Pub Date : 2021-12-17DOI: 10.1109/ICCWAMTIP53232.2021.9674079
Amin Ul Haq, J. Li, R. U. Khan, C. Mawuli, B. L. Y. Agbley, Mordecai F. Raj, Wang Zhou, Jalaluddin Khan, Abdul Haq, Abdus Saboor, Faiza Habib, Zafar Khan
Accurate diagnostic system is significantly important for timely COVID-19 identification. Diagnosing COVID-19 from chest x-ray images employing the CNN model is recommended for accurate recognition of COVID-19. The existing diagnosis techniques of COVID-19 still lack high accuracy. To handle this problem in this work, we have proposed accurate detection method for COVID-19. In the proposed method, a CNN is incorporated for the diagnosis of COVID-19 using chest x-ray images data. The experimental results illustrate that our technique is good for COVID-19 accurate diagnosis and can be easily implemented in health care systems.
{"title":"Deep Learning Approach for COVID-19 Identification","authors":"Amin Ul Haq, J. Li, R. U. Khan, C. Mawuli, B. L. Y. Agbley, Mordecai F. Raj, Wang Zhou, Jalaluddin Khan, Abdul Haq, Abdus Saboor, Faiza Habib, Zafar Khan","doi":"10.1109/ICCWAMTIP53232.2021.9674079","DOIUrl":"https://doi.org/10.1109/ICCWAMTIP53232.2021.9674079","url":null,"abstract":"Accurate diagnostic system is significantly important for timely COVID-19 identification. Diagnosing COVID-19 from chest x-ray images employing the CNN model is recommended for accurate recognition of COVID-19. The existing diagnosis techniques of COVID-19 still lack high accuracy. To handle this problem in this work, we have proposed accurate detection method for COVID-19. In the proposed method, a CNN is incorporated for the diagnosis of COVID-19 using chest x-ray images data. The experimental results illustrate that our technique is good for COVID-19 accurate diagnosis and can be easily implemented in health care systems.","PeriodicalId":358772,"journal":{"name":"2021 18th International Computer Conference on Wavelet Active Media Technology and Information Processing (ICCWAMTIP)","volume":"24 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127167141","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 : 2021-12-17DOI: 10.1109/ICCWAMTIP53232.2021.9674130
T. Yi, Li Min, Zhou Yinghai, Tang Dingyong, Wen Zepeng, YunChang Cheng
This paper studies the optimization algorithm for Ultra Wide Band(UWB) high-precision positioning technology. By observing and analyzing the characteristics of the measurement data set in the real experimental scene, we study the data fluctuation characteristics under signal interference, and make it clear that the UWB precise positioning problem under signal interference belongs to a typical error optimization problem. To solve this problem, we first use an anomaly detection algorithm and clustering algorithm to clean the experimental data set and retain more effective data features. For the optimization algorithm of precise positioning, our innovation is to give priority to the study of the spatial coordinate conversion formula with more physical meaning and the numerical distribution characteristics of various offset differences, and establish an optimization algorithm to achieve centimeter accuracy by using only one offset to correct the original distance.
{"title":"A Novel Positioning Optimization Approach Towards UWB","authors":"T. Yi, Li Min, Zhou Yinghai, Tang Dingyong, Wen Zepeng, YunChang Cheng","doi":"10.1109/ICCWAMTIP53232.2021.9674130","DOIUrl":"https://doi.org/10.1109/ICCWAMTIP53232.2021.9674130","url":null,"abstract":"This paper studies the optimization algorithm for Ultra Wide Band(UWB) high-precision positioning technology. By observing and analyzing the characteristics of the measurement data set in the real experimental scene, we study the data fluctuation characteristics under signal interference, and make it clear that the UWB precise positioning problem under signal interference belongs to a typical error optimization problem. To solve this problem, we first use an anomaly detection algorithm and clustering algorithm to clean the experimental data set and retain more effective data features. For the optimization algorithm of precise positioning, our innovation is to give priority to the study of the spatial coordinate conversion formula with more physical meaning and the numerical distribution characteristics of various offset differences, and establish an optimization algorithm to achieve centimeter accuracy by using only one offset to correct the original distance.","PeriodicalId":358772,"journal":{"name":"2021 18th International Computer Conference on Wavelet Active Media Technology and Information Processing (ICCWAMTIP)","volume":"11 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123680737","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 : 2021-12-17DOI: 10.1109/ICCWAMTIP53232.2021.9674082
Shi Qirui, Chen Hongle, Chen Juan, Wen Quan
Weight pruning is widely used for model compression and acceleration. In this work, a novel Adam optimization method for few-shot learning with GAN data augmentation is proposed. The first-order Taylor series is implemented to evaluate parameters' importance toward the loss function. And with the given compression ratio, parameters with importance above the threshold are updated by the Adam optimizer with momentum-accelerated weight decay, while others have negative updates as the penalization. After continuous iterations, the model enables to achieve corresponding sparsity ratio, with the influence of the redundant parameters reducing to a low extent. Experiments demonstrate that this method is effective on ResNet with CUB and ISIC-2018 datasets. Note that CUB and ISIC-2018 are datasets about birds and skin deceases, respectively, which represents the generalization of our method on cross-domain classification issues. As a result, the pruned model is able to retain the accuracy with high model sparse ratios. And in some specific compress ratio, like 10× for CUB dataset and 3 × for ISIC-2018 dataset, the pruned model even outperforms the origin model by 3.15% and 1.16%, respectively.
{"title":"Adam Optimized Pruning Method for Few-Shot Learning with GAN Data Augmentation","authors":"Shi Qirui, Chen Hongle, Chen Juan, Wen Quan","doi":"10.1109/ICCWAMTIP53232.2021.9674082","DOIUrl":"https://doi.org/10.1109/ICCWAMTIP53232.2021.9674082","url":null,"abstract":"Weight pruning is widely used for model compression and acceleration. In this work, a novel Adam optimization method for few-shot learning with GAN data augmentation is proposed. The first-order Taylor series is implemented to evaluate parameters' importance toward the loss function. And with the given compression ratio, parameters with importance above the threshold are updated by the Adam optimizer with momentum-accelerated weight decay, while others have negative updates as the penalization. After continuous iterations, the model enables to achieve corresponding sparsity ratio, with the influence of the redundant parameters reducing to a low extent. Experiments demonstrate that this method is effective on ResNet with CUB and ISIC-2018 datasets. Note that CUB and ISIC-2018 are datasets about birds and skin deceases, respectively, which represents the generalization of our method on cross-domain classification issues. As a result, the pruned model is able to retain the accuracy with high model sparse ratios. And in some specific compress ratio, like 10× for CUB dataset and 3 × for ISIC-2018 dataset, the pruned model even outperforms the origin model by 3.15% and 1.16%, respectively.","PeriodicalId":358772,"journal":{"name":"2021 18th International Computer Conference on Wavelet Active Media Technology and Information Processing (ICCWAMTIP)","volume":"60 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116579268","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}
With more and more network attacks on the Internet, this article uses the C4.5 decision tree classification algorithm to detect 7 types of network attacks from the perspective of network traffic. Firstly, the data is preprocessed, then extract 62 features, finally use our proposed C4.5 divided algorithm for traffic detection. This experiment uses the public data set CSE-CIC-IDS2018 for verification. The experimental results show that the method in this article can effectively detect different types of cyber attacks. The accuracy rate can reach 96.7%, and the false positive rate is only 4.5%.
{"title":"Multi-Traffic Features Network Intrusion Detection Algorithm Based on C4.5","authors":"Jingwen Zhou, Xinnan Jiang, Changan Liu, Jing Zhang, Lingling Liao, Jiazhong Lu","doi":"10.1109/ICCWAMTIP53232.2021.9674129","DOIUrl":"https://doi.org/10.1109/ICCWAMTIP53232.2021.9674129","url":null,"abstract":"With more and more network attacks on the Internet, this article uses the C4.5 decision tree classification algorithm to detect 7 types of network attacks from the perspective of network traffic. Firstly, the data is preprocessed, then extract 62 features, finally use our proposed C4.5 divided algorithm for traffic detection. This experiment uses the public data set CSE-CIC-IDS2018 for verification. The experimental results show that the method in this article can effectively detect different types of cyber attacks. The accuracy rate can reach 96.7%, and the false positive rate is only 4.5%.","PeriodicalId":358772,"journal":{"name":"2021 18th International Computer Conference on Wavelet Active Media Technology and Information Processing (ICCWAMTIP)","volume":"21 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129567806","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 : 2021-12-17DOI: 10.1109/ICCWAMTIP53232.2021.9674156
B. Junjie, Zhou Taoqi, Cai Jianfneg, Gao Shuai, Li Jiajie, Bai Junbo
The edge camber detection of hot rolled steel mainly depends on human eyes and sensors, but its accuracy is unsatisfactory and there is a certain delay. Therefore, the scheme of using machine vision to detect the bending of hot rolled steel was explored in this paper. Based on the practical engineering problems of hot-rolled plate bending detection, aiming at improving the imaging identification, effectiveness and industrial applicability, an identification method and some key technologies of edge camber detection image were explored in tins paper. Combined with the real hot-rolled steel bending image, the operators of Gaussian, Sobel and Laplace, and Hough transform were used for image preprocessing and line detection respectively. After defining the bending coefficient, the plate bending degree was divided into five grades. And the experiment shows that the machine vision method explored in this paper can accurately judge whether the hot-rolled steel is bent and grade the bending degree.
{"title":"Research On Edge Camber Detection Method Of Hot Rolled Steel Based On Hough Transform","authors":"B. Junjie, Zhou Taoqi, Cai Jianfneg, Gao Shuai, Li Jiajie, Bai Junbo","doi":"10.1109/ICCWAMTIP53232.2021.9674156","DOIUrl":"https://doi.org/10.1109/ICCWAMTIP53232.2021.9674156","url":null,"abstract":"The edge camber detection of hot rolled steel mainly depends on human eyes and sensors, but its accuracy is unsatisfactory and there is a certain delay. Therefore, the scheme of using machine vision to detect the bending of hot rolled steel was explored in this paper. Based on the practical engineering problems of hot-rolled plate bending detection, aiming at improving the imaging identification, effectiveness and industrial applicability, an identification method and some key technologies of edge camber detection image were explored in tins paper. Combined with the real hot-rolled steel bending image, the operators of Gaussian, Sobel and Laplace, and Hough transform were used for image preprocessing and line detection respectively. After defining the bending coefficient, the plate bending degree was divided into five grades. And the experiment shows that the machine vision method explored in this paper can accurately judge whether the hot-rolled steel is bent and grade the bending degree.","PeriodicalId":358772,"journal":{"name":"2021 18th International Computer Conference on Wavelet Active Media Technology and Information Processing (ICCWAMTIP)","volume":"14 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128357847","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 : 2021-12-17DOI: 10.1109/ICCWAMTIP53232.2021.9673710
Gershon Rodor, Yang Fengfan, Stephen Nkyi, Harit Rijal, Enoch Opanin Gyamfi
Cooperative communication strategies have long been touted as a solution to increase performance in wireless networks. We propose a simple form of coded cooperation using block codes and optimized selection at the relay node in this paper. The optimized distributed linear block codes for single relay cooperative technique enhances the performance by extending the maximum free distance of the overall code. MATLAB simulations proved that the proposed scheme outperforms the sub-optimal cooperation where there is no optimized selection at the relay node.
{"title":"Optimized Distributed Linear Block Codes For Single Relay Cooperative Wireless Communications","authors":"Gershon Rodor, Yang Fengfan, Stephen Nkyi, Harit Rijal, Enoch Opanin Gyamfi","doi":"10.1109/ICCWAMTIP53232.2021.9673710","DOIUrl":"https://doi.org/10.1109/ICCWAMTIP53232.2021.9673710","url":null,"abstract":"Cooperative communication strategies have long been touted as a solution to increase performance in wireless networks. We propose a simple form of coded cooperation using block codes and optimized selection at the relay node in this paper. The optimized distributed linear block codes for single relay cooperative technique enhances the performance by extending the maximum free distance of the overall code. MATLAB simulations proved that the proposed scheme outperforms the sub-optimal cooperation where there is no optimized selection at the relay node.","PeriodicalId":358772,"journal":{"name":"2021 18th International Computer Conference on Wavelet Active Media Technology and Information Processing (ICCWAMTIP)","volume":"71 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123836584","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 : 2021-12-17DOI: 10.1109/ICCWAMTIP53232.2021.9674069
Zhou Zihan, Chen Lirong, Zhang Haitao, Z. Fan
With the development of intelligent connected vehicles, in-vehicle Ethernet technologies are more and more commonly applied in automotive industry, such that cybersecurity related problems of in-vehicle Ethernet are becoming more and more obvious and should not be ignored. This paper studies an embedded intrusion detection technology for in-vehicle Ethernet. Based on rule matching a new binary rule format is designed, which can be converted with the current mainstream Snort/Suricata rule set. This technology is applied to our intrusion detection system ETH-IDS, which is a vehicle-oriented embedded Ethernet intrusion detection system and completely compliant with related AUTOSAR specifications. A multi-level comprehensive evaluation model is also proposed in the embedded environment to take the quantitative evaluation for intrusion detection systems. Related experiments are carried out in an automotive embedded environment, and the performance advantages of ETH-IDS are verified compared with Suricata. Meanwhile, the validity of the evaluation model is also verified.
{"title":"Research on Intrusion Detection Technology Based on Embedded Ethernet","authors":"Zhou Zihan, Chen Lirong, Zhang Haitao, Z. Fan","doi":"10.1109/ICCWAMTIP53232.2021.9674069","DOIUrl":"https://doi.org/10.1109/ICCWAMTIP53232.2021.9674069","url":null,"abstract":"With the development of intelligent connected vehicles, in-vehicle Ethernet technologies are more and more commonly applied in automotive industry, such that cybersecurity related problems of in-vehicle Ethernet are becoming more and more obvious and should not be ignored. This paper studies an embedded intrusion detection technology for in-vehicle Ethernet. Based on rule matching a new binary rule format is designed, which can be converted with the current mainstream Snort/Suricata rule set. This technology is applied to our intrusion detection system ETH-IDS, which is a vehicle-oriented embedded Ethernet intrusion detection system and completely compliant with related AUTOSAR specifications. A multi-level comprehensive evaluation model is also proposed in the embedded environment to take the quantitative evaluation for intrusion detection systems. Related experiments are carried out in an automotive embedded environment, and the performance advantages of ETH-IDS are verified compared with Suricata. Meanwhile, the validity of the evaluation model is also verified.","PeriodicalId":358772,"journal":{"name":"2021 18th International Computer Conference on Wavelet Active Media Technology and Information Processing (ICCWAMTIP)","volume":"27 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121392865","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}
This paper identified the emerging developmental trends of Ghana, an African country, based on online and offline textual, video and audio data sources from 2008 to 2018. In all, about 9,501 text files were collected and where needed, transcribed from these sources and content analyzed by series of text mining processes to bring out the convention for all the combined years. Topic modeling was the main adopted text mining technique. We initiated the topic modeling with the term frequency approach. The results from the mining processes were conceptually presented in relative frequency table and visually summarized as word clouds to map meaningful findings that were logically and humanly understandable. Through this approach, 45 major developmental themes emerged of which education seems to be the most frequently discussed trend in Ghanaian politics and government. In conclusion governments can detect developmental layouts from textual data.
{"title":"Text Mining A Decade Of Focal Development Trends In An African Country","authors":"Opoku-Mensah Nelson, Qin Zhiguang, Gyamfi Enoch Opanin, Danso Juliana Mantebea, Nyame Gabriel","doi":"10.1109/ICCWAMTIP53232.2021.9674166","DOIUrl":"https://doi.org/10.1109/ICCWAMTIP53232.2021.9674166","url":null,"abstract":"This paper identified the emerging developmental trends of Ghana, an African country, based on online and offline textual, video and audio data sources from 2008 to 2018. In all, about 9,501 text files were collected and where needed, transcribed from these sources and content analyzed by series of text mining processes to bring out the convention for all the combined years. Topic modeling was the main adopted text mining technique. We initiated the topic modeling with the term frequency approach. The results from the mining processes were conceptually presented in relative frequency table and visually summarized as word clouds to map meaningful findings that were logically and humanly understandable. Through this approach, 45 major developmental themes emerged of which education seems to be the most frequently discussed trend in Ghanaian politics and government. In conclusion governments can detect developmental layouts from textual data.","PeriodicalId":358772,"journal":{"name":"2021 18th International Computer Conference on Wavelet Active Media Technology and Information Processing (ICCWAMTIP)","volume":"2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115850025","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 : 2021-12-17DOI: 10.1109/ICCWAMTIP53232.2021.9674113
Mustafa R. Kadhim, WEN-HONG Tian, Guangyao Zhou, Tahseen Khan
Many clustering and cluster ensemble models have been proposed recently and have not addressed two concerns; when a single model is executed multiple times on a dataset, it predicts various labels for each data object; however, these various labels have a small correctness ratio due to the randomness in generating values in each implementation. Further, detecting which the correct label from these diverse answers is complicated, specifically when the unsupervised model works on a real-world application and needs to deliver a single correct label to the user. In this work considered these two issues by proposing a novel unsupervised constraints termed Inherited Constraints (IC) that behaves as semi-supervised constraints generation. Moreover, execute the IC needs a cluster model to utilize; thus, we proposed an unsupervised cluster ensemble model by integrating the Density Peaks cluster ensemble framework (DPE) and IC to improve the performance. This model is termed DPEIC. Further, we proposed a model termed Answer Settlements (AS) to detect a single correct label for each data object from the diverse answers obtained by DPEIC after utilized multiple times to consider the most duplicated labels as the correct ones. We compare DPEIC-AS with several state-of-the-arts to validate the strengths of this work. The experimental results indicate that DPEIC-AS outperforms the compared models at a different rate, ranging from 3% to 93%. Also, The AS assisted two state-of-the-arts methods to detect the correct labels with the highest possibilities from diverse answers.
{"title":"A Novel Side-Information for Unsupervised Cluster Ensemble","authors":"Mustafa R. Kadhim, WEN-HONG Tian, Guangyao Zhou, Tahseen Khan","doi":"10.1109/ICCWAMTIP53232.2021.9674113","DOIUrl":"https://doi.org/10.1109/ICCWAMTIP53232.2021.9674113","url":null,"abstract":"Many clustering and cluster ensemble models have been proposed recently and have not addressed two concerns; when a single model is executed multiple times on a dataset, it predicts various labels for each data object; however, these various labels have a small correctness ratio due to the randomness in generating values in each implementation. Further, detecting which the correct label from these diverse answers is complicated, specifically when the unsupervised model works on a real-world application and needs to deliver a single correct label to the user. In this work considered these two issues by proposing a novel unsupervised constraints termed Inherited Constraints (IC) that behaves as semi-supervised constraints generation. Moreover, execute the IC needs a cluster model to utilize; thus, we proposed an unsupervised cluster ensemble model by integrating the Density Peaks cluster ensemble framework (DPE) and IC to improve the performance. This model is termed DPEIC. Further, we proposed a model termed Answer Settlements (AS) to detect a single correct label for each data object from the diverse answers obtained by DPEIC after utilized multiple times to consider the most duplicated labels as the correct ones. We compare DPEIC-AS with several state-of-the-arts to validate the strengths of this work. The experimental results indicate that DPEIC-AS outperforms the compared models at a different rate, ranging from 3% to 93%. Also, The AS assisted two state-of-the-arts methods to detect the correct labels with the highest possibilities from diverse answers.","PeriodicalId":358772,"journal":{"name":"2021 18th International Computer Conference on Wavelet Active Media Technology and Information Processing (ICCWAMTIP)","volume":"55 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133830738","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 : 2021-12-17DOI: 10.1109/ICCWAMTIP53232.2021.9674107
Di Liu, Fei Cheng
With the development of network detection models, researchers have achieved good results in general target detection, but there is still no good solution for small target detection in images, especially the feature processing of small targets. At present, the most suitable feature processing method is FPN, but this fusion method will cause the feature redundancy, ambiguity and inaccuracy of small targets, and has little effect on the general large targets, but it will cause great interference and errors in the detection of small targets. For the above problems, this paper improves FPN and proposes a new SRM-FPN feature fusion method. Specifically, SRM is a spatial refinement model that learns the location of future feature points according to the context features between adjacent layers and content, while borrowing the adaptive channel merging method of the attention mechanism to optimize feature fusion. Compared with other methods, the optimized scheme combined with the existing model can effectively improve the detection effect of small targets in the image.
{"title":"SRM-FPN: A Small Target Detection Method Based on FPN Optimized Feature","authors":"Di Liu, Fei Cheng","doi":"10.1109/ICCWAMTIP53232.2021.9674107","DOIUrl":"https://doi.org/10.1109/ICCWAMTIP53232.2021.9674107","url":null,"abstract":"With the development of network detection models, researchers have achieved good results in general target detection, but there is still no good solution for small target detection in images, especially the feature processing of small targets. At present, the most suitable feature processing method is FPN, but this fusion method will cause the feature redundancy, ambiguity and inaccuracy of small targets, and has little effect on the general large targets, but it will cause great interference and errors in the detection of small targets. For the above problems, this paper improves FPN and proposes a new SRM-FPN feature fusion method. Specifically, SRM is a spatial refinement model that learns the location of future feature points according to the context features between adjacent layers and content, while borrowing the adaptive channel merging method of the attention mechanism to optimize feature fusion. Compared with other methods, the optimized scheme combined with the existing model can effectively improve the detection effect of small targets in the image.","PeriodicalId":358772,"journal":{"name":"2021 18th International Computer Conference on Wavelet Active Media Technology and Information Processing (ICCWAMTIP)","volume":"28 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134329483","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：