Pub Date : 2022-07-22DOI: 10.1109/ISPDS56360.2022.9874182
Shuhang Chen, Ziyang Luo, X. Li, Runhua He
At present, most of the tobacco quality inspection links use manual, but due to the speed and detection accuracy is not high enough, often lead to a very long quality inspection links, and mistakenly checked cigarettes into the market will cause economic losses to the company. In order to improve the speed and accuracy of cigarette inspection, a ZYNQ cigarette bar defect detection system was designed and implemented. After binarization processing and image filtering, Sobel operator is used to draw the contour of the image, and then Hough transform is used to get the image of the end face broken line. After rotation correction of the image, the judgment of defect detection is made. If the defect type is determined, the defective products are separated by sound and light alarm and automatic smoke separation device. The experiment shows that the average detection speed of the system for cigarette bar defects is less than 40ms, which meets the real-time requirements of the system. The detection accuracy is 98.67%, and the false detection rate is 0.05%, with low false detection rate.
{"title":"Research on Cigarette strip defect Detection System based on ZYNQ","authors":"Shuhang Chen, Ziyang Luo, X. Li, Runhua He","doi":"10.1109/ISPDS56360.2022.9874182","DOIUrl":"https://doi.org/10.1109/ISPDS56360.2022.9874182","url":null,"abstract":"At present, most of the tobacco quality inspection links use manual, but due to the speed and detection accuracy is not high enough, often lead to a very long quality inspection links, and mistakenly checked cigarettes into the market will cause economic losses to the company. In order to improve the speed and accuracy of cigarette inspection, a ZYNQ cigarette bar defect detection system was designed and implemented. After binarization processing and image filtering, Sobel operator is used to draw the contour of the image, and then Hough transform is used to get the image of the end face broken line. After rotation correction of the image, the judgment of defect detection is made. If the defect type is determined, the defective products are separated by sound and light alarm and automatic smoke separation device. The experiment shows that the average detection speed of the system for cigarette bar defects is less than 40ms, which meets the real-time requirements of the system. The detection accuracy is 98.67%, and the false detection rate is 0.05%, with low false detection rate.","PeriodicalId":280244,"journal":{"name":"2022 3rd International Conference on Information Science, Parallel and Distributed Systems (ISPDS)","volume":"118 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121445446","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 four-chamber view is the primary ultrasound images that clinicians diagnose whether a fetus has congenital heart disease (CHD) in the process of prenatal diagnosis and screening, which can provide clinicians with a clear view of the developmental morphology of the fetal four chambers (i.e., left atrium, left ventricle, right atrium, and right ventricle). The early diagnosis and screening for CHD depend on the clinicians' experience to a large extent. Deep learning technology has achieved great success in medical image analysis. Hence, applying deep learning technology in the four-chamber view analysis can help improve the diagnostic accuracy of CHD and make it more objective. Hence, we design a deep learning-based intelligent analysis platform (DLIAP) for fetal ultrasound four-chamber views, which includes an image input module, an image analysis module, a visualization output module, and an information query module. The DLIAP can assist the clinicians in objectively analyzing the fetal ultrasound four-chamber views and further improve the diagnostic accuracy of CHD.
{"title":"A deep learning-based intelligent analysis platform for fetal ultrasound four-chamber views","authors":"Sibo Qiao, Shanchen Pang, Yukun Dong, Haiyuan Gui, Qiwen Yuan, Zelong Zheng, Guoxuan Cui","doi":"10.1109/ISPDS56360.2022.9874029","DOIUrl":"https://doi.org/10.1109/ISPDS56360.2022.9874029","url":null,"abstract":"The four-chamber view is the primary ultrasound images that clinicians diagnose whether a fetus has congenital heart disease (CHD) in the process of prenatal diagnosis and screening, which can provide clinicians with a clear view of the developmental morphology of the fetal four chambers (i.e., left atrium, left ventricle, right atrium, and right ventricle). The early diagnosis and screening for CHD depend on the clinicians' experience to a large extent. Deep learning technology has achieved great success in medical image analysis. Hence, applying deep learning technology in the four-chamber view analysis can help improve the diagnostic accuracy of CHD and make it more objective. Hence, we design a deep learning-based intelligent analysis platform (DLIAP) for fetal ultrasound four-chamber views, which includes an image input module, an image analysis module, a visualization output module, and an information query module. The DLIAP can assist the clinicians in objectively analyzing the fetal ultrasound four-chamber views and further improve the diagnostic accuracy of CHD.","PeriodicalId":280244,"journal":{"name":"2022 3rd International Conference on Information Science, Parallel and Distributed Systems (ISPDS)","volume":"7 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126316224","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 : 2022-07-22DOI: 10.1109/ISPDS56360.2022.9874174
Yingqi Jiang, Lili Dong, Chang Tian
Maritime distress accidents usually occur in severe wind and wave environments. Infrared image enhancement technology can provide high-quality images for follow-up search and rescue work and has significant research value. This paper first analyzes the features of maritime infrared target images. According to the grayscale Gaussian distribution shape and gradient texture directionality of the target area, designs a target feature description operator and extracts the target feature image as the guide image for guided filtering. Then, the difference operation is performed between the original image and the filtering result, and the target detail layer that suppresses background noise and is not distorted is obtained. Finally, the target layer and the original image are fused with appropriate weights to obtain an enhanced image that retains the characteristics of the natural environment. The experimental results show that the method can effectively improve the clarity of the image and the detectability of the target.
{"title":"Image Enhancement of Maritime Infrared Targets Based on Joint Features of Grayscale and Texture","authors":"Yingqi Jiang, Lili Dong, Chang Tian","doi":"10.1109/ISPDS56360.2022.9874174","DOIUrl":"https://doi.org/10.1109/ISPDS56360.2022.9874174","url":null,"abstract":"Maritime distress accidents usually occur in severe wind and wave environments. Infrared image enhancement technology can provide high-quality images for follow-up search and rescue work and has significant research value. This paper first analyzes the features of maritime infrared target images. According to the grayscale Gaussian distribution shape and gradient texture directionality of the target area, designs a target feature description operator and extracts the target feature image as the guide image for guided filtering. Then, the difference operation is performed between the original image and the filtering result, and the target detail layer that suppresses background noise and is not distorted is obtained. Finally, the target layer and the original image are fused with appropriate weights to obtain an enhanced image that retains the characteristics of the natural environment. The experimental results show that the method can effectively improve the clarity of the image and the detectability of the target.","PeriodicalId":280244,"journal":{"name":"2022 3rd International Conference on Information Science, Parallel and Distributed Systems (ISPDS)","volume":"32 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122278102","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 : 2022-07-22DOI: 10.1109/ispds56360.2022.9874044
Hongjie Geng, Mingming Zhou
Film and television production is one of the sources of modern entertainment and key content. With the rapid development and popularization of digital technology, its application proportion in film and television production is constantly increasing. AE keying technology is a digital production technology that is frequently used in film and television post-production, and its role is very important. Through the application of AE keying technology, film and television post-production is more convenient and efficient, and keying synthesis technology involves a wide range, which is of great value in the industry. Based on this, this paper introduces the meaning of film and television post-production and AE keying technology, and introduces the plug-in technology of AE. At the same time, it introduces the specific application of AE keying technology from three aspects: brightness keying, chroma keying and Key light keying, so as to provide help for film and television post-production personnel.
{"title":"Application of AE keying technology in film and television post-production","authors":"Hongjie Geng, Mingming Zhou","doi":"10.1109/ispds56360.2022.9874044","DOIUrl":"https://doi.org/10.1109/ispds56360.2022.9874044","url":null,"abstract":"Film and television production is one of the sources of modern entertainment and key content. With the rapid development and popularization of digital technology, its application proportion in film and television production is constantly increasing. AE keying technology is a digital production technology that is frequently used in film and television post-production, and its role is very important. Through the application of AE keying technology, film and television post-production is more convenient and efficient, and keying synthesis technology involves a wide range, which is of great value in the industry. Based on this, this paper introduces the meaning of film and television post-production and AE keying technology, and introduces the plug-in technology of AE. At the same time, it introduces the specific application of AE keying technology from three aspects: brightness keying, chroma keying and Key light keying, so as to provide help for film and television post-production personnel.","PeriodicalId":280244,"journal":{"name":"2022 3rd International Conference on Information Science, Parallel and Distributed Systems (ISPDS)","volume":"18 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114244912","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 : 2022-07-22DOI: 10.1109/ISPDS56360.2022.9874092
Wujisiguleng Zhao, Chunyi Chen, Hongmei Meng
Monte Carlo simulation is the most widely used method to analyze and study the optical transmission characteristics of seawater channel. The scattering phase function is usually used to represent the specific scattering characteristics of seawater according to the different densities of seawater medium particles, so it is particularly important to measure the scattering phase cosine $costheta$ in the scattering phase function of the average particles in seawater. This chapter uses Monte Carlo simulation method to solve the scattering phase cosine $costheta$, and explores a suitable and efficient sampling method to simulate light scattering. First, we sample the Rayleigh phase function by using the inverse transformation, the tabulation method and the weighted algorithm, and analyze the advantages and disadvantages of these three sampling methods. The three methods are applied to the Henyey-Greentein (HG) phase function. Moreover, in order to simulate the scattering characteristics in natural media, the Henyey-Greentein (HG) phase function is improved to obtain the Rayleigh Henyey-Greenstein (RHG) phase function, and then the corresponding sampling is carried out to improve the efficiency of the sampling method.
{"title":"Study on Sampling Method of Scattering Phase Function of Medium","authors":"Wujisiguleng Zhao, Chunyi Chen, Hongmei Meng","doi":"10.1109/ISPDS56360.2022.9874092","DOIUrl":"https://doi.org/10.1109/ISPDS56360.2022.9874092","url":null,"abstract":"Monte Carlo simulation is the most widely used method to analyze and study the optical transmission characteristics of seawater channel. The scattering phase function is usually used to represent the specific scattering characteristics of seawater according to the different densities of seawater medium particles, so it is particularly important to measure the scattering phase cosine $costheta$ in the scattering phase function of the average particles in seawater. This chapter uses Monte Carlo simulation method to solve the scattering phase cosine $costheta$, and explores a suitable and efficient sampling method to simulate light scattering. First, we sample the Rayleigh phase function by using the inverse transformation, the tabulation method and the weighted algorithm, and analyze the advantages and disadvantages of these three sampling methods. The three methods are applied to the Henyey-Greentein (HG) phase function. Moreover, in order to simulate the scattering characteristics in natural media, the Henyey-Greentein (HG) phase function is improved to obtain the Rayleigh Henyey-Greenstein (RHG) phase function, and then the corresponding sampling is carried out to improve the efficiency of the sampling method.","PeriodicalId":280244,"journal":{"name":"2022 3rd International Conference on Information Science, Parallel and Distributed Systems (ISPDS)","volume":"139 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123659519","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 : 2022-07-22DOI: 10.1109/ISPDS56360.2022.9874122
Jun Yu Li, Yuejun Pan, Hao Wang, Y. Yuan, T. Guan
In order to design braces that are more in line with patient characteristics, and help clinicians achieve rapid and accurate diagnosis and treatment. Starting from practical application, this paper crawls AIS brace-related knowledge from medical websites, combines electronic cases and expert knowledge, builds AIS brace knowledge graph, and summarizes the main knowledge of AIS. Due to the complexity of the knowledge of AIS braces, this paper proposes a joint entity and relation extraction method based on the FS-E-BIESO annotation method. By comparing the two knowledge extraction algorithms BERT-BiLSTM-CRF and BiLSTM-CRF, it is concluded that BiLSTM-CRF has a better F1 value. By comparing the two knowledge extraction algorithms BERT-BiLSTM-CRF and BiLSTM-CRF, it is concluded that BiLSTM-CRF has a better F1 value. The extracted knowledge is merged to eliminate the interference knowledge, and imported into neo4j in the form of triples to construct the knowledge graph of AIS orthopedic braces.
{"title":"Research on the construction of knowledge graph of AIS orthopedic braces","authors":"Jun Yu Li, Yuejun Pan, Hao Wang, Y. Yuan, T. Guan","doi":"10.1109/ISPDS56360.2022.9874122","DOIUrl":"https://doi.org/10.1109/ISPDS56360.2022.9874122","url":null,"abstract":"In order to design braces that are more in line with patient characteristics, and help clinicians achieve rapid and accurate diagnosis and treatment. Starting from practical application, this paper crawls AIS brace-related knowledge from medical websites, combines electronic cases and expert knowledge, builds AIS brace knowledge graph, and summarizes the main knowledge of AIS. Due to the complexity of the knowledge of AIS braces, this paper proposes a joint entity and relation extraction method based on the FS-E-BIESO annotation method. By comparing the two knowledge extraction algorithms BERT-BiLSTM-CRF and BiLSTM-CRF, it is concluded that BiLSTM-CRF has a better F1 value. By comparing the two knowledge extraction algorithms BERT-BiLSTM-CRF and BiLSTM-CRF, it is concluded that BiLSTM-CRF has a better F1 value. The extracted knowledge is merged to eliminate the interference knowledge, and imported into neo4j in the form of triples to construct the knowledge graph of AIS orthopedic braces.","PeriodicalId":280244,"journal":{"name":"2022 3rd International Conference on Information Science, Parallel and Distributed Systems (ISPDS)","volume":"78 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122035678","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 : 2022-07-22DOI: 10.1109/ISPDS56360.2022.9874125
Xin Zhang, Lu Ding, Zhaohui Xu, Hui Liu
High speed dim and small target detection is an important technology in CCD vertical target coordinate measurement. Its difficulty lies in the high frame rate real-time image processing speed requirements, weak and small target capture rate and extraction accuracy is not high [1]. In order to solve these problems, FPGA is designed and applied as the core of embedded hardware platform, and high-efficiency parallel operation, background iteration and false target detection algorithm are used to realize the real-time detection of high-speed weak and small targets in CDD images with a frame rate of 4096 lines up to 50KHz. The time delay of target acquisition and output measurement results is less than 10 ms, and the real-time performance is very good. In a certain application, under the background illumination of sky, the capture rate of dim high-speed projectile (5.8 mm projectile) can reach 100%, and the measurement accuracy $sigma$ is less than 13 mm, and the acquisition rate test of targets larger than 5.8 mm reaches a higher standard.
{"title":"Design of Real-time Target Detection System in CCD Vertical Target Coordinate Measurement","authors":"Xin Zhang, Lu Ding, Zhaohui Xu, Hui Liu","doi":"10.1109/ISPDS56360.2022.9874125","DOIUrl":"https://doi.org/10.1109/ISPDS56360.2022.9874125","url":null,"abstract":"High speed dim and small target detection is an important technology in CCD vertical target coordinate measurement. Its difficulty lies in the high frame rate real-time image processing speed requirements, weak and small target capture rate and extraction accuracy is not high [1]. In order to solve these problems, FPGA is designed and applied as the core of embedded hardware platform, and high-efficiency parallel operation, background iteration and false target detection algorithm are used to realize the real-time detection of high-speed weak and small targets in CDD images with a frame rate of 4096 lines up to 50KHz. The time delay of target acquisition and output measurement results is less than 10 ms, and the real-time performance is very good. In a certain application, under the background illumination of sky, the capture rate of dim high-speed projectile (5.8 mm projectile) can reach 100%, and the measurement accuracy $sigma$ is less than 13 mm, and the acquisition rate test of targets larger than 5.8 mm reaches a higher standard.","PeriodicalId":280244,"journal":{"name":"2022 3rd International Conference on Information Science, Parallel and Distributed Systems (ISPDS)","volume":"23 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114851272","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 : 2022-07-22DOI: 10.1109/ISPDS56360.2022.9874105
Yifan Xu, Yong Bai
In recent years, object detection has been expanded to drone scenes, where remote sensing images contain a greater variety and arbitrary-oriented targets. In order to solve the problem of detection difficulty and computational intensity for remote sensing images, oriented object detection is needed and the network model is expected to be deployed on resource-limited devices. This paper proposes a lightweight object detection method for oriented object detection by leveraging and compressing YOLOv5 network model. We integrate the fine-tuning stage in network slimming with knowledge distillation to enhance the accuracy of the detection model and save training time by transferring the important feature information to the student network. Loss function is redesigned by combining Theta loss with other detection and distillation losses to make the compression model more accurate. Extensive experiments are conducted to verify the effectiveness of our proposed method on the remote sensing public dataset DOTA. The compressed model achieves an accuracy of 76.18% on the DOTA dataset, 1.7% increase compared to the original YOLOv5 model. The FLOPs are decreased by 37.0%, the number of parameters is decreased by 58.9%, the weight file size is decreased by 57.6%, and the inference time is decreased by 17.4%.
{"title":"Compressed YOLOv5 for Oriented Object Detection with Integrated Network Slimming and Knowledge Distillation","authors":"Yifan Xu, Yong Bai","doi":"10.1109/ISPDS56360.2022.9874105","DOIUrl":"https://doi.org/10.1109/ISPDS56360.2022.9874105","url":null,"abstract":"In recent years, object detection has been expanded to drone scenes, where remote sensing images contain a greater variety and arbitrary-oriented targets. In order to solve the problem of detection difficulty and computational intensity for remote sensing images, oriented object detection is needed and the network model is expected to be deployed on resource-limited devices. This paper proposes a lightweight object detection method for oriented object detection by leveraging and compressing YOLOv5 network model. We integrate the fine-tuning stage in network slimming with knowledge distillation to enhance the accuracy of the detection model and save training time by transferring the important feature information to the student network. Loss function is redesigned by combining Theta loss with other detection and distillation losses to make the compression model more accurate. Extensive experiments are conducted to verify the effectiveness of our proposed method on the remote sensing public dataset DOTA. The compressed model achieves an accuracy of 76.18% on the DOTA dataset, 1.7% increase compared to the original YOLOv5 model. The FLOPs are decreased by 37.0%, the number of parameters is decreased by 58.9%, the weight file size is decreased by 57.6%, and the inference time is decreased by 17.4%.","PeriodicalId":280244,"journal":{"name":"2022 3rd International Conference on Information Science, Parallel and Distributed Systems (ISPDS)","volume":"62 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130772091","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 : 2022-07-22DOI: 10.1109/ISPDS56360.2022.9874095
Wenyu Huo, Xisheng Li, Shengcheng Wang, Jia You
A temperature measurement compensation algorithm based on particle swarm optimization (PSO) back propagation (BP) neural network is proposed for the temperature measurement accuracy of infrared thermal imager affected by ambient temperature, measurement distance and other factors. By optimizing the initial weight and threshold of BP neural network, PSO algorithm overcomes the shortcomings of BP algorithm, such as slow convergence speed, easy to fall into local optimization and low accuracy. At the same time, the inertia weight is introduced into the PSO-BP algorithm, so that the algorithm maintains a strong global search ability and a more accurate local search ability. Compared with the single BP algorithm, the generalization ability and temperature measurement accuracy of the system are effectively improved, and the average value of the mean square error is reduced to 0.0443, which achieves the ideal effect.
{"title":"Design on Infrared Temperature Measurement Compensation Algorithm Based on PSO-BP Neural Network","authors":"Wenyu Huo, Xisheng Li, Shengcheng Wang, Jia You","doi":"10.1109/ISPDS56360.2022.9874095","DOIUrl":"https://doi.org/10.1109/ISPDS56360.2022.9874095","url":null,"abstract":"A temperature measurement compensation algorithm based on particle swarm optimization (PSO) back propagation (BP) neural network is proposed for the temperature measurement accuracy of infrared thermal imager affected by ambient temperature, measurement distance and other factors. By optimizing the initial weight and threshold of BP neural network, PSO algorithm overcomes the shortcomings of BP algorithm, such as slow convergence speed, easy to fall into local optimization and low accuracy. At the same time, the inertia weight is introduced into the PSO-BP algorithm, so that the algorithm maintains a strong global search ability and a more accurate local search ability. Compared with the single BP algorithm, the generalization ability and temperature measurement accuracy of the system are effectively improved, and the average value of the mean square error is reduced to 0.0443, which achieves the ideal effect.","PeriodicalId":280244,"journal":{"name":"2022 3rd International Conference on Information Science, Parallel and Distributed Systems (ISPDS)","volume":"39 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128720119","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 : 2022-07-22DOI: 10.1109/ISPDS56360.2022.9874076
Fei Li
In order to improve the accuracy of sentiment classification of online product reviews, a model for sentiment analysis of unbalanced reviews is proposed. Firstly, the LDA model is used to balance the original review text set, and then the word vector model and convolution neural network are combined to construct the review text vectorization feature extraction process to obtain the word feature vector, which is used as the input matrix of the balanced review set. Finally, the BiLSTM algorithm is used for sentiment classification to obtain product reviews of positive and negative sentiment categories. The results show that LDA sampling unbalance processing method is a high accuracy unbalanced text processing method. BiLSTM algorithm has better effect of sentiment classification than other deep learning algorithms. CNN-BiLSTM model based on LDA unbalance processing obtains the optimal model performance evaluation index value in the comparative experiment of different sentiment classification models, which verifies the advantages and effectiveness of the model and effectively realizes sentiment analysis on unbalanced review texts.
{"title":"An Sentiment Analysis Model of Online Product Reviews Based on Deep Learning","authors":"Fei Li","doi":"10.1109/ISPDS56360.2022.9874076","DOIUrl":"https://doi.org/10.1109/ISPDS56360.2022.9874076","url":null,"abstract":"In order to improve the accuracy of sentiment classification of online product reviews, a model for sentiment analysis of unbalanced reviews is proposed. Firstly, the LDA model is used to balance the original review text set, and then the word vector model and convolution neural network are combined to construct the review text vectorization feature extraction process to obtain the word feature vector, which is used as the input matrix of the balanced review set. Finally, the BiLSTM algorithm is used for sentiment classification to obtain product reviews of positive and negative sentiment categories. The results show that LDA sampling unbalance processing method is a high accuracy unbalanced text processing method. BiLSTM algorithm has better effect of sentiment classification than other deep learning algorithms. CNN-BiLSTM model based on LDA unbalance processing obtains the optimal model performance evaluation index value in the comparative experiment of different sentiment classification models, which verifies the advantages and effectiveness of the model and effectively realizes sentiment analysis on unbalanced review texts.","PeriodicalId":280244,"journal":{"name":"2022 3rd International Conference on Information Science, Parallel and Distributed Systems (ISPDS)","volume":"2011 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130069825","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：