Article
Nuclear Science & Technology
Hong Xu, Tao Tang, Baorui Zhang, Yuechan Liu
Summary: This paper proposes a CNN-SVM based methodology for two-phase flow regime identification, which combines the feature extraction capability of CNN and the classification accuracy of SVM. Experimental results show that the proposed method achieves higher classification accuracy and robustness.
PROGRESS IN NUCLEAR ENERGY
(2022)
Article
Forestry
Xiaobo Sun, Lin Xu, Yufeng Zhou, Yongjun Shi
Summary: In recent years, the automatic recognition of tree species based on images taken by digital cameras has been widely used, but there are still problems such as insufficient image acquisition and low recognition accuracy. This study built an image dataset of 21 tree species and used features of twigs and leaves for classification. By combining CNN models with classifiers, the accuracy of tree species recognition was improved. The experiment also analyzed the distribution of feature regions using Grad-CAM heatmap, showing the effectiveness of the proposed method.
Article
Engineering, Multidisciplinary
Tian Han, Longwen Zhang, Zhongjun Yin, Andy C. C. Tan
Summary: This paper combines CNN and SVM for bearing fault diagnosis, improving the model's generalization ability and accuracy. Experimental results show the system has advantages of less time-consuming, high accuracy, and strong generalization ability.
Article
Chemistry, Multidisciplinary
Ersin Sahin, Hueseyin Yuece
Summary: This study introduces a novel approach of using graph-based machine learning to detect leaks in pipelines, providing a theoretical contribution to the field. The generated datasets will serve as a foundation for further research and offer various benefits. The study's findings have significant implications for water resource management and environmental protection, as efficient utilization of water resources is crucial in today's society. The proposed Graph Convolutional Neural Network (GCN) model shows great potential in accurately predicting water leaks in pipeline systems, outperforming the traditional Support Vector Machine (SVM) model with a performance rate of 94%.
APPLIED SCIENCES-BASEL
(2023)
Article
Engineering, Electrical & Electronic
Caiyun Ma, Shoushui Wei, Tongshuai Chen, Jingquan Zhong, Zhenhua Liu, Chengyu Liu
Summary: This study proposed three methods for atrial fibrillation (AF) diagnosis, with the third method achieving the highest detection accuracy across different databases and experimental conditions, demonstrating high accuracy and reliable recognition for AF events.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2021)
Article
Engineering, Electrical & Electronic
Jie Li, Hongli He, Haibo He, Lusi Li, Yi Xiang
Summary: This article proposes an end-to-end framework for evaluating the health of bridges by exploring features and correlations of multiple monitoring factors, utilizing hierarchical learning structure with multiple convolutional, pooling, and dense layers to capture rich information and designing particular neural networks for each bridge factor to improve feature learning ability. Additionally, a classification scheme with multiple fully connection layers and support vector machines is designed to achieve higher evaluation accuracy, with experiments validating the superiority of the proposed model on real-world monitoring data of a specific bridge.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2021)
Article
Computer Science, Artificial Intelligence
Fatemeh Naiemi, Vahid Ghods, Hassan Khalesi
Summary: Automatic text detection and recognition in real-life images is crucial for various applications, and this paper presents a convolutional neural network-based pipeline to improve efficiency in this area. The utilization of pre-trained networks, as well as new optimized layers, demonstrate superior performance in detecting curved and vertical texts.
EXPERT SYSTEMS WITH APPLICATIONS
(2021)
Article
Remote Sensing
Anni Wang, PengLin Zhang
Summary: This research introduces a cooperative neural network that combines multi-task segmentation and graph convolution to improve the extraction of buildings. By strengthening boundaries and strategically selecting key points, this method effectively enhances the extraction of buildings.
INTERNATIONAL JOURNAL OF REMOTE SENSING
(2023)
Article
Automation & Control Systems
Shuqiang Wang, Xiangyu Wang, Yanyan Shen, Bing He, Xinyan Zhao, Prudence Wing-Hang Cheung, Jason Pui Yin Cheung, Keith Dip-Kei Luk, Yong Hu
Summary: Assessment of skeletal maturity is crucial for clinicians to make treatment decisions, but using machine learning for this task is challenging. In this article, an ensemble-based deep learning approach is proposed to automatically assess the maturity of the radius and ulna from left-hand X-ray images. Experimental results demonstrate the effectiveness of the proposed model.
IEEE TRANSACTIONS ON SYSTEMS MAN CYBERNETICS-SYSTEMS
(2022)
Article
Meteorology & Atmospheric Sciences
Zhongrui Wang, Lili Lei, Jeffrey L. Anderson, Zhe-Min Tan, Yi Zhang
Summary: Two convolutional neural network-based localization methods are proposed in this article, which can better capture the structures of the Kalman gain and generate improved analyses and forecasts in cycling assimilations.
JOURNAL OF ADVANCES IN MODELING EARTH SYSTEMS
(2023)
Article
Chemistry, Multidisciplinary
Anum Rauf, Aqsa Kiran, Malik Tahir Hassan, Sajid Mahmood, Ghulam Mustafa, Moongu Jeon
Summary: Heart-related diseases, especially caused by hypertension, are major causes of fatalities worldwide. Extracting bioactive peptides from natural food sources can provide potential alternatives to existing drugs with fewer side effects. In-silico approaches have been proven to be effective in identifying antihypertensive peptides, saving time and money. The proposed deep learning-based methodology combining CNN and SVM classifiers for feature extraction yields high accuracy in identifying antihypertensive peptides, outperforming existing state-of-the-art methods.
APPLIED SCIENCES-BASEL
(2021)
Article
Chemistry, Multidisciplinary
Dan Yang, Mengzhou Xiong, Tao Wang, Guangtao Lu
Summary: This paper proposes a non-destructive method for detecting pipeline ponding using percussion acoustic signals and a convolution neural network (CNN). The experimental results show that this method has good recognition performance and high application potential.
APPLIED SCIENCES-BASEL
(2022)
Article
Engineering, Environmental
Pengqian Liu, Changhang Xu, Jing Xie, Mingfu Fu, Yifei Chen, Zichen Liu, Zhiyuan Zhang
Summary: This study proposes a convolutional neural network-based transfer learning method for pipeline leakage detection under multiple working conditions. The results show that the proposed feature-based CNN-TL method outperforms parameter-based CNN-TL and traditional CNN methods, achieving accurate detection of pipeline leaks under various working conditions.
PROCESS SAFETY AND ENVIRONMENTAL PROTECTION
(2023)
Article
Agronomy
Yun Peng, Shengyi Zhao, Jizhan Liu
Summary: This study proposes a rapid and accurate grape leaf disease detection method based on deep feature fusion and an SVM classifier. Experimental results show that the proposed method can achieve better classification performance than using a single type of deep feature, with a training time of less than 1 second.
Article
Computer Science, Artificial Intelligence
Sameena Pathan, P. C. Siddalingaswamy, Tanweer Ali
Summary: The COVID-19 pandemic has had a devastating global impact due to the lack of automated diagnostic tools and medical resources. This study proposes the use of a Convolutional Neural Network model to assist in diagnosing COVID-19 affected chest X-ray scans, achieving high classification accuracy, which can be used in clinical and remote settings for faster screening of affected patients.
APPLIED SOFT COMPUTING
(2021)