Pub Date : 2022-08-01DOI: 10.1109/SmartIoT55134.2022.00025
Qingyu Zhang, Chunyan Wei, Qingxia Li, Xiaosen Tian, Chuanpeng Li
Compared with other sensors, high-quality depth estimation based on monocular camera has strong competitiveness and widespread application in intelligent transportation, etc. Although the barrier of training has been greatly lowered by unsupervised learning, most related works are still based on convolutional neural networks (CNNs) that suffer from unbridgeable gaps in the full-stage global information and high-resolution features while extracting multi-scale features. To break this predicament, we attempt to introduce vision transformer. However, the vision transformer with large sequence length due to image embedding brings great challenges to the computational cost. Thus, this work proposes a new pure transformer backbone named pooling pyramid vision transformer (PPViT), simultaneously shrinking out multi-scale features and reducing sequence length used for attention operation. Then, we provide two backbone settings including PPViT10 and PPViT18 whose number of parameters are close to the common ResNet18 and ResNet50, respectively. The experiments on KITTI dataset demonstrate that our work show a great potentiality of improving the capability of model and produce superior results to the previous CNN-based works. Equally important, we have lower latency than the related transformer-based work.
{"title":"Pooling Pyramid Vision Transformer for Unsupervised Monocular Depth Estimation","authors":"Qingyu Zhang, Chunyan Wei, Qingxia Li, Xiaosen Tian, Chuanpeng Li","doi":"10.1109/SmartIoT55134.2022.00025","DOIUrl":"https://doi.org/10.1109/SmartIoT55134.2022.00025","url":null,"abstract":"Compared with other sensors, high-quality depth estimation based on monocular camera has strong competitiveness and widespread application in intelligent transportation, etc. Although the barrier of training has been greatly lowered by unsupervised learning, most related works are still based on convolutional neural networks (CNNs) that suffer from unbridgeable gaps in the full-stage global information and high-resolution features while extracting multi-scale features. To break this predicament, we attempt to introduce vision transformer. However, the vision transformer with large sequence length due to image embedding brings great challenges to the computational cost. Thus, this work proposes a new pure transformer backbone named pooling pyramid vision transformer (PPViT), simultaneously shrinking out multi-scale features and reducing sequence length used for attention operation. Then, we provide two backbone settings including PPViT10 and PPViT18 whose number of parameters are close to the common ResNet18 and ResNet50, respectively. The experiments on KITTI dataset demonstrate that our work show a great potentiality of improving the capability of model and produce superior results to the previous CNN-based works. Equally important, we have lower latency than the related transformer-based work.","PeriodicalId":422269,"journal":{"name":"2022 IEEE International Conference on Smart Internet of Things (SmartIoT)","volume":"23 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131409307","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-08-01DOI: 10.1109/SmartIoT55134.2022.00034
Jinhao Liu, Xiaofan Yu, T. Simunic
Recent years have witnessed a significant increase in deploying lightweight machine learning (ML) on embedded systems. The list of applications range from self-driving vehicles to smart environmental monitoring. However, the performance of ML models after the deployment degrades because of potential drifting of the device or the environment. In this paper, we propose Self-Train, a self-supervised on-device training method for ML models to adapt to post-deployment drifting without labels. Self-Train employs offline contrastive feature learning and online drift detection with self-supervised adaptation. Experiments on images and real-world sensor datasets demonstrate consistent accuracy improvements over state-of-the-art online unsupervised methods with 2.45× at maximum, while maintaining lower execution time with a maximum of 32.7× speedup.
{"title":"Self-Train: Self-Supervised On-Device Training for Post-Deployment Adaptation","authors":"Jinhao Liu, Xiaofan Yu, T. Simunic","doi":"10.1109/SmartIoT55134.2022.00034","DOIUrl":"https://doi.org/10.1109/SmartIoT55134.2022.00034","url":null,"abstract":"Recent years have witnessed a significant increase in deploying lightweight machine learning (ML) on embedded systems. The list of applications range from self-driving vehicles to smart environmental monitoring. However, the performance of ML models after the deployment degrades because of potential drifting of the device or the environment. In this paper, we propose Self-Train, a self-supervised on-device training method for ML models to adapt to post-deployment drifting without labels. Self-Train employs offline contrastive feature learning and online drift detection with self-supervised adaptation. Experiments on images and real-world sensor datasets demonstrate consistent accuracy improvements over state-of-the-art online unsupervised methods with 2.45× at maximum, while maintaining lower execution time with a maximum of 32.7× speedup.","PeriodicalId":422269,"journal":{"name":"2022 IEEE International Conference on Smart Internet of Things (SmartIoT)","volume":"3 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123805239","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-08-01DOI: 10.1109/SmartIoT55134.2022.00035
Jianlong Zhang, Tianhong Wang, Bin Wang, Chen Chen
For object detection methods based on neural networks in computer vision, hyper-parameter is a crucial factor in the detection performance. Traditional hyperparameter optimization methods share the following shortcomings. (1) Search performance depends heavily on historical data and computational resources. (2) The open-loop structure of models may lead to unstable search results. We take missed detection targets as feedback to establish an iterative search model and propose a subspace-fusion optimization method based on mean regression. Firstly, the Successive Halving algorithm is deployed to determine the initial seeds, then detection subspaces and missed detection subspaces are generated according to the object detection results, and anchor-vector-based mean regressions are performed in the two subspaces respectively. Finally, we obtain the optimal parameters by a linear fusion of the two regression results. An early termination strategy is embedded into the search process to stop the invalid searches. Experiments show that within limited resource, this paper achieves significant improvement in search efficiency and detection performance compared with the classical methods.
{"title":"A Subspace Fusion of Hyper-parameter Optimization Method Based on Mean Regression","authors":"Jianlong Zhang, Tianhong Wang, Bin Wang, Chen Chen","doi":"10.1109/SmartIoT55134.2022.00035","DOIUrl":"https://doi.org/10.1109/SmartIoT55134.2022.00035","url":null,"abstract":"For object detection methods based on neural networks in computer vision, hyper-parameter is a crucial factor in the detection performance. Traditional hyperparameter optimization methods share the following shortcomings. (1) Search performance depends heavily on historical data and computational resources. (2) The open-loop structure of models may lead to unstable search results. We take missed detection targets as feedback to establish an iterative search model and propose a subspace-fusion optimization method based on mean regression. Firstly, the Successive Halving algorithm is deployed to determine the initial seeds, then detection subspaces and missed detection subspaces are generated according to the object detection results, and anchor-vector-based mean regressions are performed in the two subspaces respectively. Finally, we obtain the optimal parameters by a linear fusion of the two regression results. An early termination strategy is embedded into the search process to stop the invalid searches. Experiments show that within limited resource, this paper achieves significant improvement in search efficiency and detection performance compared with the classical methods.","PeriodicalId":422269,"journal":{"name":"2022 IEEE International Conference on Smart Internet of Things (SmartIoT)","volume":"14 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114092009","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-08-01DOI: 10.1109/SmartIoT55134.2022.00040
Li Cui, Xin Chen, Zhuo Ma
Merging Multi-access edge computing (MEC) and non-orthogonal multiple access (NOMA) into the sixth generation (6G) Internet of Things (IoT) can satisfy the computationally intensive task's requirement of extensible and low-energy consumption service. However, it is challenging to assigning task in MEC system due to that the channel transforms over time in dynamically varying network environments. In this paper, we propose a dynamic resource scheduling and frequency scaling algorithm (DRSFS) to allocate tasks and MEC frequency optimally. On the basis of Lyapunov optimization technique, DRSFS converts the long-range random optimization problem to a suite of determinate sub-problems and obtain the optimal solution. DRSFS can obtain an optimal offload strategy by utilizing dynamic programming theory, which can be verified by the effects of different parameters. The simulation experiment results shows the superiority of DRSFS by comparing it with other two baseline algorithms in the field of the energy consumption and the queue length.
{"title":"Dynamic Resource Scheduling and Frequency Scaling in NOMA-Based Multi-access Edge Computing System","authors":"Li Cui, Xin Chen, Zhuo Ma","doi":"10.1109/SmartIoT55134.2022.00040","DOIUrl":"https://doi.org/10.1109/SmartIoT55134.2022.00040","url":null,"abstract":"Merging Multi-access edge computing (MEC) and non-orthogonal multiple access (NOMA) into the sixth generation (6G) Internet of Things (IoT) can satisfy the computationally intensive task's requirement of extensible and low-energy consumption service. However, it is challenging to assigning task in MEC system due to that the channel transforms over time in dynamically varying network environments. In this paper, we propose a dynamic resource scheduling and frequency scaling algorithm (DRSFS) to allocate tasks and MEC frequency optimally. On the basis of Lyapunov optimization technique, DRSFS converts the long-range random optimization problem to a suite of determinate sub-problems and obtain the optimal solution. DRSFS can obtain an optimal offload strategy by utilizing dynamic programming theory, which can be verified by the effects of different parameters. The simulation experiment results shows the superiority of DRSFS by comparing it with other two baseline algorithms in the field of the energy consumption and the queue length.","PeriodicalId":422269,"journal":{"name":"2022 IEEE International Conference on Smart Internet of Things (SmartIoT)","volume":"162 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124526505","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-08-01DOI: 10.1109/SmartIoT55134.2022.00037
Jianlong Zhang, Yifan Liu, Bin Wang, Chen Chen
A Synthetic aperture radar (SAR) image change detection method based on DR-UNet-CRF iterative structure is proposed by introducing a regional dynamic convolutional network to address the problems of semantic information fading phenomenon and indeterminacy of change boundaries due to differential image computation in remote sensing image change detection. Firstly, a DR-UNet segmentation network based on the dynamic region-aware convolution (DRConv) kernel is conceived to supply a univalent potential function for the conditional random field, and a guide-mask generation method guided mask generation method with feature pyramid network (FPN) based structure is presented to guide an improved dynamic convolutional UNet to obtain accurate remote sensing change regions by learning fine spatial region delineation. Secondly, the pair-wise potential function based on image grayscale features and spatial features is designed to model the inter-pixel relationship. Finally, we use a fully connected conditional random field (CRF) model to iteratively optimize for change regions to achieve semantic compensation, thus defining the boundaries of remote sensing images more precisely. By comparing with the mainstream change detection methods, it can be considered that method in this paper has better detection performance.
{"title":"A SAR Remote Sensing Image Change Detection Method Based on DR-UNet-CRF Model","authors":"Jianlong Zhang, Yifan Liu, Bin Wang, Chen Chen","doi":"10.1109/SmartIoT55134.2022.00037","DOIUrl":"https://doi.org/10.1109/SmartIoT55134.2022.00037","url":null,"abstract":"A Synthetic aperture radar (SAR) image change detection method based on DR-UNet-CRF iterative structure is proposed by introducing a regional dynamic convolutional network to address the problems of semantic information fading phenomenon and indeterminacy of change boundaries due to differential image computation in remote sensing image change detection. Firstly, a DR-UNet segmentation network based on the dynamic region-aware convolution (DRConv) kernel is conceived to supply a univalent potential function for the conditional random field, and a guide-mask generation method guided mask generation method with feature pyramid network (FPN) based structure is presented to guide an improved dynamic convolutional UNet to obtain accurate remote sensing change regions by learning fine spatial region delineation. Secondly, the pair-wise potential function based on image grayscale features and spatial features is designed to model the inter-pixel relationship. Finally, we use a fully connected conditional random field (CRF) model to iteratively optimize for change regions to achieve semantic compensation, thus defining the boundaries of remote sensing images more precisely. By comparing with the mainstream change detection methods, it can be considered that method in this paper has better detection performance.","PeriodicalId":422269,"journal":{"name":"2022 IEEE International Conference on Smart Internet of Things (SmartIoT)","volume":"26 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133082121","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-08-01DOI: 10.1109/SmartIoT55134.2022.00010
Junlong Chen, Xilong Liu
Wireless Sensor Network (WSN) has been widely applied in Internet of Things (IoT). WSN brings great convenience to people's daily lives. In reality, the limited battery's capacity always constrains normal function time of a wireless sensor. Replacing the batteries one by one for the wireless sensors deployed in large-scale network or in dangerous places is quite infeasible. The recent research results reveal that unmanned aerial vehicle (UAV) equipped with a point-to-point far-field wireless charging unit can efficiently facilitate remote powering for WSN. The advantage of adopting charging UAV is that the distance between the wireless energy emitter and the receiver can be shortened; thus, enhancing the wireless charging efficiency. However, a UAV's energy supply usually does not allow the long-term charging mission, hence, the cost and profit of UAV wireless charging should be considered in UAV provisioned charging service. In this work, we first build a UAV wireless charging pricing model to calculate the profit of the charging service, and then, we propose the Profit-driven UAV Charging (PUC) algorithm to maximize the UAV charging profit. Through extensive simulations, we have validated that the performance of our proposed algorithm outperforms the conventional Nearest-Job-Next with Preemption (NJNP) algorithm.
{"title":"Profit-driven UAV Green Wireless Charging for WSN","authors":"Junlong Chen, Xilong Liu","doi":"10.1109/SmartIoT55134.2022.00010","DOIUrl":"https://doi.org/10.1109/SmartIoT55134.2022.00010","url":null,"abstract":"Wireless Sensor Network (WSN) has been widely applied in Internet of Things (IoT). WSN brings great convenience to people's daily lives. In reality, the limited battery's capacity always constrains normal function time of a wireless sensor. Replacing the batteries one by one for the wireless sensors deployed in large-scale network or in dangerous places is quite infeasible. The recent research results reveal that unmanned aerial vehicle (UAV) equipped with a point-to-point far-field wireless charging unit can efficiently facilitate remote powering for WSN. The advantage of adopting charging UAV is that the distance between the wireless energy emitter and the receiver can be shortened; thus, enhancing the wireless charging efficiency. However, a UAV's energy supply usually does not allow the long-term charging mission, hence, the cost and profit of UAV wireless charging should be considered in UAV provisioned charging service. In this work, we first build a UAV wireless charging pricing model to calculate the profit of the charging service, and then, we propose the Profit-driven UAV Charging (PUC) algorithm to maximize the UAV charging profit. Through extensive simulations, we have validated that the performance of our proposed algorithm outperforms the conventional Nearest-Job-Next with Preemption (NJNP) algorithm.","PeriodicalId":422269,"journal":{"name":"2022 IEEE International Conference on Smart Internet of Things (SmartIoT)","volume":"146 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122892866","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The main voucher for the circulation of bulk goods is the delivery order. The traditional warehousing enterprise receives the goods through manual input, which is inefficient and error-prone. The recognition rate of the traditional algorithm model is low and cannot be applied on a large scale. According to the characteristics of steel delivery order, through the integration of algorithm technology, this paper proposes an algorithm model based on image correction, text location, text recognition, and post verification, which solves the problem of the low recognition rate of the traditional algorithm. The character recognition rate of the recognition model is more than 95%. Finally, a visual manual correction function is developed to ensure 100% accuracy of output text data. Based on the intelligent identification technology of delivery orders, the traditional goods circulation mode is transformed from the series process of offline circulation based on traditional paper documents to the parallel process with an information-sharing cloud platform as the carrier. We build an intelligent information management system of delivery order of bulk goods is constructed. The business practice shows that the system can quickly and accurately extract the text information of delivery documents and effectively improve the efficiency of goods warehousing and circulation.
{"title":"Steel Delivery Order Recognition Based on Deep Learning and Posterior Error Correction Technology","authors":"Ming Li, Weigang Wang, Kedong Wang, Xueliang Leng, Chuan-qin Zhang, Zhongwen Guo","doi":"10.1109/SmartIoT55134.2022.00036","DOIUrl":"https://doi.org/10.1109/SmartIoT55134.2022.00036","url":null,"abstract":"The main voucher for the circulation of bulk goods is the delivery order. The traditional warehousing enterprise receives the goods through manual input, which is inefficient and error-prone. The recognition rate of the traditional algorithm model is low and cannot be applied on a large scale. According to the characteristics of steel delivery order, through the integration of algorithm technology, this paper proposes an algorithm model based on image correction, text location, text recognition, and post verification, which solves the problem of the low recognition rate of the traditional algorithm. The character recognition rate of the recognition model is more than 95%. Finally, a visual manual correction function is developed to ensure 100% accuracy of output text data. Based on the intelligent identification technology of delivery orders, the traditional goods circulation mode is transformed from the series process of offline circulation based on traditional paper documents to the parallel process with an information-sharing cloud platform as the carrier. We build an intelligent information management system of delivery order of bulk goods is constructed. The business practice shows that the system can quickly and accurately extract the text information of delivery documents and effectively improve the efficiency of goods warehousing and circulation.","PeriodicalId":422269,"journal":{"name":"2022 IEEE International Conference on Smart Internet of Things (SmartIoT)","volume":"88 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131891737","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-08-01DOI: 10.1109/SmartIoT55134.2022.00019
Pengfei Hu, Chunming He, Yiming Zhu
The Industrial Internet of Things (IIoT) enables the improvement of the productivity and intelligent level of factory. The procedure of product quality inspection has generally adopted machine intelligence algorithms instead of manual operation to improve efficiency. In this paper, we propose a product quality inspection system scheme based on software-defined edge intelligent controller (SD-EIC). By adopting the software definition and resource virtualization technologies, the hardware platform of SD-EIC is designed to support the real-time control tasks and non-real-time edge computing tasks at the same time. To this end, we propose the scheme and architecture of product quality inspection system based on SD-EIC. Multiple virtual controllers and virtual edge computing nodes are constructed on a set of SD-EIC hardware platform to realize the integrated deployment of the real-time control for terminal devices and the AI model reasoning of product defect recognition algorithm based on machine vision respectively. In addition, the management and control scheme of product quality inspection system based on industrial information model is proposed. By constructing the semantic based digital twin information model of terminal device, the flexible adjustment and parameter configuration of terminal device are realized to meet the demands of flexible production and manufacturing. The proposed product quality inspection system solution can effectively improve the utilization of hardware resources and the efficiency of product quality inspection, and reduce the overall deployment cost of the system. It can flexibly adapt to a variety of different industrial scenarios.
{"title":"The Scheme and System Architecture of Product Quality Inspection based on Software-Defined Edge Intelligent Controller (SD-EIC) in Industrial Internet of Things","authors":"Pengfei Hu, Chunming He, Yiming Zhu","doi":"10.1109/SmartIoT55134.2022.00019","DOIUrl":"https://doi.org/10.1109/SmartIoT55134.2022.00019","url":null,"abstract":"The Industrial Internet of Things (IIoT) enables the improvement of the productivity and intelligent level of factory. The procedure of product quality inspection has generally adopted machine intelligence algorithms instead of manual operation to improve efficiency. In this paper, we propose a product quality inspection system scheme based on software-defined edge intelligent controller (SD-EIC). By adopting the software definition and resource virtualization technologies, the hardware platform of SD-EIC is designed to support the real-time control tasks and non-real-time edge computing tasks at the same time. To this end, we propose the scheme and architecture of product quality inspection system based on SD-EIC. Multiple virtual controllers and virtual edge computing nodes are constructed on a set of SD-EIC hardware platform to realize the integrated deployment of the real-time control for terminal devices and the AI model reasoning of product defect recognition algorithm based on machine vision respectively. In addition, the management and control scheme of product quality inspection system based on industrial information model is proposed. By constructing the semantic based digital twin information model of terminal device, the flexible adjustment and parameter configuration of terminal device are realized to meet the demands of flexible production and manufacturing. The proposed product quality inspection system solution can effectively improve the utilization of hardware resources and the efficiency of product quality inspection, and reduce the overall deployment cost of the system. It can flexibly adapt to a variety of different industrial scenarios.","PeriodicalId":422269,"journal":{"name":"2022 IEEE International Conference on Smart Internet of Things (SmartIoT)","volume":"53 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121344198","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-08-01DOI: 10.1109/SmartIoT55134.2022.00026
Chenglin Miao, Wen Su, Yanqing Fu, Xihao Chen, D. Zang
Traffic speed prediction is an incredibly important subject of Intelligent transportation system (ITS). Efficient speed prediction methods greatly contribute to reducing traffic congestion. Most existing models focus on short term while the long-term speed prediction for a whole day is not completely developed. In this paper, a Geometric Algebra Convolutional LSTM and Graph Attention (GAConvLSTM-GAT) model is proposed to raise a potential for achieving long-term speed prediction. The proposed model is composed of a Geometric Algebra ConvLSTM (GAConvLSTM) module to extract the spatial-temporal feature, and a Graph Attention (GAT) module to make speed predictions based on the features. The experiments are evaluated by two elevated highway traffic datasets. The presented results illustrate that our GAConvLSTM model outperforms multiple baseline neural network methods.
{"title":"Long-Term Traffic Speed Prediction Based on Geometric Algebra ConvLSTM and Graph Attention","authors":"Chenglin Miao, Wen Su, Yanqing Fu, Xihao Chen, D. Zang","doi":"10.1109/SmartIoT55134.2022.00026","DOIUrl":"https://doi.org/10.1109/SmartIoT55134.2022.00026","url":null,"abstract":"Traffic speed prediction is an incredibly important subject of Intelligent transportation system (ITS). Efficient speed prediction methods greatly contribute to reducing traffic congestion. Most existing models focus on short term while the long-term speed prediction for a whole day is not completely developed. In this paper, a Geometric Algebra Convolutional LSTM and Graph Attention (GAConvLSTM-GAT) model is proposed to raise a potential for achieving long-term speed prediction. The proposed model is composed of a Geometric Algebra ConvLSTM (GAConvLSTM) module to extract the spatial-temporal feature, and a Graph Attention (GAT) module to make speed predictions based on the features. The experiments are evaluated by two elevated highway traffic datasets. The presented results illustrate that our GAConvLSTM model outperforms multiple baseline neural network methods.","PeriodicalId":422269,"journal":{"name":"2022 IEEE International Conference on Smart Internet of Things (SmartIoT)","volume":"102 2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126159632","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-08-01DOI: 10.1109/SmartIoT55134.2022.00024
Victor Gonzalo Rodriguez-Ahuanari, Miguel Angel Vega-Ramirez, Hugo Eladio Chumpitaz-Caycho, Ericka Nelly Espinoza-Gamboa, Franklin Cordova-Buiza
The study aimed to know how the smart system improves data protection in higher education institutions through a systematic review between the years 2017 to 2022. It allowed reviewing important databases such as Scielo, Ebsco, ScienceDirect and Scopus. The search achieved 229 original researches in relation to the topic intelligent system and data protection, of which according to the evaluation 201 were separated for being within the exclusion criteria. Therefore, 28 of them were examined and analyzed in detail. It is concluded that the implementation of these systems significantly improves the protection of information, as evidence has been found that it provides enormous benefits applied mainly in higher education institutions. Consequently, in the research reviewed, it has been found that these systems evolve and are continuously updated, reducing costs, as well as increasing virtual security.
{"title":"Intelligent system for data protection in higher education institutions: A systematic review","authors":"Victor Gonzalo Rodriguez-Ahuanari, Miguel Angel Vega-Ramirez, Hugo Eladio Chumpitaz-Caycho, Ericka Nelly Espinoza-Gamboa, Franklin Cordova-Buiza","doi":"10.1109/SmartIoT55134.2022.00024","DOIUrl":"https://doi.org/10.1109/SmartIoT55134.2022.00024","url":null,"abstract":"The study aimed to know how the smart system improves data protection in higher education institutions through a systematic review between the years 2017 to 2022. It allowed reviewing important databases such as Scielo, Ebsco, ScienceDirect and Scopus. The search achieved 229 original researches in relation to the topic intelligent system and data protection, of which according to the evaluation 201 were separated for being within the exclusion criteria. Therefore, 28 of them were examined and analyzed in detail. It is concluded that the implementation of these systems significantly improves the protection of information, as evidence has been found that it provides enormous benefits applied mainly in higher education institutions. Consequently, in the research reviewed, it has been found that these systems evolve and are continuously updated, reducing costs, as well as increasing virtual security.","PeriodicalId":422269,"journal":{"name":"2022 IEEE International Conference on Smart Internet of Things (SmartIoT)","volume":"4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122109382","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：