检索结果-内蒙古大学图书馆

Deep side group sparse coding network for image denoising

IET image processing 2023年第1期17卷 1-11页

作者： Yin, Haitao Wang, Tianyou Nanjing Univ Posts & Telecommun Coll Automat Nanjing 210023 Peoples R China Nanjing Univ Posts & Telecommun Coll Artificial Intelligence Nanjing 210023 Peoples R China

Recently, deep learning has made significant progress in image denoising. However, most of existing deep learning based methods are purely data-driven, without considering the knowledge of image denoising. Moreover, the parameters of deep denoising network are not explainable. According to these issues, this paper proposes a deep side group sparse coding network for image denoising, named a side group sparse coding (SGSC)-Net. First, SGSC model for image denoising by exploiting prior information regarding the group sparse coefficients consistency is developed. Specifically, the side information is constructed as the weighted combination of intermediate estimations, and updated iteratively. Then, the optimisation solution of SGSC model is turned into a deep neural network using deep unfolding, that is, SGSC-Net. The computational path of SGSC-Net fully follows the iterations of optimisation solution, and consequently the network parameters are interpretable. Furthermore, the design of SGSC-Net employs the insight of SGSC denoising model. The experimental results on well-known datasets quantitatively and qualitatively demonstrate that SGSC-Net is competitive to existing deep unfolding-based and typical deep neural network-based methods.

关键词： SGSC-Net image denoising iterative methods side group sparse coding-Net deep learning based methods deep denoising network deep learning (artificial intelligence) Neural nets typical deep neural network-based methods Optical, image and video signal processing group sparse coefficients consistency deep side group SGSC model SGSC denoising model computer vision and image processing techniques

来源：评论

学校读者我要写书评

暂无评论

An automatic plant leaf stoma detection method based on YOLOv5

引用

IET image processing 2023年第1期17卷 67-76页

作者： Li, Xin Guo, Siyu Gong, Linrui Lan, Yuan Hunan Univ Coll Elect & Informat Engn Changsha 410082 Peoples R China

The stomata on the leaf surface are mainly responsible for the material exchange between the internal and external environments of the plant, a large number of methods have been proposed to automatically measure the distribution position and number of stomatal, but few methods could achieve both stomatal count and open/closed-state judgment. Therefore, this study proposes an automatic detection method for leaf stomatal morphology analysis based on an attention mechanism and deep learning. In order to obtain more stomatal feature information and send it to the network for learning, the proposed method adds a coordinate attention (CA) mechanism to the YOLOV5 backbone part. At the same time, in order to avoid the overfitting of the model during the training process, the authors added the training trick of label smoothing. Finally, the detection ability of the proposed method for stomata is verified on the broad bean leaves stomata dataset. The experimental results show that our method achieves a detection accuracy of 0.934 and an mAP of 0.968. By comparing with other state-of-the-art algorithms, the detection capability of our method has been significantly improved. The generalization of the model is verified on the wheat leaf stomatal dataset. The experimental results show that our method can achieve a detection accuracy of 0.894 and an mAP of 0.907.

关键词： feature extraction stomata dataset deep learning automatic plant leaf stoma detection method learning (artificial intelligence) leaf stomatal morphology analysis stomatal count distribution position Neural nets YOLOV5 backbone part Agriculture, forestry and fisheries computing internal environments computer vision and image processing techniques detection accuracy external environments automatic detection method wheat leaf stomatal dataset plant diseases object detection deep learning (artificial intelligence) Agriculture detection capability botany crops material exchange leaf surface coordinate attention mechanism stomatal feature information

来源：评论

学校读者我要写书评

暂无评论

Robust graph fusion and recognition framework for fingerprint and finger-vein

引用

IET BIOMETRICS 2023年第1期12卷 13-24页

作者： Wu, Zhitao Qu, Hongxu Zhang, Haigang Yang, Jinfeng Univ Sci & Technol Liaoning Sch Elect & Informat Engn Anshan Liaoning Peoples R China Shenzhen Polytech Shenzhen Guangdong Peoples R China

The human finger is the essential carrier of biometric features. The finger itself contains multi-modal traits, including fingerprint and finger-vein, which provides convenience and practicality for finger bi-modal fusion recognition. The scale inconsistency and feature space mismatch of finger bi-modal images are important reasons for the fusion effect. The feature extraction method based on graph structure can well solve the problem of feature space mismatch for the finger bi-modalities, and the end-to-end fusion recognition can be realised based on graph convolutional neural networks (GCNs). However, this fusion recognition strategy based on GCNs still has two urgent problems: first, lack of stable and efficient graph fusion method;second, over-smoothing problem of GCNs will lead to the degradation of recognition performance. A novel fusion method is proposed to integrate the graph features of fingerprint (FP) and finger-vein (FV). Furthermore, we analyse the inner relationship between the information transmission process and the over-smoothing problem in GCNs from an optimisation perspective, and point out that the differentiated information between neighbouring nodes decreases as the number of layers increases, which is the direct reason for the over-smoothing problem. A modified deep graph convolution neural network is proposed, aiming to alleviate the over-smoothing problem. The intuition is that the differentiated features of the nodes should be properly preserved to ensure the uniqueness of the nodes themselves. Thus, a constraint term to the objective function of the GCN is added to emphasise the differentiation features of the nodes themselves. The experimental results show that the proposed fusion method can achieve more satisfied performance in finger bi-modal biometric recognition, and the proposed constrained GCN can well alleviate the problem of over-smoothing.

关键词： biometric features feature extraction finger bi-modal biometric recognition Combinatorial mathematics Sensor fusion modified deep graph convolution neural network fusion recognition strategy feature space mismatch finger bi-modal fusion recognition urgent problems fusion effect differentiation features Neural nets finger bi-modal images image fusion computer vision and image processing techniques finger-vein convolutional neural nets multimodal traits deep learning (artificial intelligence) image recognition recognition performance graph theory stable graph fusion method recognition framework over-smoothing problem efficient graph fusion method graph features finger bi-modalities graph structure biometrics (access control) feature extraction method differentiated features end-to-end fusion recognition graph convolutional neural networks human finger GCNs

来源：评论

学校读者我要写书评

暂无评论

An effective model for the iris regional characteristics and classification using deep learning alex network

引用

IET image processing 2023年第1期17卷 227-238页

作者： Balashanmugam, Thiyaneswaran Sengottaiyan, Kumarganesh Kulandairaj, Martin Sagayam Hien Dang Sona Coll Technol Dept ECE Salem Tamil Nadu India Knowledge Inst Technol Dept ECE Salem Tamil Nadu India Karunya Inst Technol & Sci Dept ECE Coimbatore Tamil Nadu India Thuyloi Univ Fac Comp Sci & Engn Hanoi 100000 Vietnam

Iris biometrics is one of the fastest-growing technologies, and it has received a lot of attention from the community. Iris-biometric-based human recognition does not require contact with the human body. Iris is a combination of crypts, wolflin nodules, concentrated furrows, and pigment spots. The existing methods feed the eye image into deep learning network which result in improper iris features and certainly reduce the accuracy. This research study proposes a model to feed preprocessed accurate iris boundary into Alexnet deep learning neural network-based system for classification. The pupil centre and boundary are initially recorded and identified from the given eye images. The iris boundary and the centre are then compared for the identification using the reference pupil centre and boundary. The iris portion, exclusive feature of the pupil area is segmented using the parameters of multiple left-right point (MLRP) algorithms. The Alexnet deep learning multilayer networks 23, 24, and 25 are replaced according to the segmented iris classes. The remaining Alexnet layers are trained using the square gradient decay factor (GDF) in accordance with the iris features. The trained Alexnet iris is validated using suitable classes. The proposed system classifies the iris with an accuracy of 99.1%. The sensitivity, specificity, and F1-score of the proposed system are 99.68%, 98.36%, and 0.995. The experimental results show that the proposed model has advantages over current models.

关键词： preprocessed accurate iris boundary image segmentation feature extraction improper iris features human body wolflin nodules image classification given eye images iris-biometric-based human recognition trained Alexnet iris Alexnet deep learning multilayer networks 23, 24 segmented iris classes Neural nets computer vision and image processing techniques neural nets left-right point algorithms deep learning (artificial intelligence) image recognition pigment spots deep learning alex network deep learning network iris portion Alexnet deep learning neural network-based system remaining Alexnet layers iris biometrics biometrics (access control) iris recognition exclusive feature iris regional characteristics concentrated furrows pupil area reference pupil centre eye image

来源：评论

学校读者我要写书评

暂无评论

EDAfuse: A encoder-decoder with atrous spatial pyramid network for infrared and visible image fusion

引用

IET image processing 2023年第1期17卷 132-143页

作者： Nie, Cairen Zhou, Dongming Nie, Rencan Yunnan Univ Sch Informat Sci & Engn Kunming Yunnan Peoples R China Yunnan Normal Univ Sch Econ & Management Kunming Yunnan Peoples R China Southeast Univ Sch Automat Nanjing Peoples R China Yunnan Key Lab Intelligent Syst & Comp Kunming Yunnan Peoples R China

Infrared and visible images come from different sensors, and they have their advantages and disadvantages. In order to make the fused images contain as much salience information as possible, a practical fusion method, termed EDAfuse, is proposed in this paper. In EDAfuse, the authors introduce an encoder-decoder with the atrous spatial pyramid network for infrared and visible image fusion. The authors use the encoding network which includes three convolutional neural network (CNN) layers to extract deep features from input images. Then the proposed atrous spatial pyramid model is utilized to get five different scale features. The same scale features from the two original images are fused by our fusion strategy with the attention model and information quantity model. Finally, the decoding network is utilized to reconstruct the fused image. In the training process, the authors introduce a loss function with saliency loss to improve the ability of the model for extracting salient features from original images. In the experiment process, the authors use the average values of seven metrics for 21 fused images to evaluate the proposed method and the other seven existing methods. The results show that our method has four best values and three second-best values. The subjective assessment also demonstrates that the proposed method outperforms the state-of-the-art fusion methods.

关键词： different scale features feature extraction Sensor fusion original images input images encoding network atrous spatial pyramid network encoder-decoder decoding network Neural nets atrous spatial pyramid model termed EDAfuse Optical, image and video signal processing image fusion computer vision and image processing techniques practical fusion method convolutional neural nets state-of-the-art fusion methods infrared images object detection convolutional neural network layers deep learning (artificial intelligence) image recognition visible images fused image information quantity model visible image fusion fusion strategy infrared image fusion

来源：评论

学校读者我要写书评

暂无评论

Multi-modal fusion method for human action recognition based on IALC

引用

IET image processing 2023年第2期17卷 388-400页

作者： Zhang, Yinhuan Xiao, Qinkun Liu, Xing Wei, Yongquan Chu, Chaoqin Xue, Jingyun Xian Technol Univ Sch Mechatron Engn Xian Peoples R China Weinan Vocat & Tech Coll Sch Construct Engn Weinan Peoples R China Xian Technol Univ Sch Elect Informat Engn Xian 710021 Peoples R China CRRC Tangshan Co Ltd Tangshan Peoples R China

In occlusion and interaction scenarios, human action recognition (HAR) accuracy is low. To address this issue, this paper proposes a novel multi-modal fusion framework for HAR. In this framework, a module called improved attention long short-term memory (IAL) is proposed, which combines the improved SE-ResNet50 (ISE-ResNet50) with long short-term memory (LSTM). IAL can extract the video sequence features and the skeleton sequence features of human behaviour. To improve the performance of HAR at a high semantic level, the obtained multi-modal sequence features are fed into a couple hidden Markov model (CHMM), and a multi-modal IAL+CHMM method called IALC is developed based on a probability graph model. To test the performance of the proposed method, experiments are conducted on the HMDB51, UCF101, Kinetics 400k, and ActivityNet datasets, and the obtained recognition accuracy are 86.40%, 97.78%, 81.12%, and 69.36% on the four datasets, respectively. The experimental results show that when the environment is complex, the proposed multi-modal fusion method for HAR based on the IALC can achieve more accurate target recognition results.

关键词： feature extraction recurrent neural nets video signal processing image representation novel multimodal fusion framework skeleton sequence features multimodal sequence features HAR occlusion computer vision and image processing techniques human action recognition accuracy image recognition image sequences multimodal fusion method Video signal processing multimodal IAL+CHMM method short-term memory probability video sequence features ISE-ResNet50 interaction scenarios accurate target recognition results IALC improved SE-ResNet50 image motion analysis human behaviour hidden Markov models Markov processes

来源：评论

学校读者我要写书评

暂无评论

Brain tumour detection in magnetic resonance imaging using Levenberg-Marquardt backpropagation neural network

引用

IET image processing 2023年第1期17卷 88-103页

作者： Ghahramani, Marzieh Shiri, Nabiollah Islamic Azad Univ Shiraz Branch Dept Elect Engn Shiraz *** Iran

Magnetic resonance imaging (MRI) is a high-quality medical image that is used to detect brain tumours in a complex and time-consuming manner. In this study, a back propagation neural network (BPNN) along with the Levenberg-Marquardt algorithm (LMA) is proposed to classify MRIs and diagnose brain tumours in a simple and fast process. The BPNN has 10 neurons in the hidden layer, and the default function of the feedforward feeds is mean squared error (MSE). The LMA is optimized as a multivariable adaptive approach and considerably decreases the MSE of the BPNN, so the errors of the tumour classification are diminished. The proposed method follows four steps including preprocessing, skull removal, feature extraction, and classification. The input MRIs are converted to greyscale, resized, and thresholding is performed in the preprocessing step and followed by skull removal. Morphological operations of closing, opening, and dilation are used to segment abnormal areas in the MRIs, and the opening operator recognizes the tumour more accurately. Using statistical analysis and a grey-level co-occurrence matrix (GLCM) 12 features are extracted from the MRIs and used as the inputs of the BPNN. To evaluate the proposed method, 670 normal and 670 abnormal brain MRIs are used as input data, and the classification is performed in 0.494 s. The accuracy, sensitivity, specificity, precision, dice, recall, and MSE are 98.7%, 97.61%, 99.7%, 97.61%, 98.6%, 97.61%, and 0.005, respectively. The approach is accurate and fast for medical images classification.

关键词： medical image classification image segmentation feature extraction tumour classification default function Biology and medical computing input MRIs high-quality medical image image classification biomedical MRI Medical magnetic resonance imaging and spectroscopy multivariable adaptive approach brain tumour detection Biophysics of neurophysiological processes brain tumour diagnosis Neural nets backpropagation mean squared error opening operator greyscale computer vision and image processing techniques Biomedical magnetic resonance imaging and spectroscopy statistical analysis grey-level cooccurrence matrix neural nets skull removal hidden layer preprocessing step image recognition MSE preprocessing removal Levenberg-Marquardt algorithm magnetic resonance imaging image thresholding abnormal brain MRIs BPNN morphological operations brain LMA Other topics in statistics Patient diagnostic methods and instrumentation tumours Levenberg-Marquardt backpropagation neural network medical image processing Probability theory, stochastic processes, and statistics

来源：评论

学校读者我要写书评

暂无评论

image super-resolution based on self-similarity generative adversarial networks

引用

IET image processing 2023年第1期17卷 157-165页

作者： Wang, Shuang Sun, Zhengxing Li, Qian Nanjing Univ State Key Lab Novel Software Technol Nanjing 210023 Peoples R China Jiangsu Vocat Inst Commerce Nanjing Peoples R China Natl Univ Def Technol Coll Meteorol & Oceanog Changsha Peoples R China

Self-attention has been successfully leveraged for long-range feature-wise similarities in deep learning super-resolution (SR) methods. However, most of the SR methods only explore the features on the original scale, but do not take full advantage of self-similarities features on different scales especially in generative adversarial networks (GAN). In this paper, self-similarity generative adversarial networks (SSGAN) are proposed as the SR framework. The framework establishes the multi-scale feature correlation by adding two modules to the generative network: downscale attention block (DAB) and upscale attention block (UAB). Specifically, DAB is designed to restore the repetitive details from the corresponding downsampled image, which achieves multi-scale feature restoration through self-similarity. And UAB improves the baseline up-sampling operations and captures low-resolution to high-resolution feature mapping, which enhances the cross-scale repetitive features to reconstruct the high-resolution image. Experimental results demonstrate that the proposed SSGAN achieve better visual performance especially in the similar pattern details.

关键词： Optical, image and video signal processing computer vision and image processing techniques Neural nets

来源：评论

学校读者我要写书评

暂无评论

THE RAPID RISE OF AI ART

引用

Engineering and Technology 2023年第2期18卷 20-25页

作者： Cousins, Stephen

GENERATIVE AI art has exploded onto the scene over the past few months through advanced online platforms like DALL-E2, Midjourney and Stable Diffusion, which enable anyone with access to a smartphone or PC to create h... 详细信息

关键词： Spatial and pictorial databases image processing Knowledge based systems Humanities computing artificial intelligence visual databases Optical, image and video signal processing computer vision and image processing techniques art

来源：评论

学校读者我要写书评

暂无评论

Long short-distance topology modelling of 3D point cloud segmentation with a graph convolution neural network

引用

IET computer vision 2023年第3期17卷 251-264页

作者： Zhang, Wen Jing Su, Song Zhi Hong, Qing Qi Wang, Bei Zhan Sun, Li Xiamen Univ Sch Informat Xiamen Peoples R China Univ Sheffield Dept Comp Sci Sheffield England Xiamen Univ Sch Informat Xiamen 361005 Peoples R China

3D point cloud segmentation is a non-trivial problem due to its irregular, sparse, and unordered data structure. Existing methods only consider structural relationships of a 3D point and its spatial neighbours. However, the inner-point interactions and long-distance context of a 3D point cloud have been less investigated. In this study, we propose an effective plug-and-play module called the Long Short-Distance Topologically Modelled (LSDTM) Graph Convolutional Neural Network (GCNN) to learn the underlying structure of 3D point clouds. Specifically, we introduce the concept of subgraph to model the contextual-point relationships within a short distance. Then the proposed topology can be reconstructed by recursive aggregation of subgraphs, and importantly, to propagate the contextual scope to a long range. The proposed LSDTM can parse the point cloud data with maximisation of preserving the geometric structure and contextual structure, and the topological graph can be trained end-to-end through a seamlessly integrated GCNN. We provide a case study of triple-layer ternary topology and experimental results on ShapeNetPart, Stanford 3D Indoor Semantics and ScanNet datasets, indicating a significant improvement on the task of 3D point cloud segmentation and validating the effectiveness of our research.

关键词： image segmentation feature extraction solid modelling Combinatorial mathematics graph convolution neural network data structures learning (artificial intelligence) contextual-point relationships 3D point cloud segmentation contextual structure Neural nets unordered data structure computational geometry Long Short-Distance Topologically Modelled Graph Convolutional Neural Network computer vision and image processing techniques convolutional neural nets irregular data structure Computational geometry long-distance context graph theory point cloud data short-distance topology modelling sparse, data structure File organisation inner-point interactions Graphics techniques geometric structure

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

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

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：