei Surveying & Mapping Technol Co Ltd Beijing 100086 Peoples R China Chinese Acad Sci Aerosp Informat Res Inst Beijing 100101 Peoples R China Natl Inst Nat Hazards Beijing 100085 Peoples R China

Nowadays, real-time monitoring of highway operation by unmanned aerial vehicle (UAV) technology is one of the research frontiers for urban remote sensing. In general, the existing stitching algorithms can meet the basic requirements in terms of accuracy, but their splicing speed cannot meet the real-time stitching requirements of UAV. The cause is that the time consumption sharply increases when stitching plenty of UAV images-this is the bottleneck problem. Herein, we proposed a novel splicing method based on the Superpoint network and a self-designed algorithm of matrix iteration. In this method, we take advantage of an advanced deep learning algorithm-Superpoint to efficiently extract image feature points for calculating the geometric transformation matrix, and make the Superpoint model more suitable for highway. More importantly, for the purpose of further improving the stitching speed and realizing real-time stitching for a large number of UAV images, we specially designed an algorithm of matrix iteration to accurately represent the image transformation relationships, i.e., a matrix is iterated through each adjacent transformation matrix relationship. It is the first time that an algorithm of transformation matrix iteration has been designed to address the bottleneck problem in stitching plenty of UAV images. As a result, the experiments indicate that the proposed method has remarkably enhanced the stitching speed and accuracy for plenty of UAV images. Notably, even in the condition of no air triangulation parameters, it can realize real-time stitching.

关键词： Feature extraction Autonomous aerial vehicles real-time systems Road transportation Remote sensing deep learning image stitching image mosaic matrix iteration real-time stitching unmanned aerial vehicle (UAV)

来源：评论

学校读者我要写书评

暂无评论

Convergence of deep learning and Edge Computing using Model Optimization 13

Convergence of Deep Learning and Edge Computing using Model ...

引用

13th Iranian/3rd International Machine Vision and image processing Conference (MVIP)

作者： Babaei, Peyman Islamic Azad Univ Comp Dept West Tehran Branch Tehran Iran

ISBN: (纸本)9798350350494;9798350350500

Edge systems are undergoing a groundbreaking computing evolution to support artificial intelligence, deep learning, and complex computational algorithms. Using cloud servers to perform deep learning model inference poses challenges such as response delays, increased communication costs, and data privacy concerns. Therefore, significant efforts have been made to push the processing of deep learning models to edge systems, which has led to the creation of edge intelligence as the intersection of learning and edge computing. learning models, especially deep convolutional neural networks, have made significant achievements in machine vision, which provide high accuracy and predictability by spending computing power and memory. If these models are optimized and deployed on edge systems, there will be a revolution in the applications of edge systems in real time. In this paper, by using optimization techniques such as quantization, weight pruning, and weight clustering, the possibility of deploying a typical convolutional neural network model on edge systems that have limited computing resources and memory is investigated. The results show that by using a collaborative algorithm, despite the slight decrease in the accuracy of the model, it is possible to achieve a small-sized model that can even be deployed on microcontrollers.

关键词： Model optimization Quantization Weight pruning Weight clustering Edge computing deep learning

来源：评论

学校读者我要写书评

暂无评论

Temporal signed gestures segmentation in an image sequence using deep reinforcement learning

引用

ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE 2024年 131卷

作者： Kalandyk, Dawid Kapuscinski, Tomasz Rzeszow Univ Technol Doctoral Sch Al Powstancow Warszawy 12 PL-35959 Rzeszow Poland Rzeszow Univ Technol Dept Comp & Control Engn Wincentego Pola 2 PL-35959 Rzeszow Poland

Continuous sign language recognition is challenging due to coarticulatory distortions, which occur at the beginning and end of each gesture. These distortions depend on the temporal context and introduce additional intraclass variability. To address this issue, a new approach is proposed that extracts segments from the image sequence corresponding to undistorted parts of gestures. This should simplify the task by reducing it to the easier problem of isolated gestures recognition. The proposed approach uses deep reinforcement learning for segmentation and a novel image sequence processing scheme to extract gradient changes over time. A dataset recorded by deaf people and annotated according to the proposed approach, was prepared to evaluate the method. The proposed deep learning architectures achieved leave -one -subject -out recognition accuracies in the range of 0.70 to 0.76. Considering the inability to compare with other works, the authors also proposed other evaluation protocols to thoroughly examine the employed approach. This work will be developed, and the main aspiration of the authors will be to create an integrated framework that converts the raw form of RGB video into a string of words representing the Sign Language user's intentions.

关键词： deep reinforcement learning image sequence segmentation Gesture spotting Gesture database

来源：评论

学校读者我要写书评

暂无评论

Multi-Domain Joint Synthetic Aperture Radar Ship Detection Method Integrating Complex Information with deep learning

引用

REMOTE SENSING 2024年第19期16卷 3555-3555页

作者： Tian, Chaoyang Lv, Zongsen Xue, Fengli Wu, Xiayi Liu, Dacheng Chinese Acad Sci Aerosp Informat Res Inst Dept Space Microwave Remote Sensing Syst Beijing 100190 Peoples R China Univ Chinese Acad Sci Sch Elect Elect & Commun Engn Beijing 100049 Peoples R China

With the flourishing development of deep learning, synthetic aperture radar (SAR) ship detection based on this method has been widely applied across various domains. However, most deep-learning-based detection methods currently only use the amplitude information from SAR images. In fact, phase information and time-frequency features can also play a role in ship detection. Additionally, the background noise and the small size of ships also pose challenges to detection. Finally, satellite-based detection requires the model to be lightweight and capable of real-time processing. To address these difficulties, we propose a multi-domain joint SAR ship detection method that integrates complex information with deep learning. Based on the imaging mechanism of line-by-line scanning, we can first confirm the presence of ships within echo returns in the eigen-subspace domain, which can reduce detection time. Benefiting from the complex information of single-look complex (SLC) SAR images, we transform the echo returns containing ships into the time-frequency domain. In the time-frequency domain, ships exhibit distinctive features that are different from noise, without the limitation of size, which is highly advantageous for detection. Therefore, we constructed a time-frequency SAR image dataset (TFSID) using the images in the time-frequency domain, and utilizing the advantages of this dataset, we combined space-to-depth convolution (SPDConv) and Inception depthwise convolution (InceptionDWConv) to propose Efficient SPD-InceptionDWConv (ESIDConv). Using this module as the core, we proposed a lightweight SAR ship detector (LSDet) based on YOLOv5n. The detector achieves a detection accuracy of 99.5 with only 0.3 M parameters and 1.2 G operations on the dataset. Extensive experiments on different datasets demonstrated the superiority and effectiveness of our proposed method.

关键词： synthetic aperture radar (SAR) single-look complex (SLC) complex information deep learning ship detection

来源：评论

学校读者我要写书评

暂无评论

A Machine learning-Based Tropospheric Prediction Approach for High-Precision real-time GNSS Positioning

引用

SENSORS 2024年第10期24卷 2957页

作者： Chen, Jianping Gao, Yang Univ Calgary Dept Geomat Engn Calgary AB T2N 1N4 Canada

For high-precision positioning applications, various GNSS errors need to be mitigated, including the tropospheric error, which remains a significant error source as it can reach up to a few meters. Although some commercial GNSS correction data providers, such as the Quasi-Zenith Satellite System (QZSS) Centimeter Level Augmentation Service (CLAS), have developed real-time precise regional troposphere products, the service is available only in limited regional areas. The International GNSS Service (IGS) has provided precise troposphere correction data in TRO format post-mission, but its long latency of 1 to 2 weeks makes it unable to support real-time applications. In this work, a real-time troposphere prediction method based on the IGS post-processing products was developed using machine learning techniques to eliminate the long latency problem. The test results from tropospheric predictions over a year using the proposed method indicate that the new method can achieve a prediction accuracy (RMSE) of 2 cm, making it suitable for real-time applications.

关键词： LSTM FFNN troposphere prediction deep learning machine learning neural network

来源：评论

学校读者我要写书评

暂无评论

HyperSLICE: HyperBand optimized spiral for low-latency interactive cardiac examination

引用

MAGNETIC RESONANCE IN MEDICINE 2024年第1期91卷 266-279页

作者： Jaubert, Olivier Montalt-Tordera, Javier Knight, Daniel Arridge, Simon Steeden, Jennifer Muthurangu, Vivek UCL UCL Ctr Translat Cardiovasc Imaging London England Royal Free London NHS Fdn Trust Dept Cardiol London England UCL Dept Comp Sci London England UCL Ctr Translat Cardiovasc Imaging 30 Guilford St London WC1N 1EH England

Purpose: Interactive cardiac MRI is used for fast scan planning and MR-guided interventions. However, the requirement for real-time acquisition and near-real-time visualization constrains the achievable spatio-temporal resolution. This study aims to improve interactive imaging resolution through optimization of undersampled spiral sampling and leveraging of deep learning for low-latency reconstruction (deep artifact suppression).Methods: A variable density spiral trajectory was parametrized and optimized via HyperBand to provide the best candidate trajectory for rapid deep artifact suppression. Training data consisted of 692 breath-held CINEs. The developed interactive sequence was tested in simulations and prospectively in 13 subjects (10 for image evaluation, 2 during catheterization, 1 during exercise). In the prospective study, the optimized framework-HyperSLICE- was compared with conventional Cartesian real-time and breath-hold CINE imaging in terms quantitative and qualitative image metrics. Statistical differences were tested using Friedman chi-squared tests with post hoc Nemenyi test (p < 0.05).Results: In simulations the normalized RMS error, peak SNR, structural similarity, and Laplacian energy were all statistically significantly higher using optimized spiral compared to radial and uniform spiral sampling, particularly after scan plan changes (structural similarity: 0.71 vs. 0.45 and 0.43). Prospectively, HyperSLICE enabled a higher spatial and temporal resolution than conventional Cartesian real-time imaging. The pipeline was demonstrated in patients during catheter pull back, showing sufficiently fast reconstruction for interactive ***: HyperSLICE enables high spatial and temporal resolution interactive imaging. Optimizing the spiral sampling enabled better overall image quality and superior handling of image transitions compared with radial and uniform spiral trajectories.

关键词： cardiac MRI deep learning image reconstruction interventional real-time

来源：评论

学校读者我要写书评

暂无评论

real-time Classification of Vehicles Using Machine learning Algorithm on the Extensive Dataset

引用

IEEE ACCESS 2024年 12卷 98338-98351页

作者： Pemila, M. Pongiannan, R. K. Narayanamoorthi, R. Sweelem, Emad A. Hendawi, Essam Abu El-Sebah, Mohamed I. SRM Inst Sci & Technol Dept Elect & Elect Engn Kattankulathur 603203 India SRM Inst Sci & Technol Sch Comp Dept Comp Technol Kattankulathur 603203 India Elect Res Inst Dept PV Cells Cairo 12622 Egypt Taif Univ Coll Engn Dept Elect Engn Taif 21944 Saudi Arabia Elect Res Inst Dept Power Elect & Energy Convers Cairo 12622 Egypt

Vehicle classification (VC) is a prominent research domain within image processing and machine learning (ML) for identifying vehicle volumes and traffic rule violations. In developed countries, nearly 40% of daily accidents are fatal, while in developing countries, the figure rises to 70%. Traditionally, vehicle detection and classification have been performed manually by experts, which is difficult, time-consuming, and prone to errors. Furthermore, incorrect detection and classification can result in hazardous situations. This highlights the need for more reliable techniques to identify and classify vehicles accurately and practically. In existing applications, numerous automated methods have been proposed. However, employing deep and machine learning algorithms on complex datasets of vehicle images has failed to achieve accuracy in various climate conditions and has been time-consuming. This paper presents an accurate, robust, real-time system to classify vehicles from onsite roads. The proposed system utilizes a random wavelet transform for pre-processing, edge and region-based segmentation for feature extraction, an embedded method for feature selection, and the XGBoost algorithm for VC. The proposed work classifies vehicles under complex weather, illumination, color, and occlusion conditions over 10 datasets, including a novel dataset named SRM2KTR, containing 75,436 vehicle images on an FPGA platform. The results show 98.81% accuracy, outperforming the state-of-the-art (98%). The system was demonstrated with four different classifiers, classifying images in 0.16 ns with an average accuracy of 97.79%. The system exhibits high accuracy, rapid identification time, and robustness in practical use.

关键词： Feature extraction Accuracy Task analysis Machine learning algorithms Reluctance motors image edge detection Classification algorithms Gradient methods Boosting Vehicle classification machine learning eXtreme gradient boost algorithm

来源：评论

学校读者我要写书评

暂无评论

real-time deep learning-Based Object Recognition in Augmented reality

Real-Time Deep Learning-Based Object Recognition in Augmente...

引用

Conference on real-time processing of image, Depth, and Video Information

作者： Egipko, V Zhdanova, M. Gapon, N. Voronin, V. Semenishchev, E. Moscow State Univ Technol STANKIN Ctr Cognit Technol & Machine Vis Moscow Russia Don State Tech Univ Rostov Na Donu Russia

ISBN: (纸本)9781510673199;9781510673182

Augmented reality is a visualization technology that displays information by adding virtual images to the real world. Effective implementation of augmented reality requires recognition of the current scene. Identifying objects in real-time video on computationally limited hardware requires significant effort. One way to solve this problem is to create a hybrid system that, based on machine learning and computer vision technology, processes and analyzes visual data to identify and classify real-world objects. The proposed architecture is based on a combination of the Vuforia augmented system, which provides good performance by balancing prediction accuracy and efficiency. First, the Vuforia neural network architecture allows convenient interaction with AR in Unity and provides initial conditions for detecting 3D objects. The augmented reality construction algorithm is based on the ARCore framework and the OpenGL interface for embedded systems. The system integrates recognition data with an AR platform to display corresponding 3D models, allowing users to interact with them through the functionality of the AR application. This method also involves the development of an enhanced user interface for AR, making the augmented environment more accessible for navigation and control. Experimental research has shown that the proposed method significantly improves the accuracy of object recognition and the ease of working with 3D models in AR.

关键词： deep learning augmented reality computer vision object recognition robotic systems real-time processing

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：