检索结果-内蒙古大学图书馆

Fault Diagnosis of Harmonic Reducers Based On a FDG-Resnet Joint Methodology

Fault Diagnosis of Harmonic Reducers Based On a FDG-Resnet J...

2023 International Conference on Sensing, Measurement and Data Analytics in the Era of Artificial Intelligence, ICSMD 2023

作者： Lv, Jing Long, Zhuo Wang, Juan Ma, Xiaoguang College of Electronics and Information Engineering Heilongjiang University of Science and Technology Harbin China Engineering and Also with the Foshan Graduate School Northeastern University Northeastern University Faculty of Robot Science Foshan China

ISBN: (纸本)9798350318012

Harmonic reducers(HRs) are widely used in various industrial fields due to its advantages of high precision, high transmission efficiency and compact structure. However, the vibration signal of the HRs are complex, making it difficult to diagnose faults accurately. In this paper, a method based on frequency domain Gramian angular field (FDG) and Deep residual network was proposed to classify the fault states of the HRs. First, the vibration signals were converted into twodimensional images using Gramian angular field (GAF) coding, and a two-dimensional discrete Fourier transform (2D-DFT) were performed to transform the images from the spatial domain to the frequency domain;The training accuracy of traditional models is not very high due to their low number of network layers. The deep residual network not only has a very deep network layer, but also can guarantee the accuracy of the model under certain conditions, so the deep residual network is introduced to train the model. Finally, experiments were designed using data from a real industrial HRs health monitoring platform to validate the proposed method. Under fixed operating conditions, the proposed method achieved an accuracy of 99%, which was an improvement of 0.3% compared to ResNet50, 3.5% compared to VGGNet-16, and 2.9% compared to the Swin-Tiny model. Under variable load conditions, it achieved an accuracy of 99.4%, an increase of 0.7% compared to ResNet50, 5.5% compared to VGGNet-16, and 5.6% compared to Swin-Tiny. Under variable speed conditions, it achieved an accuracy of 94.3%, an improvement of 0.9% compared to ResNet50, 0.1% compared to VGGNet-16, and 9.3% compared to Swin-Tiny. © 2023 IEEE.

关键词： Network layers

来源：评论

学校读者我要写书评

暂无评论

Efficient Attentional Underwater Image Enhancement Generative Adversarial Network

Efficient Attentional Underwater Image Enhancement Generativ...

引用

Chinese Control and Decision Conference, CCDC

作者： Haopeng Zhang Hongli Xu Xiaosheng Yu Junxiang Wang Chengdong Wu Faculty of Robot Science and Engineering Northeastern University Shenyang China

ISBN: (数字)9798350387780

ISBN: (纸本)9798350387797

In the domain of underwater imaging, the attenuation of light, varying light conditions, and suspended particles often degrade image clarity, resulting in a loss of visual information and color distortion. The proposed Efficient Attentional Underwater Image Enhancement Generative Adversarial Network (EAU-GAN) offers a novel approach to address these challenges by leveraging the power of attention mechanisms in deep learning. This model employs an attention-driven generator that adaptively focuses on critical regions in the underwater image, thereby effectively highlighting obscured details while suppressing unnecessary noise. The discriminator, trained adversarially with the generator, ensures the generation of high-quality, naturalistic images. The innovation of EAU-GAN lies in its attention-driven generator that adaptively concentrates on crucial regions within underwater imagery. This targeted focus enhances obscured details and attenuates extraneous noise, markedly improving image clarity and color fidelity. Experimental results show that EAU-GAN outperforms other methods in both qualitative and quantitative aspects.

关键词： Technological innovation Visualization Image color analysis Image synthesis Imaging Generative adversarial networks Distortion

来源：评论

学校读者我要写书评

暂无评论

Attention-Based Generative Grasping Deformable Convolutional Neural Network

Attention-Based Generative Grasping Deformable Convolutional...

引用

International Conference on robotics and Automation engineering (ICRAE)

作者： Kang-Xiong Qin Ru-Jie Fu Sheng-Wei Zhang Yang Jiang Faculty of Robot Science and Engineering Northeastern University Shenyang China

ISBN: (数字)9798331518301

ISBN: (纸本)9798331518318

Traditional generative grasping convolutional neural network suffers from insufficient feature selection capability and limited adaptability to complex geometric transformations in unstructured environments. To address these issues, this study proposes an optimizational method based on the Convolutional Block Attention Module (CBAM) and Deformable Convolutional Networks (DCN) to improve the performance of Generative Grasping Convolutional Neural Network (GG-CNN) in robotic grasping tasks. The Network is trained and evaluated using the Cornell and Jacquard datasets, the simulation and real-world experiments are conducted. The experimental results show that the AGG-DCNN exhibits better performance in model training and grasping tasks, achieving an accuracy of 97.19% on the Jacquard dataset. Additionally, the network shows significant improvements in convergence speed and robustness in learning accuracy.

关键词： Training Accuracy Deformation Grasping Interference Feature extraction Robustness Surface texture Convolutional neural networks robots

来源：评论

学校读者我要写书评

暂无评论

Weakly Supervised Segmentation Framework for Thyroid Nodule Based on High-confidence Labels and High-rationality Losses

arXiv

引用

arXiv 2025年

作者： Chi, Jianning Li, Zelan Lin, Geng Sun, MingYang Yu, Xiaosheng Faculty of Robot Science and Engineering Northeastern University Shenyang110169 China

Weakly supervised segmentation methods can delineate thyroid nodules in ultrasound images efficiently using training data with coarse labels, but suffer from: 1) low-confidence pseudo-labels that simply follow topological priors, introducing significant label noise, and 2) low-rationality loss functions that rigidly compare segmentation with labels, ignoring discriminative information for nodules with diverse and complex shapes. To solve these problems, we clarify the objective together with required references of weakly supervised ultrasound image segmentation, and present the framework with high-confidence pseudo-labels to represent topological and anatomical information, and high-rationality losses to learn multi-level discriminative information. Specifically, we fuse geometric transformations of four-point annotations and results from the MedSAM model prompted by certain annotations to generate high-confidence box, foreground, and background labels. We design a high-rationality learning strategy comprising: 1) Alignment loss that measures the spatial projection consistency between the segmentation and box label, and the topological continuity of the segmentation within foreground label, guiding the network to perceive the location arrangement of nodule features;2) Contrastive loss that pulls features sampled from labeled foreground regions, while pushing features sampled from labeled foreground and background regions, guiding the network to capture the regional distribution of nodule and background features;3) Prototype correlation loss that measures the consistency between correlation maps derived by comparing features with foreground and background prototypes respectively, shrinking the uncertain regions to precise nodule edge delineation. Experimental results demonstrate that our method achieves state-of-the-art performance on the publicly available TN3K and DDTI datasets. The code is publicly available at HCL-HRL. © 2025, CC BY.

关键词： Ultrasonic imaging

来源：评论

学校读者我要写书评

暂无评论

Collaborative Path Planning for Multiple Autonomous Underwater Vehicles Based on Improved Particle Swarm Optimization Algorithm

Collaborative Path Planning for Multiple Autonomous Underwat...

引用

Chinese Control and Decision Conference, CCDC

作者： Mian Wang Hongli Xu Faculty of Robot Science and Engineering Northeastern University Shenyang China

ISBN: (数字)9798350387780

ISBN: (纸本)9798350387797

To solve the collaborative path planning problem of multiple autonomous underwater vehicles in complex environments, an improved particle swarm optimization algorithm is proposed to address the problems of traditional particle swarm optimization algorithms being prone to falling into local optima and having slow convergence speed, resulting in unreasonable search results. Firstly, a dynamic optimization formation transformation obstacle avoidance strategy was adopted for the multi AUV formation problem, and a formation obstacle avoidance strategy was designed that includes multiple modes of formation zero transformation, isomorphic transformation, and heterogeneous transformation. At the same time, this algorithm nonlinearly functionalizes the invariant inertia weights and acceleration constants in traditional particle swarm optimization algorithms, thereby optimizing the optimization ability and solving the problem of traditional particle swarm algorithms being prone to falling into local optima.

关键词： Autonomous underwater vehicles Heuristic algorithms Collaboration Search problems Path planning Particle swarm optimization Vehicle dynamics

来源：评论

学校读者我要写书评

暂无评论

Cross-level Diffusion-based Affinity Learning for Unsupervised Salient Object Detection

Cross-level Diffusion-based Affinity Learning for Unsupervis...

引用

Chinese Control and Decision Conference, CCDC

作者： Pengfei Lyu Xiaosheng Yu Chengdong Wu Faculty of Robot Science and Engineering Northeastern University Shenyang China

ISBN: (数字)9798350387780

ISBN: (纸本)9798350387797

Numerous bottom-up salient object detection (SOD) methods rely on local similarity (affinity) to construct, which ignores the relationships among non-adjacent regions. To solve this problem, we propose a novel framework based on absorbing Markov chain (AMC) with cross-level diffusion-based affinity learning, namely CLD. In the AMC model, high transition probability implies reliable relationships among regions. For this purpose, a dense transition probability matrix is learned by a cross-level diffusion strategy. First, to comprehensively reveal the relationship among regions, multi-level features are extracted and are used to calculate multiple initial sparse affinity matrices. Second, the tensor product in the iterative diffusion process has the property of capturing the high-order relationships among features. Thus, by iteratively updating sparse affinity matrices, multiple-level features are effectively fused. Finally, through a weight learning paradigm, outlier data is removed and noise is fully suppressed. Meanwhile, learning weights are also assigned for each sparse affinity matrix and they are integrated into a dense transition probability matrix. Furthermore, to fit salient objects at multiple scales, we propose a refinement model MCLD which is a Bayesian framework under the guidance of high-level features. The MCLD aggregates saliency maps generated by CLD at different scales, and calibrates their shortcomings to produce more stable results. Extensive experiments on five challenging datasets indicate that the proposed method outperforms ten state-of-the-art algorithms.

关键词： Tensors Aggregates Noise Diffusion processes Object detection Feature extraction Bayes methods

来源：评论

学校读者我要写书评

暂无评论

Observation Time Difference: an Online Dynamic Objects Removal Method for Ground Vehicles

Observation Time Difference: an Online Dynamic Objects Remov...

引用

IEEE International Conference on robotics and Automation (ICRA)

作者： Rongguang Wu Chenglin Pang Xuankang Wu Zheng Fang Faculty of Robot Science and Engineering Northeastern University Shenyang China

ISBN: (数字)9798350384574

ISBN: (纸本)9798350384581

In the process of urban environment mapping, the sequential accumulations of dynamic objects will leave a large number of traces in the map. These traces will usually have bad influences on the localization accuracy and navigation performance of the robot. Therefore, dynamic objects removal plays an important role for creating clean map. However, conventional dynamic objects removal methods usually run offline. That is, the map is reprocessed after it is constructed, which undoubtedly increases additional time costs. To tackle the problem, this paper proposes a novel method for online dynamic objects removal for ground vehicles. According to the observation time difference between the object and the ground where it is located, dynamic objects are classified into two types: suddenly appear and suddenly disappear. For these two kinds of dynamic objects, we propose downward retrieval and upward retrieval methods to eliminate them respectively. We validate our method on SemanticKITTI dataset and author-collected dataset with highly dynamic objects. Compared with other state-of-the-art methods, our method is more efficient and robust, and reduces the running time per frame by more than 60% on average. Our method will be open-sourced on GitHub 1 .

关键词： Location awareness Costs Accuracy Navigation Urban areas Land vehicles Object recognition

来源：评论

学校读者我要写书评

暂无评论

FlowTrack: Point-level Flow Network for 3D Single Object Tracking

arXiv

引用

arXiv 2024年

作者： Li, Shuo Cui, Yubo Li, Zhiheng Fang, Zheng Faculty of Robot Science and Engineering Northeastern University Shenyang China

3D single object tracking (SOT) is a crucial task in fields of mobile robotics and autonomous driving. Traditional motion-based approaches achieve target tracking by estimating the relative movement of target between two consecutive frames. However, they usually overlook local motion information of the target and fail to exploit historical frame information effectively. To overcome the above limitations, we propose a point-level flow method with multi-frame information for 3D SOT task, called FlowTrack. Specifically, by estimating the flow for each point in the target, our method could capture the local motion details of target, thereby improving the tracking performance. At the same time, to handle scenes with sparse points, we present a learnable target feature as the bridge to efficiently integrate target information from past frames. Moreover, we design a novel Instance Flow Head to transform dense point-level flow into instance-level motion, effectively aggregating local motion information to obtain global target motion. Finally, our method achieves competitive performance with improvements of 5.9% on the KITTI dataset and 2.9% on NuScenes. The code will be made publicly available soon. © 2024, CC BY.

关键词： Target tracking

来源：评论

学校读者我要写书评

暂无评论

Design and Validation of Pipeline Inspection robot based on Panoramic Vision

Design and Validation of Pipeline Inspection Robot based on ...

引用

Chinese Control and Decision Conference, CCDC

作者： Zixi Jia Shengming Li Bingze Li Faculty of Robot Science and Engineering Northeastern University Shenyang China

ISBN: (数字)9798350387780

ISBN: (纸本)9798350387797

Pipeline-type workpieces are prevalent in industrial production. However, manually inspecting the inner walls is an inefficient quality control activity that may pose safety risks to inspection personnel, especially for specialized workpieces with internal ramps. The complex internal structure further increases the difficulty of manual inspection. Therefore, designing a robot capable of navigating inside tube-shaped workpieces and stitching images of the inner walls into a single composite image is crucial for improving the efficiency of quality control processes for such workpieces. To meet this demand, we have designed a robot as an experimental platform capable of traveling and climbing ramps within pipes. The robot can handle a maximum climbing angle of 30 degrees and is equipped with four 180-degree wide-angle cameras arranged at 90-degree intervals, facilitating the capture of images around the pipe. Addressing the challenge of narrow environments inside pipes where conventional stitching algorithms may not be effectively applied, we have designed a stitching method capable of seamlessly stitching larger disparity images in narrow pipe environments. This algorithm optimizes the processes of feature point matching and optimal seam line selection, achieving higher-quality image stitching from the four cameras. The robot we designed will significantly enhance the efficiency of inspecting the inner walls of tube-shaped workpieces, particularly those with ramps.

关键词： Service robots Splicing Pipelines robot vision systems Quality control Production Inspection

来源：评论

学校读者我要写书评

暂无评论

Slot-Based Object-Centric Reinforcement Learning Algorithm

Slot-Based Object-Centric Reinforcement Learning Algorithm

引用

IEEE International Conference on Cyber Technology in Automation, Control, and Intelligent Systems

作者： Chao Chen Fei Wang Xinyao Wang Faculty of Robot Science and Engineering Northeastern University China Faculty of Robot Science and Engineering Northeastern University Shen Yang LiaoNing China Bussiness School Liaoning University

ISBN: (数字)9798331506056

ISBN: (纸本)9798331506063

With the development of reinforcement learning algorithms, it has become capable of handling a wide variety of complex tasks in simulated environments. However, applying reinforcement learning algorithms to real-world tasks still faces significant challenges. When facing vision-based tasks, these algorithms often embed images into a unified representation, which is effective for most simple tasks but inefficient when dealing with tasks involving interactions between multiple objects. This inefficiency arises from the lack of prior information about the world, neglecting the fundamental characteristic that the world is composed of multiple objects. This paper proposes an algorithm called SOCRL (Slot Based Object-Centric reinforcement learning), which enhances the efficiency of reinforcement learning algorithms in multi-object tasks by decoupling images into independent components. Experimental results demonstrate that our algorithm exhibits better generalization and higher sample efficiency compared to previous algorithms.

关键词： Training Image coding Filtering Semantics Reinforcement learning Interference Prediction algorithms Robustness Intelligent systems Faces

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：