检索结果-内蒙古大学图书馆

2022 IEEE International Conference on Unmanned systems, ICUS 2022

作者： Zheng, Haonan Zhao, Jiajia Tan, Wei School of Electronics and Information Northwestern Polytechnical University Shaanxi Xi'an710072 China Science and Technology on Complex System Control and Intelligent Agent Cooperation Laboratory Beijing100074 China

ISBN: (数字)9781665484565

ISBN: (纸本)9781665484565

In the field of artificial intelligence, neural network is one of the key technologies used for image classification and recognition. However, recent work has demonstrated that deep neural networks are easily attacked by adversarial examples to make misjudgments. Adversarial examples are almost indistin-guishable from normal examples and yet cannot be classified correctly by neural networks. The existence of adversarial examples is a major obstacle to the practical application and deployment of neural networks, so the research on adversarial defense algorithms is an important topic in the field of AI security. This paper proposes an adversarial example defense algorithm based on a triplet network and voting decision mechanism. Firstly, two neural networks with different structures are trained based on normal dataset. Secondly, the first network is fine-tuned using the adversarial examples generated by these two networks, resulting in a third neural network. Then, these three neural networks are used as sub-networks in parallel to construct a triplet network. Through adversarial training and differences in structures, the transferability of adversarial examples among the three sub-networks is weakened. Finally, the final classification result is obtained by majority voting, based on the parallel output results of the three sub-networks. Through the complementarity between these three sub-networks, the defense against adversarial examples is realized. The experimental results demonstrate the effectiveness of this algorithm. © 2022 IEEE.

关键词： Deep neural networks

来源：评论

学校读者我要写书评

暂无评论

A Comparative Study on Adversarial Attacks for Visible and Infrared Images

A Comparative Study on Adversarial Attacks for Visible and I...

引用

2022 IEEE International Conference on Unmanned systems, ICUS 2022

作者： Guo, Haiwang Tan, Wei Zhao, Jiajia School of Electronics and Information Northwestern Polytechnical University Shaanxi Xi'an710072 China Agent Cooperation Laboratory Science and Technology on Complex System Control and Intelligent Beijing100074 China

ISBN: (数字)9781665484565

ISBN: (纸本)9781665484565

In recent years, the booming development of artificial intelligence has driven the development and progress of society. With its excellent performance, deep learning has been widely used in many fields, including some security-sensitive fields such as unmanned systems, auto-driving and so on. Research has found that deep neural networks are vulnerable to adversarial samples, which can lead to misclassification of neural networks and can cause serious safety incidents. Hence, the study of adversarial samples has received increasing attention, and researchers have conducted studies from different perspectives, and relevant research results have emerged. However, most of these studies have focused on visible images, and few have studied infrared images, which play an important role in military and other important fields. In this paper, a comparative study on adversarial attacks for visible and infrared images is presented. A number of adversarial attack experiments are conducted on ResNet18 and VGG16 models for visible and infrared images. By comparing the experimental results, the characteristics of infrared adversarial samples are derived relative to those with visible adversarial samples. Then, a gradient attack method from a frequency domain perspective is proposed. The adversarial perturbation noise is added to the low frequency component of images. It is verified experimentally that the proposed method improves the transferability of infrared adversarial samples compared to some baseline methods. © 2022 IEEE.

关键词： Deep neural networks

来源：评论

学校读者我要写书评

暂无评论

Improved RRT Autonomous Exploration Method Based on Hybrid Clustering Algorithm

Improved RRT Autonomous Exploration Method Based on Hybrid C...

引用

International Conference on Autonomous Unmanned systems, ICAUS 2021

作者： Yin, Yizhen Ma, Hongjun Liang, Xinkai College of Information Science and Engineering Northeastern University Liaoning Shenyang China Science and Technology On Complex System Control and Intelligent Agent Cooperation Laboratory Beijing Electro-Mechanical Engineering Institute Beijing China

ISBN: (纸本)9789811694912

In autonomous robot exploration, it is very important that the robot can autonomously plan the exploration path of the unknown environment and create a map representation of the surrounding environment. In the robot's autonomous environment traversal method, one branch is the frontier-based detection method. This method consumes more and more computing resources as the map expands and is not suitable for large-scale environment detection. Another branch is the random exploration method based on multiple rapidly-exploring randomized trees. This method uses the edge of the tree to search for the boundary, but the selection of the next exploration point is very demanding. This paper proposes an exploration point selection strategy based on the hybrid clustering method, which speeds up the RRT exploration method. Simulation experiments show that the improved RRT exploration method proposed in this paper takes less time to explore and build maps, especially in a large-scale mapping environment, which effectively improves the completeness and efficiency of mapping. © 2022, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.

关键词： Cluster analysis

来源：评论

学校读者我要写书评

暂无评论

Active Disturbance Rejection Gain Scheduling control of Hypersonic Vehicle Based on Guardian Maps

Active Disturbance Rejection Gain Scheduling Control of Hype...

引用

Data Driven control and Learning systems (DDCLS)

作者： Tingyu Zhang Yongshuai Wang Mingwei Sun Mingrui Hao Zengqiang Chen College of Artificial Intelligence Nankai University Tianjin China Science and Technology on Complex System Control and Intelligent Agent Cooperation Laboratory Beijing China

ISBN: (数字)9798350361674

ISBN: (纸本)9798350361681

This paper presented an adaptive gain scheduling algorithm for hypersonic vehicles with wide envelopes. First, the linear parameter variable parameter (LPV) model of hypersonic vehicles is established in the flight envelope. Then, the active disturbance rejection controller is considered to cope with disturbances and uncertainties, and the control gain is scheduled using the guardian maps method to adapt to the wide envelope of velocity and height. Finally, numerical simulation results show good control performance and meet the required performance indexes in all operating conditions, demonstrating the effectiveness and robustness of the proposed method.

关键词： Uncertainty PI control Scheduling algorithms Atmospheric modeling Numerical simulation Scheduling Robustness

来源：评论

学校读者我要写书评

暂无评论

Lightweight Vision Transformer for Small Data Sets 38

Lightweight Vision Transformer for Small Data Sets

引用

38th Youth Academic Annual Conference of Chinese Association of Automation, YAC 2023

作者： Fang, Wenxing Su, Yanxu Institutes of Physical Science and Information Technology Anhui University Hefei China School of Artificial Intelligence Anhui University Hefei230601 China Southeast University Key Laboratory of Measurement and Control of Complex System of Engineering Ministry of Education Nanjing210096 China

ISBN: (纸本)9798350303636

With the rapid development of Transformers in the field of computer vision, models based on Transformers have become highly competitive architectures in the area. Although variants of Transformer models have achieved increasing accuracy on image classification tasks, the size of the training set and the number of parameters required by the models have increased dramatically. When dealing with small datasets, such models face problems such as overfitting and undergeneralization, leading to poor accuracy on the test set. We propose a new lightweight vision transformer (LVT) to address these issues. We reconstructed the backbone network, which learns the relationship between pixels through local window self-attention and global self-attention computation. We also use the attention pooling approach to fuse the token sequences generated by the backbone network more meticulously. We trained on the CIFAR-10 and CIFAR-100 datasets from scratch and compared them with a modern convolutional neural network. The experimental results show that LVT outperforms the modern convolutional neural network in terms of accuracy and efficiency. On the test set of CIFAR-10, we have obtained an accuracy of 96.83%, which indicates that our model can effectively solve the problems facing the training of small datasets and has a wide range of application prospects. © 2023 IEEE.

关键词： lightweight pooling self-attention windows

来源：评论

学校读者我要写书评

暂无评论

WATNet: A Wavelet-Aware Lightweight Hybrid Model for Fast Low-Light Enhancement 7

WATNet: A Wavelet-Aware Lightweight Hybrid Model for Fast Lo...

引用

7th Asian Conference on Artificial Intelligence technology, ACAIT 2023

作者： Wang, Qiangqiang Huang, Peiliang Li, Lei Han, Longfei Anhui University AHU-IAI Ai Joint Laboratory Hefei China Hefei Comprehensive National Science Center Institute of Artificial Intelligence Hefei China Northwestern Polytechnical University School of Automation Xi'an China Science And Technology On Complex System Control And Intelligent Agent Cooperation Laboratory Beijing China Beijing Technology And Business University Beijing China University of Science And Technology of China School of Information Science And Technology Hefei China

ISBN: (纸本)9798350359145

Low-light enhancement task is an essential component of computer low-level visual tasks, which involves processing images captured under dim lighting conditions to make them appear as if they were taken under normal illumination. Currently, deep neural networks have become the mainstream approach for image processing. However, recent works have devoted considerable efforts to designing high-performance models, which often come with high computational complexity and inference time, making real-time processing unfeasible. We observed that some convolutional methods are due to the need for deep layers which results in a large number of parameters. Moreover, enhancing details and removing noise in low-light images remains an open challenge. In order to solve the above problems, we propose a lightweight baseline that combines CNN and sparse grid attention transformer blocks to enable the model to capture a global receptive field at an early stage. Specifically, we propose a High-Frequency Wavelet-aware Block(HFWB) that focuses on processing high-frequency information in the wavelet domain to refine details and suppress noise. With a processing time of only 10.6ms, the performance of our model outperforms that of the current state-of-the-art lightweight models on benchmark low-light datasets. Compared to state-of-the-art models in the LOL dataset, our model achieves a reduction in inference time of over 90% and requires only about 1% of the FLOPS. © 2023 IEEE.

关键词： Deep neural networks

来源：评论

学校读者我要写书评

暂无评论

Estimation of smoke echo signal using single-photon lidar 13

Estimation of smoke echo signal using single-photon lidar

引用

13th International Conference on Information Optics and Photonics, CIOP 2022

作者： Zhang, Yinbo Li, Sining Sun, Jianfeng Jiang, Peng Zhou, Shuaijun Zhang, Xin Hou, Yue Zhang, Hailong National Key Laboratory of Science and Technology on Tunable Laser Institute of Opto-Electronic Harbin Institute of Technology Heilongjiang Harbin150001 China Science and Technology on Complex System Control and Intelligent Agent Cooperation Laboratory Beijing100074 China Beijing Aerospace Automatic Control Institute Beijing100854 China

ISBN: (纸本)9781510660632

The non-uniform distribution of smoke and laser spot seriously limits the imaging ability of single-photon lidar through smoke. To this end, based on the collision theory between photons and smoke particles, this paper establishes the time-domain distribution model of scattered photons (Gamma), which considers the lidar system parameters and the characteristics of smoke particles. The shape parameter and inverse scale parameter are defined as the maximum scattering number K and the average scattering number β, respectively. The indoor test effectively illustrates the estimation ability of the model for smoke. The parameter estimation results show that the average scattering number β and the maximum scattering number K increase linearly with the increase of attenuation coefficient. This study is of great significance for the suppression of smoke backscattering and is expected to improve the weather adaptability of single-photon lidar. © 2022 SPIE.

关键词： Backscattering

来源：评论

学校读者我要写书评

暂无评论

Object Detection and Multiple Objective Optimization Manipulation Planning for Underwater Autonomous Capture in Oceanic Natural Aquatic Farm

引用

Journal of Field Robotics 2025年第0期

作者： Hai, Huang Tao, Jiang Xinyu, Bian Hao, Zhou Xu, Yang Gang, Wang Hongde, Qin Xinyue, Han National Key Laboratory of Science and Technology on Underwater Vehicle Harbin Engineering University Harbin China State Key Laboratory of Management and Control for Complex System Institute of Automation Chinese Academy of Science Bei Jing China

Underwater autonomous capture operations offer significant potential for reducing labor and health risks in sea organism industries. This study presents a comprehensive solution for cross-domain underwater object detection and autonomous capture. A novel unsupervised domain adaptive learning method is proposed, integrating multiscale domain adaptive modules and attention mechanisms into a Faster Region-Convolutional Neural Network framework. This approach enhances feature alignment across diverse aquatic domains without parameter tuning. Additionally, an efficient, parameterless constrained multiobjective optimization algorithm is introduced for underwater autonomous mobile capture, integrating parameterized trajectory planning with innovative features, such as adaptive mutation strategies and constraint violation tolerance. The proposed approaches are extensively validated through simulations, tank experiments, and real-world oceanic trials in the Natural Aquatic Farm of Zhangzidao Island. Results demonstrate the system's robustness in complex underwater environments with varying currents, with experimental outcomes validating the accuracy and reliability of detection and capture capabilities. This research significantly advances autonomous underwater systems' capabilities in object detection and capture tasks, addressing complex challenges in realistic organism capture applications across diverse aquatic environments. © 2025 Wiley Periodicals LLC.

关键词： Health risks

来源：评论

学校读者我要写书评

暂无评论

Distance estimation of obstacles in monocular vision based on deep learning 8

Distance estimation of obstacles in monocular vision based o...

引用

8th Symposium on Novel Photoelectronic Detection technology and Applications

作者： Shan, Liu Ming, Li Long Hui, Wei Dong Beijing Electro-Mechanical Engineering Institute Science and Technology on Complex System Control and Intelligent Agent Cooperation Laboratory Beijing100074 China China Aerospace Science and Industry Corporation Limited Beijing100048 China

ISBN: (数字)9781510653122

ISBN: (纸本)9781510653115

Aiming at the forward-looking distance perception requirements of obstacles such as mountains, forests and trees during the ultra-low-altitude flight of subsonic aircraft, an obstacle distance estimation method based on deep learning is proposed. Firstly, the relationship between adjacent hyper-pixels is modeled by the conditional random fields, the neural network model is trained using the data set, and then the loss function of the pose parameter regression is designed, and the representation of the neural network model is input to the constructed loss function to optimize the weight of the regression model. Simulation experiments show that the proposed method can effectively perceive obstacles within 3km ahead, and the distance estimation error is not more than 20% of the relative distance, which can meet the obstacle avoidance requirements of most sudden obstacles during ultra-low-altitude flight. © 2022 SPIE

关键词： Vision

来源：评论

学校读者我要写书评

暂无评论

Research on Intelligent Credibility Evaluation Method of Aircraft HWIL Simulation system Based on complex Network 17th

Research on Intelligent Credibility Evaluation Method of Air...

引用

17th Chinese Intelligent systems Conference, CISC 2021

作者： Huapin, Geng Ni, Li Science and Technology on Complex System Control and Intelligent Agent Cooperation Laboratory Beijing China Beijing Electromechanical Engineering Institute Beijing China Beijing University of Aeronautics and Astronautics Beijing China

ISBN: (纸本)9789811663192

Simulation credibility is an effective means to measure the effectiveness and correctness of simulation, and it is the cornerstone of simulation. According to the uncertainty of physical effect model and the complex interaction of model information in aircraft HWIL simulation system, The idea of time-varying directed weighted network and uncertainty measurement is introduced into the process of credibility evaluation of simulation system based on complex network and credibility evaluation theory;the credibility evaluation model of time-varying topological structure based on uncertainty measurement of model information interaction is established to quantitatively evaluate the credibility of simulation system. Taking a typical HWIL simulation system as an example, the effectiveness of the credibility evaluation method of simulation system based on complex network is fully tested. © 2022, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.

关键词： Hardware-in-the-loop simulation

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：