检索结果-内蒙古大学图书馆

arXiv 2022年

作者： Zhang, Tong Gao, Peng Dong, Hao Zhuang, Yin Wang, Guanqun Zhang, Wei Chen, He Beijing Key Laboratory of Embedded Real-time Information Processing Technology Beijing Institute of Technology Beijing100081 China The Shang Hai AI laboratory Shanghai100024 China Peking University Beijing100871 China

Currently, under supervised learning, a model pretrained by a large-scale nature scene dataset and then fine-tuned on a few specific task labeling data is the paradigm that has dominated the knowledge transfer learning. It has reached the status of consensus solution for task-aware model training in remote sensing domain (RSD). Unfortunately, due to different categories of imaging data and stiff challenges of data annotation, there is not a large enough and uniform remote sensing dataset to support large-scale pretraining in RSD. Moreover, pretraining models on large-scale nature scene datasets by supervised learning and then directly fine-tuning on diverse downstream tasks seems to be a crude method, which is easily affected by inevitable labeling noise, severe domain gaps and task-aware discrepancies. Thus, in this paper, considering the self-supervised pretraining and powerful vision transformer (ViT) architecture, a concise and effective knowledge transfer learning strategy called ConSecutive PreTraining (CSPT) is proposed based on the idea of not stopping pretraining in natural language processing (NLP), which can gradually bridge the domain gap and transfer knowledge from the nature scene domain to the RSD. The proposed CSPT also can release the huge potential of unlabeled data for task-aware model training. Finally, extensive experiments are carried out on twelve datasets in RSD involving three types of downstream tasks (e.g., scene classification, object detection and land cover classification) and two types of imaging data (e.g., optical and SAR). The results show that by utilizing the proposed CSPT for task-aware model training, almost all downstream tasks in RSD can outperform the previous method of supervised pretraining-then-fine-tuning and even surpass the state-of-the-art (SOTA) performance without any expensive labeling consumption and careful model design. Copyright © 2022, The Authors. All rights reserved.

关键词： Supervised learning

来源：评论

学校读者我要写书评

暂无评论

Radar range Extended target detection using Prototypical Network

Radar range Extended target detection using Prototypical Net...

引用

Signal, information and Data processing (ICSIDP), IEEE International Conference on

作者： Wenhao Ma Mingchen Yuan Xinyang Wang Liang Zhang Yanhua Wang Radar Technology Research Institute Beijing Institute of Technology Beijing China Zhengzhou Research Institute Beijing Institute of Technology Zhengzhou China Xi'an Modern Control Technology Research Institute Xi'an China Electromagnetic Sensing Research Center of CEMEE State Key Laboratory Beijing Institute of Technology Beijing China Beijing Key Laboratory of Embedded•Real-Time Information: Processing Technology Beijing China Beijing Institute of Technology Chongqing Innovation Center Chongqing China Advanced Technology Research Institute Beijing Institute of Technology Jinan China

ISBN: (数字)9798331515669

ISBN: (纸本)9798331515676

In radar ground stationary or slow-moving target detection, the target detection algorithm based on High-Resolution Range Profile (HRRP) is an important technical approach. Traditional target detection algorithms based on HRRP are difficult to apply in target detection tasks under multiple terrains. To this end, this paper proposes a feature detection algorithm based on deep learning for target polarized HRRP. This paper studies an end-to-end target detection algorithm based on prototypical networks. This algorithm utilizes prototypical networks to extract deep features from input HRRP and constructs the feature distribution of clutter. Then, it calculates the distance between the features of the input test sample and the clutter feature center, and compares it with the threshold calculated based on the false alarm probability to complete the detection. Under the backgrounds of bare soil and highway, the algorithm improves by 22% compared to the traditional energy detection algorithm for two types of vehicle targets at a signal-to-clutter ratio (SCR) of 0dB.

关键词： Road transportation Radar detection Object detection Radar Soil Probability Feature extraction Classification algorithms Clutter Signal to noise ratio

来源：评论

学校读者我要写书评

暂无评论

Multi-scale object detection by top-down and bottom-up feature pyramid network

引用

Journal of Systems Engineering and Electronics 2019年第1期30卷 1-12页

作者： ZHAO Baojun ZHAO Boya TANG Linbo WANG Wenzheng WU Chen School of Information and Electronics Beijing Institute of Technology Beijing Key Laboratory of Embedded Real-time Information Processing Technology Beijing Institute of Technology

While moving ahead with the object detection technology, especially deep neural networks, many related tasks, such as medical application and industrial automation, have achieved great success. However, the detection of objects with multiple aspect ratios and scales is still a key problem. This paper proposes a top-down and bottom-up feature pyramid network(TDBU-FPN),which combines multi-scale feature representation and anchor generation at multiple aspect ratios. First, in order to build the multi-scale feature map, this paper puts a number of fully convolutional layers after the backbone. Second, to link neighboring feature maps, top-down and bottom-up flows are adopted to introduce context information via top-down flow and supplement suboriginal information via bottom-up flow. The top-down flow refers to the deconvolution procedure, and the bottom-up flow refers to the pooling procedure. Third, the problem of adapting different object aspect ratios is tackled via many anchor shapes with different aspect ratios on each multi-scale feature map. The proposed method is evaluated on the pattern analysis, statistical modeling and computational learning visual object classes(PASCAL VOC)dataset and reaches an accuracy of 79%, which exhibits a 1.8% improvement with a detection speed of 23 fps.

关键词： convolutional neural network (CNN) feature pyramid network (FPN) object detection deconvolution.

来源：评论

学校读者我要写书评

暂无评论

Lightweight CNN for HRRP Recognition Based on Attention Mechanism Structured Pruning

Lightweight CNN for HRRP Recognition Based on Attention Mech...

引用

IEEE International Conference on Radar

作者： Yanhua Wang Zhilong Zhang Mingchen Yuan Jiandong Liao Liang Zhang Radar Research Lab School of Information and Electronics Beijing Institute of Technology Beijing China Electromagnetic Sensing Research Center of CEMEE State Key Lab Beijing Institute of Technology Beijing China Beijing Key Laboratory of Embedded Real-time Information Processing Technology Beijing China Chongqing Innovation Center Beijing Institute of Technology Chongqing China Advanced Technology Research Institute Beijing Institute of Technology Shandong China

Deep convolutional neural network (CNN) has been widely investigated for radar target high resolution range profile (HRRP) recognition. However, the deep structures of CNN require high storage and computational capabilities, thus restricting its applications with limited resources. In this paper, we design a lightweight CNN model with structure pruning based on the channel-wise attention mechanism. Specifically, the attention value is used to represent the importance of the filters in CNN. Furthermore, Greedy strategy and fine-tuning are adopted in the pruning process to minimize the loss of model performance. The results of public dataset show that the recognition accuracy of the proposed method decreases by less than 0.13% when the pruning ratio is 80%.

关键词：

来源：评论

学校读者我要写书评

暂无评论

An Adaptive Phase Noise Filtering Approach For Multi-Frequency Interferometric SAR

An Adaptive Phase Noise Filtering Approach For Multi-Frequen...

引用

IEEE International Conference on Radar

作者： Zhen Wang Zegang Ding Yan Wang Xinnong Ma Linghao Li Tao Zeng School of Information and electronics Beijing Institute of Technology Beijing China Beijing Key Laboratory of Embedded Real-time Information Processing Technology

ISBN: (纸本)9781665468893

Multi-frequency (MF) interferometric SAR (InSAR) provides an effective way for the surveying of steep terrains. Phase noise filtering is a key step in MF-InSAR processing which determines the quality of the interferograms and affects the accuracy of inversed elevation. Different from the traditional filtering method of single frequency (SF) InSAR, an adaptive filtering approach for MF-InSAR based on MF phase fusion is proposed in this paper. The main contributions of this method is summarized into two aspects. First, the local fringe frequency estimation (LFFE) is extended from SF to MF case. Second, the non-linear model is introduced in the adaptive phase filtering in the local window based on MF-LFFE to adapt to complex terrains. The computer simulation and the real dual-frequency airborne experiment validate the proposed approach.

关键词： Phase noise Adaptation models Filtering Computer simulation Computational modeling Atmospheric modeling Estimation

来源：评论

学校读者我要写书评

暂无评论

A Novel Image Recognition Method Based on CNN Algorithm 5th

A Novel Image Recognition Method Based on CNN Algorithm

引用

5th International Conference on Smart Computing and Communication, SmartCom 2020

作者： Guo, Yuwen Zhang, Xin Yang, Qi Guo, Hong College of Computer Science and Technology Wuhan University of Science and Technology Wuhan China Hubei Province Key Laboratory of Intelligent Information Processing and Real-Time Industrial System Beijing China

ISBN: (纸本)9783030747169

Handwritten digital image recognition is a special problem in digital recognition, which is much more difficult due to uncertainties such as writer and writing environment. But the accuracy of handwritten digital recognition is particularly important in some critical areas such as public security and finance. In this paper, we used machine learning and deep learning technology, to identify MNIST handwritten digital data sets. And the support vector machine and convolutional neural network were used to image recognition, among them, the support vector machine compares the accuracy of the identification of the two kernel functions. Convolutional neural networks are improved based on the classic LeNet-5 model. The Experiments result shown that the identification accuracy of Gaussian kernel function model and linear kernel function model is 94.03% and 90.47% for support vector machines, respectively. In contrast, Gaussian kernel function performs better, but the overall accuracy is low. The best performance is based on the classic convolutional neural network LeNet-5 improved model, the test set recognition rate can reach 99.05%. © 2021, Springer Nature Switzerland AG.

关键词： Support vector machines

来源：评论

学校读者我要写书评

暂无评论

Distorted SAR target recognition with virtual SVM and AP-HOG feature 5

Distorted SAR target recognition with virtual SVM and AP-HOG...

引用

IET International Radar Conference 2020, IET IRC 2020

作者： Yao, Di Liu, Zhenyuan Li, Feng Li, Yang Radar Research Laboratory School of Information and Electronics Beijing Institute of Technology Beijing100081 China Beijing Key Laboratory of Embedded Real-Time Information Processing Technology Beijing100081 China

ISBN: (纸本)9781839535635

Synthetic aperture radar (SAR) plays an indispensable role in national defence and civil fields due to its unique advantages. Although many methods have been proposed, the recognition to SAR images still faces many difficulties, especially to distorted images. This paper first improves the morphological methods attribute profiles (APs) which is widely employed with optical, and hyperspectral images in particular. We combine the APs method with histogram of oriented gradients (HOG) feature (called AP-HOG) to make it suitable for SAR images. Then, the SVM model is used to identify the labelled samples (i.e., Support Vectors) that are closest to the separating hyperplane with maximum margin. We generate different degrees of distorted images based on support vectors, and use the distorted image and the support vector as a new training set for retraining to improve the recognition effect of the distorted image. Finally, we perform experimental verification based on the MSTAR dataset, which proves the effectiveness of the proposed method. © 2020 Institution of Engineering and technology. All rights reserved.

关键词： Synthetic aperture radar

来源：评论

学校读者我要写书评

暂无评论

SAR target recognition using feature extraction of multi-scale monogenic components 5

SAR target recognition using feature extraction of multi-sca...

引用

IET International Radar Conference 2020, IET IRC 2020

作者： Li, Feng Yao, Weijun Li, Yang Radar Research Laboratory School of Information and Electronics Beijing Institute of Technology Beijing100081 China Beijing Key Laboratory of Embedded Real-Time Information Processing Technology Beijing100081 China

ISBN: (纸本)9781839535635

In this paper, we combine the monogenic signal with the histogram of oriented gradient (HOG) feature, and propose a new low-level feature: Monogenic-HOG (MG-HOG);then use the bag of words (BoW) model to construct the middle-level feature. Most of the existed features are directly extracted from grayscale synthetic aperture radar (SAR) images. As an extended representation of analytic signals in high-dimensional space, the monogenic signal has excellent characteristics such as rotation and scale invariance in the potential extraction process, and it is also valuable in the image field. However, in the past related work, it is mostly used directly as the feature representation of the image. After experimental validation on the moving and stationary target acquisition and recognition (MSTAR) data set, the proposed feature extraction method has better performance than a number of existed methods. © 2020 Institution of Engineering and technology. All rights reserved.

关键词： Feature extraction

来源：评论

学校读者我要写书评

暂无评论

SAR target recognition combining class information and classification hyperplane 5

SAR target recognition combining class information and class...

引用

IET International Radar Conference 2020, IET IRC 2020

ISBN: (纸本)9781839535635

Automatic target recognition (ATR) in synthetic aperture radar (SAR) images plays an important role. An important step in SAR target recognition is to transform the raw data to a feature space with the property of within-class compactness and between-class separation. However, due to the problem of target aspect angle sensitivity in SAR images, the difference of some targets belonging to the same class in different poses is greater than the difference of targets belonging to different class in similar poses, which often causes misjudgement. To solve this problem, we first use the twin support vector machine (TWSVM) to obtain the classification hyperplane of each class. Then, based on the obtained hyperplane, the projection matrix is constructed by fusing the prior class information of the target. In this way, the sample can be close to the classification hyperplane of the same class and distant from the classification hyperplane of a different class. This algorithm can improve TWSVM's ability to classify SAR images. Experiments are conducted on the moving and stationary target acquisition and recognition (MSTAR) database. The results verify the effectiveness of the proposed algorithm. © 2020 Institution of Engineering and technology. All rights reserved.

关键词： Synthetic aperture radar

来源：评论

学校读者我要写书评

暂无评论

A Nonlinear Clutter Map Detector for Rockfall Detection under Moving Debris and Dust Clutter

A Nonlinear Clutter Map Detector for Rockfall Detection unde...

引用

Signal, information and Data processing (ICSIDP), IEEE International Conference on

作者： Ruiqi Lin Weiming Tian Youwang Chen Delin Fang Ningyuan Chang Yunkai Deng Radar Technology Research Institute Beijing Institute of Technology Beijing China Beijing Key Laboratory of Embedded Realtime Information Processing Technology Beijing Institute of Technology Beijing China School of Information and Electronics Beijing Institute of Technology Beijing China Chongqing Innovation Center Beijing Institute of Technology Chongqing China Advanced Technology Research Institute Beijing Institute of Technology Jinan China

ISBN: (数字)9798331515669

ISBN: (纸本)9798331515676

The rockfall detection requires the Doppler radar to detect falling rocks amidst strong and heterogeneous ground clutter. Clutter map detectors are employed to provide robust detection in such scenario. In addition to ground clutter, falling rock causes moving debris and dust clutter, whose intensity has abrupt changes and causes false alarms. This paper proposes a nonlinear clutter map detector to handle such clutter. Employing a nonlinear transform of the echo power, this detector generates a maximum and a minimum sequence which characterize the intensity of abrupt clutter and stable clutter, respectively. It provides stable performance in slow-changing ground clutter and abrupt moving debris and dust clutter. Field data shows that the proposed nonlinear iterative clutter map can reduce 76.2% of false alarms caused by debris dust clutter, without affecting the detection of falling rocks.

关键词： Detectors Transforms Rocks Data processing Doppler radar Iterative methods Clutter

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：