检索结果-内蒙古大学图书馆

Catalysing assistive solutions by deploying light-weight deep learning model on edge devices

JOURNAL OF EXPERIMENTAL & THEORETICAL ARTIFICIAL INTELLIGENCE 2025年第3期37卷 465-486页

作者： Manjari, Kanak Verma, Madhushi Singal, Gaurav Chamola, Vinay Bennett Univ Dept Comp Sci Engn Greater Noida India NSUT Dept Comp Sci & Engn Delhi India BITS Dept Elect & Elect Engn Pilani India

Nowadays, real-time object detection, which is a crucial task, is being performed through image processing and deep learning techniques. As there are several high-performance computing edge devices available, selecting the best-fit device for a particular problem is a tough task and keeping in mind the cost, performance, and weight of the device in mind. One faces several challenges while performing this task in real-time such as a lack of resources in terms of power and mobility. We have provided an insight into the computation power of devices in terms of Frames per Second (FPS) by deploying object detection models on them. This paper will provide insight into selecting the appropriate combination of device and object detection models for real-time applications. Raspberry Pi 3 (RPi3), Raspberry Pi 4 (RPi4), Intel Neural Compute Stick 2 (NCS2), and Nvidia Jetson NANO are popular devices with high computation power used for real-time applications. The memory constraints of devices along with the deployment of different You Only Look Once (YOLO) and Single-Shot Detector (SSD) are the two object detection models that have been explained in this paper. A deep learning inference optimiser, TensorRT, has been used in NANO to achieve high throughput in the performance of object detection. The precision, recall, and F1 score achieved on deploying each tested model have been presented. After observing the devices during experimentation, RPi4+NCS2 showed the best execution with the blend of factors i.e. speed, portability, and user-friendliness.

关键词： Assistive technology edge devices raspberry neural compute stick jetson NANO

来源：评论

学校读者我要写书评

暂无评论

A lightweight real-time algorithm for plum harvesting detection in orchards under complex conditions

引用

SIGNAL image AND VIDEO processing 2025年第4期19卷 1-12页

作者： Zhang, Dongyan Chen, Nuo Mao, Siyu Wu, Chenxv Gu, Dandan Zhang, Lanxiang Univ Northeast Forestry Coll Comp & Control Engn Harbin 150040 Peoples R China

Accurate and efficient algorithms for detection are essential technologies utilized by robots for plum harvesting. To facilitate effective implementation on edge computing devices that have restricted processing capabilities, this research introduces a lightweight plum detection algorithm derived from YOLOv5s. Initially, the backbone network of the original architecture is substituted with the Revcol network. Subsequently, the feature space pyramid pooling (SPPF) is refined by integrating a Large Separable Kernel Attention (LSKA) mechanism, and the C3 module of the original model is improved through the use of Distributed Shift Convolutions (DSConv). Finally, a custom-designed lightweight detection head is employed to construct the lightweight plum detection model. The results show that the average mAP of the improved detection model is 96.3%, and the number of floating point operations (FLOPs) and parameters are reduced to 84.8% and 91.2% respectively. On edge computing devices, the inference speed reaches 44.64 frames per millisecond. This algorithm features high accuracy, lightweight design, and fast inference speed, providing valuable references for real-time object detection deployment of plum harvesting robots on edge computing devices.

关键词： deep learning Plum detection Target detection Smart picking real-time

来源：评论

学校读者我要写书评

暂无评论

Joint Spatial and Frequency Domain learning for Lightweight Spectral image Demosaicing

引用

IEEE TRANSACTIONS ON image processing 2025年 34卷 1119-1132页

作者： Wu, Fangfang Huang, Tao Xu, Junwei Cao, Xun Dong, Weisheng Dong, Le Shi, Guangming Xidian Univ Sch Comp Sci & Technol Xian 710071 Peoples R China Xidian Univ Sch Artificial Intelligence Xian 710071 Peoples R China Yealink Xiamen 361009 Peoples R China Nanjing Univ Sch Elect Sci & Engn Nanjing 210093 Peoples R China

Conventional spectral image demosaicing algorithms rely on pixels' spatial or spectral correlations for reconstruction. Due to the missing data in the multispectral filter array (MSFA), the estimation of spatial or spectral correlations is inaccurate, leading to poor reconstruction results, and these algorithms are time-consuming. deep learning-based spectral image demosaicing methods directly learn the nonlinear mapping relationship between 2D spectral mosaic images and 3D multispectral images. However, these learning-based methods focused only on learning the mapping relationship in the spatial domain, but neglected valuable image information in the frequency domain, resulting in limited reconstruction quality. To address the above issues, this paper proposes a novel lightweight spectral image demosaicing method based on joint spatial and frequency domain information learning. First, a novel parameter-free spectral image initialization strategy based on the Fourier transform is proposed, which leads to better initialized spectral images and eases the difficulty of subsequent spectral image reconstruction. Furthermore, an efficient spatial-frequency transformer network is proposed, which jointly learns the spatial correlations and the frequency domain characteristics. Compared to existing learning-based spectral image demosaicing methods, the proposed method significantly reduces the number of model parameters and computational complexity. Extensive experiments on simulated and real-world data show that the proposed method notably outperforms existing spectral image demosaicing methods.

关键词： Frequency-domain analysis Imaging image reconstruction Transformers Feature extraction Correlation Fourier transforms Interpolation Three-dimensional displays Optimization Spectral image demosaicing multispectral filter array (MSFA) deep learning spatial domain frequency domain transformer

来源：评论

学校读者我要写书评

暂无评论

Meta learning based Object Tracking Technology: A Survey

引用

KSII TRANSACTIONS ON INTERNET AND INFORMATION SYSTEMS 2024年第8期18卷 2067-2081页

作者： Baek, Ji-Won Chung, Kyungyong Kyonggi Univ Dept Comp Sci 154-42 Gwanggyosan Ro Suwon 16227 Gyeonggi Do South Korea Kyonggi Univ Div AI Comp Sci & Engn 154-42 Gwanggyosan Ro Suwon 16227 Gyeonggi Do South Korea

Recently, image analysis research has been actively conducted due to the accumulation of big image data and the development of deep learning. image analytics research has different characteristics from other data such as data size, real-time, image quality diversity, structural complexity, and security issues. In addition, a large amount of data is required to effectively analyze images with deep-learning models. However, in many fields, the data that can be collected is limited, so there is a need for meta learning based image analysis technology that can effectively train models with a small amount of data. This paper presents a comprehensive survey of meta-learning-based object-tracking techniques. This approach comprehensively explores object tracking methods and research that can achieve high performance in datalimited situations, including key challenges and future directions. It provides useful information for researchers in the field and can provide insights into future research directions.

关键词： Convolution Neural Network Object Tracking Meta learning deep learning Object Detection

来源：评论

学校读者我要写书评

暂无评论

Scanning imaging through the scattering medium based on the guided filter assisted by a WTMM and NLM

引用

APPLIED OPTICS 2025年第10期64卷 2408-2418页

作者： Zhang, Yiyang Ding, Zhewen Li, Runze Meng, Xianglei Shi, Yan Zhan, Chunlian Zhao, Chunliu China Jiliang Univ Coll Opt & Elect Technol Hangzhou 310018 Zhejiang Peoples R China State Adm Market Regulat Key Lab Light Sensing & Image Metrol Hangzhou 310018 Zhejiang Peoples R China Zhejiang Prov Key Lab Optoelect Metrol & Precis In Hangzhou 310018 Zhejiang Peoples R China

In a scanning imaging system through a scattering medium, the quality of the imaging result is related to the energy distribution of the focusing point. In actual imaging, the energy of the focusing point cannot be perfectly concentrated. The scattering noise is always surrounded by the focusing point, which reduces the signal-to-noise ratio and results in poor image quality. To improve the quality of imaging, further processing of the imaging results is required, while most conventional image processing methods can only achieve one of the goals such as noise reduction, image smoothing, edge sharpening, or maintaining image integrity. In this paper, a scanning imaging system through the scattering medium based on an adaptive guided filter assisted by the wavelet transform modulus maximum (WTMM) and non-local mean (NLM) is proposed for the first time, to the best of our knowledge, which can make the imaging results simultaneously have the advantages of low noise, high contrast, and clear details. To verify the validity of the proposed method, a scanning imaging system through the scattering medium was setup. Transmissive imaging was performed at different positions from the focal plane. The experimental results show that the background noise is significantly restrained, single-pixel response and edge continuity are good, and details are clear with the proposed method. Compared with the traditional methods and the deep learning methods, the proposed method can improve the PSNR and SSIM by up to 10.68 dB, 0.75 and 5.34 dB, 0.72, respectively. Finally, the performance of the proposed method in high-noise environments, its application in the field of real-time imaging, and future improvements are discussed. The method proposed in this paper can effectively improve the quality of scanning imaging results through the scattering medium, which is expected to promote the application of this technology in endoscopic imaging and other fields. (c) 2025 Optica Publishing Group. A

关键词： deep learning Imaging systems Imaging techniques real time imaging Scattering media Wavelet transforms

来源：评论

学校读者我要写书评

暂无评论

deep learning-based instance segmentation on 3D laser triangulation data for inline monitoring of particle size distributions in construction and demolition waste recycling

引用

RESOURCES CONSERVATION AND RECYCLING 2024年 205卷

作者： Wu, Xiaoye Kroell, Nils Greiff, Kathrin Rhein Westfal TH Aachen Dept Anthropogen Mat Cycles Wuellnerstr 2 D-52062 Aachen Germany

Overlapping material flow presentations in construction and demolition waste (CDW) recycling make an inline particle size distribution (PSD) monitoring challenging. Here, we aim to build a deep-learning-based segmentation model for overlapping particles in 3D -laser -triangulation images of CDW. Our model was trained on three specially designed datasets with two transfer learning processes. U -net was employed as the backbone and MultiStar algorithm was used to describe particle shapes. The final model demonstrated an impressive performance on test set, with a mean average precision (mAP) of 92.8% at IoU = 0.5. Comparing with the traditional segmentation algorithm based on image processing methods, the mAP can only reach to 27.4% on the same images. The shown model performance paves the way toward novel sensor technology applications for real -time PSD monitoring in CDW recycling.

关键词： Sensor-based material flow characterization Digital image processing deep learning Artificial intelligence Instance segmentation Circular economy

来源：评论

学校读者我要写书评

暂无评论

A real-time Tracking Approach for Moving Objects Based on an Integrated Algorithm of YOLOv7 and SORT

引用

JOURNAL OF CIRCUITS SYSTEMS AND COMPUTERS 2025年第4期34卷

作者： Wang, Lei Liu, Tongtong Yellow River Conservancy Tech Inst Coll Elect Engn Kaifeng 475004 Henan Peoples R China

Mobile target tracking remains a significant issue in smart cities. Due to complex changes in time and space of targets, real-time tracking remains a challenging problem. As a result, this paper proposes a real-time tracking approach for moving objects by combining the advantages of YOLOv7 and SORT algorithms. First, we use the YOLOv7 algorithm for object detection, which has the characteristics of high accuracy and efficiency. Then, we apply the SORT algorithm to the target tracking stage, which estimates and updates the target state through Kalman filtering. The collaborative function of the two parts is expected to achieve high-quality tracking of moving targets. Besides, this paper also demonstrates experiments and analysis on image datasets. The experimental results show that the proposed algorithm has achieved good performance in real-time tracking of moving targets. Compared with traditional methods, it can more accurately predict the position and trajectory of targets and has better real-time performance. In addition, the proposed algorithm is equally effective for target tracking in complex scenes, such as multi-target tracking and target occlusion. Future research can further optimize the performance of algorithms to cope with more complex scenarios and problems.

关键词： Moving target tracking real-time image processing YOLOv7 Kalman filtering deep learning

来源：评论

学校读者我要写书评

暂无评论

On-road obstacle detection in real time environment using an ensemble deep learning model

引用

SIGNAL image AND VIDEO processing 2024年第6-7期18卷 5387-5400页

作者： Balasundaram, Ananthakrishnan Shaik, Ayesha Prasad, A. Pratheepan, Yogarajah Vellore Inst Technol Ctr Cyber Phys Syst Chennai Tamil Nadu India Vellore Inst Technol VIT Sch Comp Sci & Engn Chennai Tamil Nadu India Ulster Univ Sch Comp Engn & Intelligent Syst Londonderry North Ireland

Obstacle detection on road is a challenging task in autonomous vehicle driving. Although obstacle detection is carried out with the help of sensors which are accurate and precise in real-time, they are not cost effective and computationally intensive. So, a computer vision and deep learning-based approach can be considered as a potential alternative. The proposed system is an ensemble of two instance segmentation algorithms namely You Only Look Once v7 (YOLOv7) and Mask Region-Based Convolutional Neural Network (Mask R-CNN) used to detect various obstacles on road. The system was tested on popular obstacle detection datasets such as INRIA, KAIST, and COCO2017. The ensemble model achieved mean average precision score of 90.1, 95.4 and 88.9 and mean intersection over union score of 93.3, 89.7 and 91.4 for KAIST, INRIA and COCO datasets respectively. A custom dataset for detecting obstacles was developed and the proposed model was tested on the custom dataset as well.

关键词： Obstacle detection YOLOV7 Mask R-CNN deep learning Ensemble model

来源：评论

学校读者我要写书评

暂无评论

Sensor-based characterization of construction and demolition waste at high occupancy densities using synthetic training data and deep learning

引用

WASTE MANAGEMENT & RESEARCH 2024年第9期42卷 788-796页

作者： Kronenwett, Felix Maier, Georg Leiss, Norbert Gruna, Robin Thome, Volker Laengle, Thomas Inst Optron Fraunhofer IOSB Syst Technol & Image Exploitat Karlsruhe Germany Fraunhofer IBP Inst Bldg Phys Holzkirchen Germany Inst Optron Fraunhofer IOSB Syst Technol & Image Exploitat Fraunhoferstr 1 D-76131 Karlsruhe Germany

Sensor-based monitoring of construction and demolition waste (CDW) streams plays an important role in recycling (RC). Extracted knowledge about the composition of a material stream helps identifying RC paths, optimizing processing plants and form the basis for sorting. To enable economical use, it is necessary to ensure robust detection of individual objects even with high material throughput. Conventional algorithms struggle with resulting high occupancy densities and object overlap, making deep learning object detection methods more promising. In this study, different deep learning architectures for object detection (Region-based CNN/Region-based Convolutional Neural Network (Faster R-CNN), You only look once (YOLOv3), Single Shot MultiBox Detector (SSD)) are investigated with respect to their suitability for CDW characterization. A mixture of brick and sand-lime brick is considered as an exemplary waste stream. Particular attention is paid to detection performance with increasing occupancy density and particle overlap. A method for the generation of synthetic training images is presented, which avoids time-consuming manual labelling. By testing the models trained on synthetic data on real images, the success of the method is demonstrated. Requirements for synthetic training data composition, potential improvements and simplifications of different architecture approaches are discussed based on the characteristic of the detection task. In addition, the required inference time of the presented models is investigated to ensure their suitability for use under real-time conditions.

关键词： Machine learning object detection synthetic data sensor-based sorting construction and demolition waste circular economy

来源：评论

学校读者我要写书评

暂无评论

Applying deep learning and Heatmap Techniques for Sinusitis Diagnosis Using Enhanced 3-D image Segmentation

引用

IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT 2025年 74卷

作者： Huang, Pin-Siang Horng, Shi-Jinn Yang, Chia-Jung Chen, Hongmei Zhou, Wanlei Natl Taiwan Univ Sci & Technol Dept Comp Sci & Informat Engn Taipei 10607 Taiwan Asia Univ Dept Comp Sci & Informat Engn Taichung 41354 Taiwan China Med Univ China Med Univ Hosp Dept Med Res Taichung Taiwan Univ Technol Sydney Sch Comp Sci Ultimo NSW 2007 Australia Taitung MacKay Mem Hosp Dept Radiol Taitung 950 Taiwan Southwest Jiaotong Univ Sch Comp & Artificial Intelligence Chengdu 610032 Peoples R China City Univ Macau Fac Data Sci Macau Peoples R China

Sinusitis, characterized by inflammation of the sinus cavities, presents diagnostic challenges due to the subjective and time-consuming nature of manual computed tomography (CT) scan annotations. This study introduces a deep learning (DL) method combined with heatmap analysis to automate 3-D segmentation of sinusitis in CT images. The proposed approach reduces the dependency on fully segmented annotations, instead requiring only binary classification labels during training. A classification model identifies areas likely affected by inflammation, highlighted by gradient-weighted class activation mapping (Grad-CAM) to produce heatmaps, which are refined using a watershed algorithm for precise segmentation. This approach reduces annotation preparation time from 30 min to under 1 min while achieving a Dice similarity coefficient (DSC) of 90.04%. Built on the EfficientNet-B2 architecture, the model outperforms state-of-the-art methods like U-Net, U-Net++, and statistical-based approaches. Additionally, a novel instrumentation system integrated with CT scanners calculates inflammation areas in real time, streamlining diagnostic workflows. This method accelerates the process while improving the consistency and reliability of sinusitis diagnosis, potentially transforming clinical practices with advanced image processing techniques.

关键词： image segmentation Computed tomography deep learning Watersheds Three-dimensional displays Accuracy Medical diagnostic imaging Convolutional neural networks Annotations Training 3-D segmentation computed tomography (CT) scan deep learning (DL) diagnostic automation Dice similarity coefficient (DSC) gradient-weighted class activation mapping (GRAD-CAM) image annotation medical image processing sinusitis watershed algorithm

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：