Pub Date : 2022-12-09DOI: 10.1109/ACAIT56212.2022.10137934
Y. Liu
In order to solve the problems of poor output performance and large output oscillation of traditional low-temperature solid oxide fuel cells, artificial intelligence technology was introduced in this paper to analyze the performance of low-temperature solid oxide fuel cells. Firstly, the steady-state control system of the battery was constructed, and the three-dimensional structure design and electrical performance optimization of the battery were realized. Then, the electrode potential induction analysis model was constructed to analyze the carbon / metal oxide electrode materials with stable mechanical and electrochemical properties. Thirdly, combined with the three-phase regulation of the battery electrode, the microstructure area in the fuel cell electric field is controlled. Finally, according to the fuel cell output voltage, the fuel cell ion mass conservation model is constructed. Artificial intelligence is used to obtain the optimal solution of fuel cell voltage output, so as to complete the analysis of fuel cell steady-state performance. The simulation results show that this method can control the output of the low-temperature solid oxide fuel cell well and reduce the output oscillation of the cell, which has a certain theoretical reference significance for the performance of the cell.
{"title":"Performance Analysis of Low Temperature Solid Oxide Fuel Cell Based on Artificial Intelligence Technology","authors":"Y. Liu","doi":"10.1109/ACAIT56212.2022.10137934","DOIUrl":"https://doi.org/10.1109/ACAIT56212.2022.10137934","url":null,"abstract":"In order to solve the problems of poor output performance and large output oscillation of traditional low-temperature solid oxide fuel cells, artificial intelligence technology was introduced in this paper to analyze the performance of low-temperature solid oxide fuel cells. Firstly, the steady-state control system of the battery was constructed, and the three-dimensional structure design and electrical performance optimization of the battery were realized. Then, the electrode potential induction analysis model was constructed to analyze the carbon / metal oxide electrode materials with stable mechanical and electrochemical properties. Thirdly, combined with the three-phase regulation of the battery electrode, the microstructure area in the fuel cell electric field is controlled. Finally, according to the fuel cell output voltage, the fuel cell ion mass conservation model is constructed. Artificial intelligence is used to obtain the optimal solution of fuel cell voltage output, so as to complete the analysis of fuel cell steady-state performance. The simulation results show that this method can control the output of the low-temperature solid oxide fuel cell well and reduce the output oscillation of the cell, which has a certain theoretical reference significance for the performance of the cell.","PeriodicalId":398228,"journal":{"name":"2022 6th Asian Conference on Artificial Intelligence Technology (ACAIT)","volume":"17 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-12-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114969972","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 : 2022-12-09DOI: 10.1109/ACAIT56212.2022.10137926
Benguo Ye, Xiaofeng Yuan, Gang Peng, Weizhen Zeng
LSTM is a sequential model containing the long short-term memory cells gated recurrent units. Compared to the traditional RNN, LSTM introduces three gates which solve the exploding and vanishing gradient problems of RNN. In this paper, we propose a new speech emotion model by combining CNN and LSTM. The model is implemented based on the CASIA data sets, the Python librosa library and the opensmile tool to get the speech emotion features by extracting the multi-feature of the fusion acoustics which would then be compared to the features based on different configurations to evaluate the recognition accuracy. The experimental results show that the features extracted from the emobase2010 configuration can achieve 84% recognition accuracy based on the CASIA dataset. Compared with other models, the recognition accuracy of the model introduced in1 this paper is 3.3% higher than that of the SVM model, but 6.3% lower than that of the ConvLSTM model.
{"title":"A Novel Speech Emotion Model Based on CNN and LSTM Networks","authors":"Benguo Ye, Xiaofeng Yuan, Gang Peng, Weizhen Zeng","doi":"10.1109/ACAIT56212.2022.10137926","DOIUrl":"https://doi.org/10.1109/ACAIT56212.2022.10137926","url":null,"abstract":"LSTM is a sequential model containing the long short-term memory cells gated recurrent units. Compared to the traditional RNN, LSTM introduces three gates which solve the exploding and vanishing gradient problems of RNN. In this paper, we propose a new speech emotion model by combining CNN and LSTM. The model is implemented based on the CASIA data sets, the Python librosa library and the opensmile tool to get the speech emotion features by extracting the multi-feature of the fusion acoustics which would then be compared to the features based on different configurations to evaluate the recognition accuracy. The experimental results show that the features extracted from the emobase2010 configuration can achieve 84% recognition accuracy based on the CASIA dataset. Compared with other models, the recognition accuracy of the model introduced in1 this paper is 3.3% higher than that of the SVM model, but 6.3% lower than that of the ConvLSTM model.","PeriodicalId":398228,"journal":{"name":"2022 6th Asian Conference on Artificial Intelligence Technology (ACAIT)","volume":"2013 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-12-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127435211","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 : 2022-12-09DOI: 10.1109/ACAIT56212.2022.10137834
R. Liu, Wei Pu, Yangyang Zou, Linfeng Jiang, Zhiyong Ye
Ischemic strokes are the most common acute brain disorder, and seriously threaten patients’ lives. In order to help physicians determine the location of ischemic stroke lesions and other information as early as possible, many scholars have used convolutional neural networks and Transformer segmentation networks to segment lesions on CT perfusion images. However, convolutional neural networks are not capable of extracting spatial information sufficiently, which leads to loss of effective lesion information. In addition, the global attention mechanism module of Transformer is computationally intensive at runtime, which is not suitable for use in high-resolution input and intensive prediction tasks. We designed a DSE-ResNet module to solve these problems to establish spatial channel information correlation. Then we innovatively propose the Pool-UNet model, which combines the Poolformer structure with a convolutional neural network. It can efficiently model the global context and learn multi-scale features while maintaining a grasp of the lowlevel details. The segmentation results on the ISLES-2018 dataset show that PoolUNet achieves 67.82% precision, 56.54% recall, 56.04% Dice coefficient, and 21.14 mm Haushofer distance. Compared with the classical UNet, R2UNet, and TransUNet 3 segmentation models, Pool-UNet improved at least 0.26%, 1.52%, 1.07%, and 0. 17mm in accuracy, recall, Dice coefficient, and Hausdorff distance, respectively. Pool-UNet has a competitive advantage over other classical and advanced medical segmentation algorithms.
{"title":"Pool-UNet: Ischemic Stroke Segmentation from CT Perfusion Scans Using Poolformer UNet","authors":"R. Liu, Wei Pu, Yangyang Zou, Linfeng Jiang, Zhiyong Ye","doi":"10.1109/ACAIT56212.2022.10137834","DOIUrl":"https://doi.org/10.1109/ACAIT56212.2022.10137834","url":null,"abstract":"Ischemic strokes are the most common acute brain disorder, and seriously threaten patients’ lives. In order to help physicians determine the location of ischemic stroke lesions and other information as early as possible, many scholars have used convolutional neural networks and Transformer segmentation networks to segment lesions on CT perfusion images. However, convolutional neural networks are not capable of extracting spatial information sufficiently, which leads to loss of effective lesion information. In addition, the global attention mechanism module of Transformer is computationally intensive at runtime, which is not suitable for use in high-resolution input and intensive prediction tasks. We designed a DSE-ResNet module to solve these problems to establish spatial channel information correlation. Then we innovatively propose the Pool-UNet model, which combines the Poolformer structure with a convolutional neural network. It can efficiently model the global context and learn multi-scale features while maintaining a grasp of the lowlevel details. The segmentation results on the ISLES-2018 dataset show that PoolUNet achieves 67.82% precision, 56.54% recall, 56.04% Dice coefficient, and 21.14 mm Haushofer distance. Compared with the classical UNet, R2UNet, and TransUNet 3 segmentation models, Pool-UNet improved at least 0.26%, 1.52%, 1.07%, and 0. 17mm in accuracy, recall, Dice coefficient, and Hausdorff distance, respectively. Pool-UNet has a competitive advantage over other classical and advanced medical segmentation algorithms.","PeriodicalId":398228,"journal":{"name":"2022 6th Asian Conference on Artificial Intelligence Technology (ACAIT)","volume":"37 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-12-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124935013","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 : 2022-12-09DOI: 10.1109/ACAIT56212.2022.10137804
Lei Wang, Yue Jin
Aesthetic quality evaluation of images has an important role in the field of visual analysis, and the widespread use of high-quality image editing has gradually increased the importance of image aesthetic evaluation in automatic image processing tasks. Previous researchers have mostly explored the mapping relationship between images and labeled scores using convolutional neural networks, but the aesthetic features of different regions on images have not been explored sufficiently, when an image is rich in background information and it is necessary to correlate the aesthetic features of different regions to evaluate the image, convolutional neural networks often cannot extract the aesthetic features of the image adequately due to the lack of the advantage of global feature modeling. We introduce a novel Transformer architecture for image aesthetic quality assessment(IAFormer), IAFormer can model the global aesthetic features of an image, and it is a framework that unifies the aesthetic quality assessment of images and the aesthetic cropping of images, while the aesthetic quality of the image is evaluated, the aesthetic weights on different patches within the image can be calculated to give valid reference information for the aesthetic cropping task.
{"title":"IAFormer: A Transformer Network for Image Aesthetic Evaluation and Cropping","authors":"Lei Wang, Yue Jin","doi":"10.1109/ACAIT56212.2022.10137804","DOIUrl":"https://doi.org/10.1109/ACAIT56212.2022.10137804","url":null,"abstract":"Aesthetic quality evaluation of images has an important role in the field of visual analysis, and the widespread use of high-quality image editing has gradually increased the importance of image aesthetic evaluation in automatic image processing tasks. Previous researchers have mostly explored the mapping relationship between images and labeled scores using convolutional neural networks, but the aesthetic features of different regions on images have not been explored sufficiently, when an image is rich in background information and it is necessary to correlate the aesthetic features of different regions to evaluate the image, convolutional neural networks often cannot extract the aesthetic features of the image adequately due to the lack of the advantage of global feature modeling. We introduce a novel Transformer architecture for image aesthetic quality assessment(IAFormer), IAFormer can model the global aesthetic features of an image, and it is a framework that unifies the aesthetic quality assessment of images and the aesthetic cropping of images, while the aesthetic quality of the image is evaluated, the aesthetic weights on different patches within the image can be calculated to give valid reference information for the aesthetic cropping task.","PeriodicalId":398228,"journal":{"name":"2022 6th Asian Conference on Artificial Intelligence Technology (ACAIT)","volume":"9 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-12-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125276357","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}
Diabetic peripheral neuropathy (DPN), one of the common long-term complications of diabetes, may affect the physical condition and quality of life of patients. Corneal confocal microscopy (CCM) is a rapid non-invasive ophthalmic imaging technique that can be used to observe the form of nerve fibers in sub-basal corneal nerve plexus directly. Analysis of the nerve fibers in CCM images quantifies features of nerve fibers, can apply to clinical diagnosis of DPN. This paper presents an attention deep learning model for detecting nerve fibers from CCM images, which combine context encoder network and squeeze-and-excitation networks. The algorithm with attention mechanism can solve the problem of the segmentation result is easily influenced by high level noise in CCM images and imbalance of nerve fiber pixels and background pixels to a certain degree. The proposed algorithm shows the best performance among common image segmentation deep learning model.
{"title":"Context Encoder Network with Channel-Wise Attention Mechanism for Nerve Fibers Detection in Corneal Confocal Microscopy Images","authors":"Wenyuan Li, Zheng Tang, Lulu Zhao, Wanyong Tian, Taotao Qi","doi":"10.1109/ACAIT56212.2022.10137928","DOIUrl":"https://doi.org/10.1109/ACAIT56212.2022.10137928","url":null,"abstract":"Diabetic peripheral neuropathy (DPN), one of the common long-term complications of diabetes, may affect the physical condition and quality of life of patients. Corneal confocal microscopy (CCM) is a rapid non-invasive ophthalmic imaging technique that can be used to observe the form of nerve fibers in sub-basal corneal nerve plexus directly. Analysis of the nerve fibers in CCM images quantifies features of nerve fibers, can apply to clinical diagnosis of DPN. This paper presents an attention deep learning model for detecting nerve fibers from CCM images, which combine context encoder network and squeeze-and-excitation networks. The algorithm with attention mechanism can solve the problem of the segmentation result is easily influenced by high level noise in CCM images and imbalance of nerve fiber pixels and background pixels to a certain degree. The proposed algorithm shows the best performance among common image segmentation deep learning model.","PeriodicalId":398228,"journal":{"name":"2022 6th Asian Conference on Artificial Intelligence Technology (ACAIT)","volume":"22 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-12-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127018361","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 : 2022-12-09DOI: 10.1109/ACAIT56212.2022.10137828
Zhiqiang Ma, Jinyi Li, Junpeng Zhang
Aiming at the problem of uneven regional distribution of speech caused by the lack of labeled data in the Mongolian speech data set, this paper proposes a Mongolian speech data augmentation model based on a conditional generation confrontation network. The model uses conditional speech generators and multiple fusion discriminators for adversarial learning, and uses Mongolian text and specified regional features to generate Mongolian speech with specified regional features. The original data set was augmented by using the methods of speech rate perturbation and spectrogram enhancement, and compared with the end-to-end Mongolian speech recognition model trained on different augment data sets and the original data sets, it was found that the word error rate in the end-to-end Mongolian speech recognition model trained on the augment data set of the specified regional characteristics is 3.1%; Compared with the end-to-end Mongolian speech recognition model trained on the original data set, the speech rate disturbance data set, and the spectrogram enhancement data set, the word error rate dropped by 2%, 0.5%, and 0.8%.
{"title":"Speech Augmentation Using Conditional Generative Adversarial Nets in Mongolian Speech Recognition","authors":"Zhiqiang Ma, Jinyi Li, Junpeng Zhang","doi":"10.1109/ACAIT56212.2022.10137828","DOIUrl":"https://doi.org/10.1109/ACAIT56212.2022.10137828","url":null,"abstract":"Aiming at the problem of uneven regional distribution of speech caused by the lack of labeled data in the Mongolian speech data set, this paper proposes a Mongolian speech data augmentation model based on a conditional generation confrontation network. The model uses conditional speech generators and multiple fusion discriminators for adversarial learning, and uses Mongolian text and specified regional features to generate Mongolian speech with specified regional features. The original data set was augmented by using the methods of speech rate perturbation and spectrogram enhancement, and compared with the end-to-end Mongolian speech recognition model trained on different augment data sets and the original data sets, it was found that the word error rate in the end-to-end Mongolian speech recognition model trained on the augment data set of the specified regional characteristics is 3.1%; Compared with the end-to-end Mongolian speech recognition model trained on the original data set, the speech rate disturbance data set, and the spectrogram enhancement data set, the word error rate dropped by 2%, 0.5%, and 0.8%.","PeriodicalId":398228,"journal":{"name":"2022 6th Asian Conference on Artificial Intelligence Technology (ACAIT)","volume":"77 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-12-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115541039","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 : 2022-12-09DOI: 10.1109/ACAIT56212.2022.10137889
Yiwei Zhang
In the production of multi-scale block-fused art images, the image quality is poor due to the sudden change and occlusion of color shift. This paper puts forward a compensation method for color shift of multi-scale block-fused art images based on augmented reality technology. Based on the attenuation of optical parameters and the control of color balance, a multi-dimensional color fusion model of background light correlation of multi-scale block fused art images is established, and the augmented reality model of multi-scale block fused art images is constructed by combining the analysis method of surface features and illumination features distribution model of art images. Through the parameter analysis of each level feature map model of the similarity degree of the previous frame under color deviation, The gray texture and color texture features of multi-scale block-fused art images with complementary advantages and disadvantages are extracted, and augmented reality technology is adopted to realize gray scale enhancement and color enhancement in the process of color compensation of art images. Combined parameter identification method is adopted to realize color adjustment and feedback compensation control of output stability of art images. According to the characteristics of high-order moment output stability of color features, color offset compensation and optimal imaging processing of multi-scale block-fused art images are realized by calculating and counting boundary corner information and texture parameter analysis. The test shows that this method performs the color offset processing of multi-scale block fusion art image sensor, improves the color offset compensation ability of art images and the true color imaging quality of images, and increases the peak signal-to-noise ratio of output images.
{"title":"Color Offset Compensation Method of Art Image Based on Augmented Reality Technology","authors":"Yiwei Zhang","doi":"10.1109/ACAIT56212.2022.10137889","DOIUrl":"https://doi.org/10.1109/ACAIT56212.2022.10137889","url":null,"abstract":"In the production of multi-scale block-fused art images, the image quality is poor due to the sudden change and occlusion of color shift. This paper puts forward a compensation method for color shift of multi-scale block-fused art images based on augmented reality technology. Based on the attenuation of optical parameters and the control of color balance, a multi-dimensional color fusion model of background light correlation of multi-scale block fused art images is established, and the augmented reality model of multi-scale block fused art images is constructed by combining the analysis method of surface features and illumination features distribution model of art images. Through the parameter analysis of each level feature map model of the similarity degree of the previous frame under color deviation, The gray texture and color texture features of multi-scale block-fused art images with complementary advantages and disadvantages are extracted, and augmented reality technology is adopted to realize gray scale enhancement and color enhancement in the process of color compensation of art images. Combined parameter identification method is adopted to realize color adjustment and feedback compensation control of output stability of art images. According to the characteristics of high-order moment output stability of color features, color offset compensation and optimal imaging processing of multi-scale block-fused art images are realized by calculating and counting boundary corner information and texture parameter analysis. The test shows that this method performs the color offset processing of multi-scale block fusion art image sensor, improves the color offset compensation ability of art images and the true color imaging quality of images, and increases the peak signal-to-noise ratio of output images.","PeriodicalId":398228,"journal":{"name":"2022 6th Asian Conference on Artificial Intelligence Technology (ACAIT)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-12-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129418306","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}
Spoken Language Understanding (SLU) is a very important module in intelligent dialogue systems. It is usually constructed based on a bi-directional long and short-term memory network (BiLSTM). It has some shortcomings, such as relative single representation of feature space and fuzzy semantic features. For this reason, this study constructs a SLU model which combines the temporal characteristics of context and the characteristics of multi-layer representation space. The model combines a bi-directional long and short-term memory network and a multi-head self-attention to extract different feature information of contextual temporal features and multisemantic representation space of the text, respectively; then, the two features are fused using a residual linking method to enhance the features of word dependence at different locations within the text; meanwhile, the gate mechanism is then used to enable the intent detection task to establish an influence relationship on the slot filling task. Finally, the SNIPS dataset, the ATIS dataset, and the slot-gated model are selected for comparison experiments. The slot filling F1 value is increased by 4.14% and 1.1% respectively, and the accuracy of semantic framework is increased by 4.25% and 2.50% respectively. The results show the effectiveness of the model of SLU task.
{"title":"A Model of Spoken Language Understanding Combining with Multi-Head Self-Attention","authors":"Dafei Lin, Jiangfeng Zhou, Xinlai Xing, Xiaochuan Zhang","doi":"10.1109/ACAIT56212.2022.10137905","DOIUrl":"https://doi.org/10.1109/ACAIT56212.2022.10137905","url":null,"abstract":"Spoken Language Understanding (SLU) is a very important module in intelligent dialogue systems. It is usually constructed based on a bi-directional long and short-term memory network (BiLSTM). It has some shortcomings, such as relative single representation of feature space and fuzzy semantic features. For this reason, this study constructs a SLU model which combines the temporal characteristics of context and the characteristics of multi-layer representation space. The model combines a bi-directional long and short-term memory network and a multi-head self-attention to extract different feature information of contextual temporal features and multisemantic representation space of the text, respectively; then, the two features are fused using a residual linking method to enhance the features of word dependence at different locations within the text; meanwhile, the gate mechanism is then used to enable the intent detection task to establish an influence relationship on the slot filling task. Finally, the SNIPS dataset, the ATIS dataset, and the slot-gated model are selected for comparison experiments. The slot filling F1 value is increased by 4.14% and 1.1% respectively, and the accuracy of semantic framework is increased by 4.25% and 2.50% respectively. The results show the effectiveness of the model of SLU task.","PeriodicalId":398228,"journal":{"name":"2022 6th Asian Conference on Artificial Intelligence Technology (ACAIT)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-12-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130389127","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 : 2022-12-09DOI: 10.1109/ACAIT56212.2022.10137947
He Yan, Yao Li, Kuilin Huang, Xiaotang Wang
To solve the problem that the global features extracted by the ResNet-50 network has insufficient recognition capability in similar vehicle re-identification task, a new Re-ID method combining metric learning is proposed. Firstly, the fine-grained features of vehicles are extracted by using triplet constraints, and then combined with the global features extracted by the backbone network as vehicle features. Secondly, the similarity of different vehicle features is judged and ranked by Euclidean distance, so as to obtain more accurate results. Finally, a comparative experiment is conducted on the VeRi-776 dataset for different network models. The results show that our method has high recognition accuracy in Re-ID tasks. Compared with ResNet-50, the mean average accuracy (mAP) is improved by 2.30 %, rank-l increased by 2.31 %, and the rank-5 increased by 2.05 %. It is verified that this model can effectively improve the recognition accuracy in vehicle Re-ID.
{"title":"A Vehicle Re-Identification Method Based on Fine-Grained Features and Metric Learning","authors":"He Yan, Yao Li, Kuilin Huang, Xiaotang Wang","doi":"10.1109/ACAIT56212.2022.10137947","DOIUrl":"https://doi.org/10.1109/ACAIT56212.2022.10137947","url":null,"abstract":"To solve the problem that the global features extracted by the ResNet-50 network has insufficient recognition capability in similar vehicle re-identification task, a new Re-ID method combining metric learning is proposed. Firstly, the fine-grained features of vehicles are extracted by using triplet constraints, and then combined with the global features extracted by the backbone network as vehicle features. Secondly, the similarity of different vehicle features is judged and ranked by Euclidean distance, so as to obtain more accurate results. Finally, a comparative experiment is conducted on the VeRi-776 dataset for different network models. The results show that our method has high recognition accuracy in Re-ID tasks. Compared with ResNet-50, the mean average accuracy (mAP) is improved by 2.30 %, rank-l increased by 2.31 %, and the rank-5 increased by 2.05 %. It is verified that this model can effectively improve the recognition accuracy in vehicle Re-ID.","PeriodicalId":398228,"journal":{"name":"2022 6th Asian Conference on Artificial Intelligence Technology (ACAIT)","volume":"32 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-12-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132493624","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 : 2022-12-09DOI: 10.1109/ACAIT56212.2022.10137812
Zhenjia Chen, Lihui Wang
In order to improve the blind localization accuracy of radio signal sources, this paper proposes a radio frequency (RF) signal source target localization method based on Monte Carlo algorithm. Combined with the free-space propagation loss characteristics of electromagnetic waves, real-time electromagnetic spectrum detection data based on distributed detection nodes are analyzed. Based on the spatial distribution of electromagnetic spectrum, the target location method of radio frequency signal source is studied. Distributed detection nodes detect electromagnetic spectrum data of the same radio frequency signal source target at different spatial locations. The Monte Carlo algorithm is used for the cooperative detection target localization method of the RF signal source. With the increase of random times, the estimation result with the smallest cumulative distance error is used as the estimation result of the target position of the radio frequency signal source. The measured data show that the method can improve the blind detection and positioning accuracy of radio signal sources in the target area.
{"title":"RF Signal Source Target Detection Method Based on Monte Carlo Algorithm in Distributed Electromagnetic Spectrum Monitoring System","authors":"Zhenjia Chen, Lihui Wang","doi":"10.1109/ACAIT56212.2022.10137812","DOIUrl":"https://doi.org/10.1109/ACAIT56212.2022.10137812","url":null,"abstract":"In order to improve the blind localization accuracy of radio signal sources, this paper proposes a radio frequency (RF) signal source target localization method based on Monte Carlo algorithm. Combined with the free-space propagation loss characteristics of electromagnetic waves, real-time electromagnetic spectrum detection data based on distributed detection nodes are analyzed. Based on the spatial distribution of electromagnetic spectrum, the target location method of radio frequency signal source is studied. Distributed detection nodes detect electromagnetic spectrum data of the same radio frequency signal source target at different spatial locations. The Monte Carlo algorithm is used for the cooperative detection target localization method of the RF signal source. With the increase of random times, the estimation result with the smallest cumulative distance error is used as the estimation result of the target position of the radio frequency signal source. The measured data show that the method can improve the blind detection and positioning accuracy of radio signal sources in the target area.","PeriodicalId":398228,"journal":{"name":"2022 6th Asian Conference on Artificial Intelligence Technology (ACAIT)","volume":"36 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-12-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131081437","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：