检索结果-内蒙古大学图书馆

Robust detection of headland boundary in paddy fields from continuous RGB-D images using hybrid deep neural networks

引用

COMPUTERS AND ELECTRONICS IN AGRICULTURE 2023年第1期207卷

作者： Li, Dongfang Li, Boliao Long, Sifang Feng, Huaiqu Wang, Yongwei Wang, Jun Zhejiang Univ Dept Biosyst Engn 866 Yuhangtang Rd Hangzhou 310058 Peoples R China

Accurate and robust headland boundary detection in the field is crucial for formulating turning strategies for agro-machinery. The headland area is challenging to be detected because of its similar appearance to the farmland, coupled with complex field conditions such as high weed pressure and discontinuous headland edges. In this study, a machine vision-based method for headland boundary detection was proposed, which introduced depth information in addition to RGB images to improve its detection accuracy. A deep learning-based network combining convolutional neural network (CNN) and recurrent neural network (RNN) was constructed for headland semantic segmentation. An interactive attention module (IAM) was proposed to fuse complementary information in RGB-D images adaptively. A time series information processing module (TPM) composed of a set of bidirectional convolution long short-term memories (ConvLSTMs) was used to extract interrelated information from consecutive images. image preprocessing techniques and a proposed distance-based boundary point clus-tering algorithm were applied to the headland segmentation mask to obtain the boundary line on the working side of the agro-machinery. Ablation study results confirmed the effectiveness of IAM and TPM in improving the headland segmentation performance. The network presented achieved excellent headland segmentation per-formance, with a mean intersection over union (mIoU) of up to 95.7%. The mean deviation of boundary line extraction on 256*144 resolution images was 3.57 pixels. The detection rate reached 25.8 frames per second (FPS), which could provide real-time reference lines for agro-machinery turning path planning.

关键词： Machine vision Headland detection Multimodal fusion RGB-D ConvLSTM

来源：评论

学校读者我要写书评

暂无评论

VSIP 2021 - Proceedings of 2021 3rd International Conference on Video, Signal and image processing

VSIP 2021 - Proceedings of 2021 3rd International Conference...

引用

3rd International Conference on Video, Signal and image processing, VSIP 2021

ISBN: (纸本)9781450385886

The proceedings contain 22 papers. The topics discussed include: real time hand gesture recognition in industry;epidemic prevention system based on voice recognition combined with intelligent recognition of mask and helmet;activity recognition in industrial environment using two layers learning;a new method of specific emitter feature extraction based on IQ imbalance;mixup augmentation for deep hashing;multi-resolution Gabor descriptor for corrosion detection in pipeline video sequences;image deep steganography detection based on knowledge distillation in teacher-student network;a multi-scale framework for visual grounding;a comparison of three swarm-based optimization algorithms in wind turbine radar clutter micro-motion parameters estimation;the influence of accounting information system quality and human resource competency on information quality;measurement and analysis of electrophysiological propagation on the cardiac slice-based biosensor;and improve the field-of-view of cameras: consideration on the micro lens array.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Aerial image-Based Semantic Segmentation for Forest Fire Identification

Aerial Image-Based Semantic Segmentation for Forest Fire Ide...

引用

Artificial Intelligence Robotics, Signal and image processing (AIRoSIP), International Conference on

作者： Nadilla Asyifa Salma Mahmud Dwi Sulistiyo Risnandar Febryanti Sthevanie Gia Septiana Wulandari School of Computing Telkom University Bandung Indonesia School of Computing Telkom University Research Center for Data and Information Sciences BRIN Bandung Indonesia

ISBN: (数字)9798350327472

ISBN: (纸本)9798350327489

Fire is a serious disaster because it can endanger the safety of living things and cause great losses. Detection of forest fires using sensor-based devices or through analysis of weather data becomes less efficient because it is not real-time, its scope depends on sensor deployment, and sometimes it misses small fire spreads that are potentially dangerous. The key idea in this paper is to utilize a semantic segmentation approach to precisely detect forest fires and thus prevent the spread of larger fires. The proposed method uses aerial images since it is a practical way to tackle problems in fire monitoring and detection. Since similar research is still limited, let alone necessary to be implemented in countries potentially susceptible to forest fires, this study demonstrates deep learning-based semantic segmentation methods as preliminary research for identifying forest fires. We analyzed the implementation of the well-known FCN and U-Net models with some backbone options and used the FLAME dataset to train the models and evaluate them quantitatively. Based on the experimental results, FCN with VGG-16 backbone became the model with the highest Mean IoU value of 0.825. We also tested the best model on several aerial images arbitrarily taken from the internet to see the qualitative performance and analyze its possibility of being applied to situations outside the research dataset.

关键词： Analytical models Accuracy Semantic segmentation Forestry real-time systems Reliability Robots Monitoring Testing Meteorology

来源：评论

学校读者我要写书评

暂无评论

deep Separable Spatiotemporal learning for Fast Dynamic Cardiac MRI

arXiv

引用

arXiv 2024年

作者： Wang, Zi Xiao, Min Zhou, Yirong Wang, Chengyan Wu, Naiming Li, Yi Gong, Yiwen Chang, Shufu Chen, Yinyin Zhu, Liuhong Zhou, Jianjun Cai, Congbo Wang, He Guo, Di Yang, Guang Qu, Xiaobo Department of Electronic Science Intelligent Medical Imaging R&D Center Fujian Provincial Key Laboratory of Plasma and Magnetic Resonance National Institute for Data Science in Health and Medicine Xiamen University China Institute of Artificial Intelligence Xiamen University China Human Phenome Institute Fudan University China Department of Imaging Xiamen Cardiovascular Hospital of Xiamen University School of Medicine Xiamen University China Department of Cardiovascular Medicine Heart Failure Center Ruijin Hospital Lu Wan Branch Shanghai Jiaotong University School of Medicine China Shanghai Institute of Cardiovascular Diseases Zhongshan Hospital Fudan University China Department of Radiology Zhongshan Hospital Fudan University Department of Medical Imaging Shanghai Medical School Shanghai Institute of Medical Imaging China Fujian Province Key Clinical Specialty Construction Project Medical Imaging Department Xiamen Key Laboratory of Clinical Transformation of Imaging Big Data and Artificial Intelligence China School of Computer and Information Engineering Xiamen University of Technology China Department of Bioengineering and Imperial-X Imperial College London United Kingdom Department of Bioengineering Imperial College London United Kingdom

Dynamic magnetic resonance imaging (MRI) plays an indispensable role in cardiac diagnosis. To enable fast imaging, the k-space data can be undersampled but the image reconstruction poses a great challenge of high-dimensional processing. This challenge necessitates extensive training data in deep learning reconstruction methods. In this work, we propose a novel and efficient approach, leveraging a dimension-reduced separable learning scheme that can perform exceptionally well even with highly limited training data. We design this new approach by incorporating spatiotemporal priors into the development of a deep Separable Spatiotemporal learning network (deepSSL), which unrolls an iteration process of a 2D spatiotemporal reconstruction model with both temporal low-rankness and spatial sparsity. Intermediate outputs can also be visualized to provide insights into the network behavior and enhance interpretability. Extensive results on cardiac cine datasets demonstrate that the proposed deepSSL surpasses state-of-the-art methods both visually and quantitatively, while reducing the demand for training cases by up to 75%. Additionally, its preliminary adaptability to unseen cardiac patients has been verified through a blind reader study conducted by experienced radiologists and cardiologists. Furthermore, deepSSL enhances the accuracy of the downstream task of cardiac segmentation and exhibits robustness in prospectively undersampled real-time cardiac MRI. Copyright © 2024, The Authors. All rights reserved.

关键词： image reconstruction

来源：评论

学校读者我要写书评

暂无评论

ODAR: A Lightweight Object Detection Framework for Autonomous Driving Robots

ODAR: A Lightweight Object Detection Framework for Autonomou...

引用

International Conference on Digital image Computing - Techniques and Applications (DICTA)

作者： Duong, Le Hoang Trung, Huynh Thanh Tam, Pham Minh Ko, Gwangzeen Moon, Jung Ick Jo, Jun Hung, Nguyen Quoc Viet Griffith Univ Nathan Qld Australia Hanoi Univ Sci & Technol Hanoi Vietnam Elect & Telecommun Res Inst Daejeon South Korea

ISBN: (纸本)9781665417099

Object detection is an emerging and essential problem in recent years, which has been widely applied in many aspects of daily life such as video surveillance, self-driving robots, and automatic payment. The rapid development of deep learning models allows object detectors to work in real-time with high accuracy. However, such a sophisticated model often requires robust computing infrastructure such as powerful graphics processing units (GPUs). This requirement might cause a severe issue for embedded systems with small, power-efficient artificial intelligence (AI) systems like Jetson Nano, which are often restricted in both memory storage and computing sheer power. In this work, we aim to address this challenge by proposing a lightweight object detection framework that is specialized for the Internet of Things (IoT) devices with low-power processors such as Jetson Nano. In order to detect the object with different size, our framework employs a backbone residual CNN-based network as the feature extractor. We then design a multi-layer model to combine the feature at different levels of granularity, before using the processed feature to locate and classify the object. We also apply augmentation techniques to enhance the robustness of the framework to adversarial factors. Extensive experiments on real devices in many scenarios, such as autonomous cars or wireless robot recharging systems, showed that our technique can achieve nearly on par results with the state-of-the-art YOLOv5 while requires only one-fourth of computation power.

关键词： object detection autonomous robot driving image processing deep neural network

来源：评论

学校读者我要写书评

暂无评论

real-time License Plate Recognition and Vehicle Tracking System Based on deep learning

Real-Time License Plate Recognition and Vehicle Tracking Sys...

引用

22nd Asia-Pacific Network Operations and Management Symposium, APNOMS 2021

作者： Chen, Guan-Wen Yang, Chun-Min Lk, Tsi-Ui Institute of Computer Science and Engineering National Yang Ming Chiao Tung University Hsinchu City30010 Taiwan

ISBN: (纸本)9784885523328

Traditional license plate recognition technology mostly uses traditional image processing methods to find out the characteristics of the license plate, and then crop and recognize the characters. The process needs to be modified due to the different environments, scenes and conditions. In recent years, many studies have implemented license plate and character recognition by using deep learning algorithms. Although it has a good recognition accuracy, the calculation speed still cannot reach the level of real-time recognition. This research proposes a real-time license plate recognition system based on YOLOv3, which uses deep learning model to realize the vehicle license plate recognition, lane identification and vehicle trajectory tracking. In this study, a web-based platform is established to present the result of license plate recognition and trajectory, and the streaming roadside video in the campus. In the platform, license plates of driving vehicles can be identified in real-time, and the user can search and track specific vehicle intuitively. In the experiment, the average accuracy of the system performs 84.3% in real-time license plate recognition, and 100% in lane identification. The system can process in 40 FPS, which can meet the level of real-time system. In the future, the system can cooperate with traffic access control in campus or community to improve the efficiency of traffic control. © 2021 IEICE.

关键词： real time systems

来源：评论

学校读者我要写书评

暂无评论

MEMG: Mobile Encrypted Traffic Classification With Markov Chains and Graph Neural Network 23

MEMG: Mobile Encrypted Traffic Classification With Markov Ch...

引用

23rd IEEE International Conference on High Performance Computing and Communications, 7th IEEE International Conference on Data Science and Systems, 19th IEEE International Conference on Smart City and 7th IEEE International Conference on Dependability in Sensor, Cloud and Big Data Systems and Applications, HPCC-DSS-SmartCity-DependSys 2021

作者： Cai, Wei Gou, Gaopeng Jiang, Minghao Liu, Chang Xiong, Gang Li, Zhen Institute of Information Engineering Chinese Academy of Sciences China School of Cyber Security University of Chinese Academy of Sciences Beijing China

ISBN: (纸本)9781665494571

In recent years, user privacy and information Security have attracted widespread attention, encryption ratio of mobile traffic has increased tremendously, which has brought considerable challenges to traditional traffic classification meth-ods. Machine learning methods and deep learning methods have become mainstream methods to solve this problem. However, the existing machine learning methods require manual features and cannot adapt to the newly generated traffic patterns. deep learning methods are capable of learning features from the raw traffic sequences automatically but will increase the calculation costs. To address these challenges, in this paper, we propose a Mobile Encrypted Traffic Classification with Markov Chains and Graph Neural Network (MEMG). We use the Markov chains to mine the hidden topological information of the flow. Then we build the flow graph structure on this basis, add the sequence information of traffic in the node feature in the graph. We also design a Graph Neural Network-based classifier to learn the topological and sequential features from the graph. The classifier can map the graph structure to the embedding space, and classify different graph structures by embedding vector differences. We have done comprehensive experiments on both the real-world dataset and the public dataset. The real-world dataset contains the traffic of 29 commonly used mobile encrypted applications collected by us recently, and the total number of traffic exceeds 116,000. Our method outperforms the state-of-the-art methods by 6.1% and 3.5% of the accuracy rate on our dataset and public dataset, respectively. We also lessen the training time overhead and GPU memory usage by 40% and 46%, respectively. © 2021 IEEE.

关键词： Training deep learning Privacy Smart cities Graphics processing units Markov processes Traffic control

来源：评论

学校读者我要写书评

暂无评论

Research on intelligent damage detection of far-sea cage based on machine vision and deep learning

引用

AQUACULTURAL ENGINEERING 2022年 96卷

作者： Liao, Wenxuan Zhang, Shubin Wu, Yinghao An, Dong Wei, Yaoguang China Agr Univ Coll Informat & Elect Engn Beijing 100083 Peoples R China China Agr Univ Natl Innovat Ctr Digital Fishery Beijing 100083 Peoples R China Beijing Engn & Technol Res Ctr Internet Things Ag Beijing 100083 Peoples R China Minist Agr Key Lab Agr Informat Acquisit Technol Beijing 100083 Peoples R China

Far-sea cage is an essential way for aquaculture. In the process of far-sea cage aquaculture, the damage of net structure can cause severe economic property losses to farmers, so it is necessary to check the cage's integrity. At present, the typical way to inspect cages is to hire professional divers for manual inspection. This type of inspection is time-consuming and has security concerns. This paper proposes a method for detecting the damage of a far-sea cage based on machine vision and deep learning, which can detect the structure of a far-sea cage in real time and accurately detect the damaged area of the cage. Firstly, the cage image data were collected by autonomous cruising ROV. According to the characteristics of the captured images, an improved multi-scale fusion algorithm was proposed to better the performance of denoising and smoothing effect of the original method. Secondly, we use the MobileNet-SSD and key-frame extraction detection method to detect the damage of underwater cage video. The MobileNet-SSD model has been optimized in model size and detection speed compared with the SSD model. In the experiment, the simulated damaged images of the far-sea cage were used for testing. The experimental results have shown that the scheme can improve the efficiency of far-sea cage inspection and accurately detect the damaged areas in the cage in real-time.

关键词： Aquaculture Far-sea Cage Damage Detection deep learning

来源：评论

学校读者我要写书评

暂无评论

A Multi-Mode 8K-MAC HW-Utilization-Aware Neural processing Unit with a Unified Multi-Precision Datapath in 4nm Flagship Mobile SoC

A Multi-Mode 8K-MAC HW-Utilization-Aware Neural Processing U...

引用

2022 IEEE International Solid-State Circuits Conference, ISSCC 2022

作者： Park, Jun-Seok Park, Changsoo Kwon, Suknam Kim, Hyeong-Seok Jeon, Taeho Kang, Yesung Lee, Heonsoo Lee, Dongwoo Kim, James Lee, YoungJong Park, Sangkyu Jang, Jun-Woo Ha, SangHyuck Kim, MinSeong Bang, Jihoon Lim, Suk Hwan Kang, Inyup Samsung Electronics Hwaseong Korea Republic of Samsung Advanced Institute of Technology Suwon Korea Republic of

ISBN: (纸本)9781665428002

Recent work on neural-network accelerators has focused on obtaining high performance in order to meet the needs of real-time applications with vastly different performance requirements, including high precision computation, efficiency for various deep learning (DL) layer types, and extremely low power to run always-on applications. Applying a single mode or datatype uniformly across these different scenarios would be less efficient than using different operating modes according to different operating scenarios. For example, super-resolution typically requires FP16 precision for higher image quality, while NNs for face-detection need only INT4 or INT8 precision. Using higher precision than INT8 for face detection would result in higher power consumption. A highly programmable NPU capable of covering the diverse workloads observed in the real world is therefore desired. In this paper, we present a neural processing unit (NPU) optimized with the following features: i) reconfigurable data prefetching and operational flow for high compute utilization, ii) multi-precision MACs supporting INT4, 8, 16, and float16, iii) a dynamic operation mode to cover extremely low-power or low-latency requirements. These features provide the flexibility needed by real world applications within the power constraints of various product domains. © 2022 IEEE.

关键词： Face recognition

来源：评论

学校读者我要写书评

暂无评论

基于深度学习的公路货车车型识别

引用

交通运输工程学报 2023年第1期23卷 267-279页

作者：张念张亮太原理工大学土木工程学院山西太原030024 太原科技大学交通与物流学院山西太原030024

为判断公路货车车型,并提升货车车型识别的速度与精度,提出基于深度学习的方法对公路货车及其轮轴进行精细化目标检测;采用道路监控拍摄和网络爬取的方式获得了16 403张公路货车侧方图像,建立了货车侧方图像数据集,并采用Retinex理论和... 详细信息

为判断公路货车车型,并提升货车车型识别的速度与精度,提出基于深度学习的方法对公路货车及其轮轴进行精细化目标检测;采用道路监控拍摄和网络爬取的方式获得了16 403张公路货车侧方图像,建立了货车侧方图像数据集,并采用Retinex理论和加入限制对比度的自适应直方图均衡化(CLAHE)等视觉增强方法预处理所采集图像中的光照不均图像和夜视图像;通过理论分析和对比试验选取单阶段检测网络YOLOv3作为公路货车车型识别的目标检测网络,并从调整先验框和模型输入大小以及引入注意力机制3个方面优化了检测模型;针对单帧图像可能同时出现多辆货车的情况,采用基于目标位置信息挖掘的算法分析了货车与轮轴的位置信息,提出一种通过轮轴中心点与货车预测框位置信息判定公路货车与轮轴隶属关系的方法。研究结果表明:图像经过预处理可显著增强车辆的特征信息,优化后检测模型的网络性能得到提高,通过对目标位置信息的挖掘与利用可以很好地解决货车车型判定问题;优化后的检测模型实时检测速度可达47帧·s^(-1),对公路货车车型的识别综合准确率达到了94.4%。该方法实现了对公路货车车型的无接触、快速和准确识别,为公路货车车型识别提供了新的手段,符合智慧交通系统的建设需要,可进一步提升道路服务水平。

关键词：交通信息工程公路货车深度学习车型识别注意力机制图像处理

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：