Pub Date : 2021-10-13DOI: 10.1109/icisfall51598.2021.9627337
Sen Wang, Peng Li, Wei-hua Niu
The complex equipment such as aeroengines has a complicated internal structure. Due to long-term exposure to extremely harsh external environmental conditions such as temperature and pressure, aeroengines often have various forms of failure, which seriously affect the normal flight of the aircraft. It is difficult for traditional models to extract accurate fault information from complex vibration signals, which increases the difficulty of troubleshooting for aircraft engines. Aiming at this problem, a fault diagnosis model using the combination of variational mode decomposition and convolutional neural network is proposed. First, the original signal is decomposed by variational mode decomposition, and then the decomposed signal is reconstructed into a two-dimensional characteristic matrix. Finally, the reconstructed matrix is used as the input of the convolutional neural network to realize the classification of typical failure modes. Compared with the traditional method, this method can extract the internal fault characteristics of the vibration signal better, and the fault recognition accuracy rate is higher.
{"title":"Research on Fault Diagnosis Model of Convolutional Neural Network Based on Signal Decomposition","authors":"Sen Wang, Peng Li, Wei-hua Niu","doi":"10.1109/icisfall51598.2021.9627337","DOIUrl":"https://doi.org/10.1109/icisfall51598.2021.9627337","url":null,"abstract":"The complex equipment such as aeroengines has a complicated internal structure. Due to long-term exposure to extremely harsh external environmental conditions such as temperature and pressure, aeroengines often have various forms of failure, which seriously affect the normal flight of the aircraft. It is difficult for traditional models to extract accurate fault information from complex vibration signals, which increases the difficulty of troubleshooting for aircraft engines. Aiming at this problem, a fault diagnosis model using the combination of variational mode decomposition and convolutional neural network is proposed. First, the original signal is decomposed by variational mode decomposition, and then the decomposed signal is reconstructed into a two-dimensional characteristic matrix. Finally, the reconstructed matrix is used as the input of the convolutional neural network to realize the classification of typical failure modes. Compared with the traditional method, this method can extract the internal fault characteristics of the vibration signal better, and the fault recognition accuracy rate is higher.","PeriodicalId":240142,"journal":{"name":"2021 IEEE/ACIS 20th International Fall Conference on Computer and Information Science (ICIS Fall)","volume":"50 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114670500","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}
Cooperation is an obvious intelligence feature for Swarm Intelligent Systems (SIS), and where, autonomous task cooperation is one of the core issues for such systems that run in complicated environments, such as factories and aerospace. Focusing on this topic, the mechanisms for modeling and coordinating tasks are deeply studied in this article. After analyzing typical features of task cooperation, a hierarchical architecture of task models is firstly proposed and established, where mission, sub-mission and atom task models are designed, from the top-level to the bottom. Meanwhile, a policy-based parsing mechanism is proposed to autonomously decompose a mission into a set of atom tasks with special constraints. On this basis, resource related entities for various atom tasks are designed uniformly, and for the purpose of behavior synchronization, the concept of virtual logic task is further introduced and Topic-based communication channel is adopted over DDS middleware. Finally, a typical example is given and analyzed. In general, this work provides a feasible idea and method for constructing cooperative swarm intelligent systems.
{"title":"Multi-level Modeling and Cooperation Mechanisms of Tasks for Swarm Intelligent Systems","authors":"Peng Li, Wenwen Fu, Hongjun You, Yuxi Liu, Zhihao Wu, Ran Qu, Yahui Li, Kailong Zhang","doi":"10.1109/icisfall51598.2021.9627407","DOIUrl":"https://doi.org/10.1109/icisfall51598.2021.9627407","url":null,"abstract":"Cooperation is an obvious intelligence feature for Swarm Intelligent Systems (SIS), and where, autonomous task cooperation is one of the core issues for such systems that run in complicated environments, such as factories and aerospace. Focusing on this topic, the mechanisms for modeling and coordinating tasks are deeply studied in this article. After analyzing typical features of task cooperation, a hierarchical architecture of task models is firstly proposed and established, where mission, sub-mission and atom task models are designed, from the top-level to the bottom. Meanwhile, a policy-based parsing mechanism is proposed to autonomously decompose a mission into a set of atom tasks with special constraints. On this basis, resource related entities for various atom tasks are designed uniformly, and for the purpose of behavior synchronization, the concept of virtual logic task is further introduced and Topic-based communication channel is adopted over DDS middleware. Finally, a typical example is given and analyzed. In general, this work provides a feasible idea and method for constructing cooperative swarm intelligent systems.","PeriodicalId":240142,"journal":{"name":"2021 IEEE/ACIS 20th International Fall Conference on Computer and Information Science (ICIS Fall)","volume":"23 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123410163","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 : 2021-10-13DOI: 10.1109/icisfall51598.2021.9627396
Haixiang Wang, Pencheng Wen, L. Bai
In this paper, we study the obstacle avoidance method of swarm UAVs. This method is used to avoid the longtime split of the formation when the swarm passes through the obstacle area. A control method of the dense formation is designed based on the behavioral approach with the constraints of formation boundary. This distributed method just needs each UAV to communicate with neighboring individuals of the swarm. A new route planning method based on Particle Swarm Optimization (PSO) algorithm is proposed to plan a safe and flyable route matching the formation width for swarm UAVs in the area with obstacles. The planned route serves as the consensus information of the swarm, which is equivalent to a virtual UAV. During avoiding obstacles, swarm UAVs are treated as a whole, and the swarm forms a dense formation by following the planned route. Simulation results are presented to demonstrate the effectiveness and rationality of the proposed method.
{"title":"A Distributed Obstacle Avoidance Method for Swarm UAVs based on Behavioral Approach and Route Planning","authors":"Haixiang Wang, Pencheng Wen, L. Bai","doi":"10.1109/icisfall51598.2021.9627396","DOIUrl":"https://doi.org/10.1109/icisfall51598.2021.9627396","url":null,"abstract":"In this paper, we study the obstacle avoidance method of swarm UAVs. This method is used to avoid the longtime split of the formation when the swarm passes through the obstacle area. A control method of the dense formation is designed based on the behavioral approach with the constraints of formation boundary. This distributed method just needs each UAV to communicate with neighboring individuals of the swarm. A new route planning method based on Particle Swarm Optimization (PSO) algorithm is proposed to plan a safe and flyable route matching the formation width for swarm UAVs in the area with obstacles. The planned route serves as the consensus information of the swarm, which is equivalent to a virtual UAV. During avoiding obstacles, swarm UAVs are treated as a whole, and the swarm forms a dense formation by following the planned route. Simulation results are presented to demonstrate the effectiveness and rationality of the proposed method.","PeriodicalId":240142,"journal":{"name":"2021 IEEE/ACIS 20th International Fall Conference on Computer and Information Science (ICIS Fall)","volume":"62 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122893299","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 : 2021-10-13DOI: 10.1109/icisfall51598.2021.9627460
Taozheng Zhang, Jiaqi Guo
Chinese sentiment analysis is a very important branch of natural language processing. It has been receiving much attention in recent years. The bidirectional long and short-term memory network (Bi-LSTM) model has been well applied in the field of sentiment analysis because of its own characteristics. This experiment hopes to further explore the performance and application of the Bi-LSTM model in sentiment analysis. There are three main steps in the experiment. First, the collected Chinese reviews are segmented and vectorized. Then, the Bi-LSTM is trained and tested. Finally, the sentiment analysis result is obtained. With the help of the hyper-parameter adjustment and the dropout mechanism, the evaluation indicators of the experimental model have reached about 89%. What's more, based on the same experimental environment and experimental data, this experiment tested the accuracy of CNN, LSTM, CNN_LSTM, and Bi-LSTM. In addition, the trained Bi-LSTM was used to analyze reviews from Taobao and JD.COM. The specific operation is to collect reviews on a certain product from Taobao and JD.COM to perform a specific analysis with the model. Then find the advantages and disadvantages of the model in practical applications, so that the model can continue to be improved.
{"title":"Research on Chinese Sentiment Analysis Based on Bi-LSTM Networks","authors":"Taozheng Zhang, Jiaqi Guo","doi":"10.1109/icisfall51598.2021.9627460","DOIUrl":"https://doi.org/10.1109/icisfall51598.2021.9627460","url":null,"abstract":"Chinese sentiment analysis is a very important branch of natural language processing. It has been receiving much attention in recent years. The bidirectional long and short-term memory network (Bi-LSTM) model has been well applied in the field of sentiment analysis because of its own characteristics. This experiment hopes to further explore the performance and application of the Bi-LSTM model in sentiment analysis. There are three main steps in the experiment. First, the collected Chinese reviews are segmented and vectorized. Then, the Bi-LSTM is trained and tested. Finally, the sentiment analysis result is obtained. With the help of the hyper-parameter adjustment and the dropout mechanism, the evaluation indicators of the experimental model have reached about 89%. What's more, based on the same experimental environment and experimental data, this experiment tested the accuracy of CNN, LSTM, CNN_LSTM, and Bi-LSTM. In addition, the trained Bi-LSTM was used to analyze reviews from Taobao and JD.COM. The specific operation is to collect reviews on a certain product from Taobao and JD.COM to perform a specific analysis with the model. Then find the advantages and disadvantages of the model in practical applications, so that the model can continue to be improved.","PeriodicalId":240142,"journal":{"name":"2021 IEEE/ACIS 20th International Fall Conference on Computer and Information Science (ICIS Fall)","volume":"20 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124799269","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 : 2021-10-13DOI: 10.1109/icisfall51598.2021.9627434
Jun Zhao, Changyu Wang, T. Ai, Yanfei Huang, Gang Zhao
Aiming at the demand of improving anti-interference and real-time performance of a new type laser semiactive four-quadrant proportional guidance device, this paper proposes a controller architecture based on ARM+FPGA which improves the real-time performance and complex code pattern detection ability of the system by increasing the co-processing ability, and adopts LQG/LTR algorithm to design the control strategy of the servo tracking system to improve the fast and stable tracking ability and robustness. The experimental results show that the pulse acquisition accuracy of the system reaches 30ns, the analog acquisition synchronization accuracy reaches 13ns, the analog acquisition accuracy reaches 0.3%, the gain control response time is less than 3.77 ms, and the pseudorandom code recognition delay is less than 100ms, which effectively improves the real-time performance and robustness of the laser semi-active proportional guidance device.
{"title":"Design of Controller for Laser Semi-Active Proportional Navigation Device Based on LQG/LTR","authors":"Jun Zhao, Changyu Wang, T. Ai, Yanfei Huang, Gang Zhao","doi":"10.1109/icisfall51598.2021.9627434","DOIUrl":"https://doi.org/10.1109/icisfall51598.2021.9627434","url":null,"abstract":"Aiming at the demand of improving anti-interference and real-time performance of a new type laser semiactive four-quadrant proportional guidance device, this paper proposes a controller architecture based on ARM+FPGA which improves the real-time performance and complex code pattern detection ability of the system by increasing the co-processing ability, and adopts LQG/LTR algorithm to design the control strategy of the servo tracking system to improve the fast and stable tracking ability and robustness. The experimental results show that the pulse acquisition accuracy of the system reaches 30ns, the analog acquisition synchronization accuracy reaches 13ns, the analog acquisition accuracy reaches 0.3%, the gain control response time is less than 3.77 ms, and the pseudorandom code recognition delay is less than 100ms, which effectively improves the real-time performance and robustness of the laser semi-active proportional guidance device.","PeriodicalId":240142,"journal":{"name":"2021 IEEE/ACIS 20th International Fall Conference on Computer and Information Science (ICIS Fall)","volume":"53 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134125071","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 : 2021-10-13DOI: 10.1109/icisfall51598.2021.9627364
M. Garcia-Ruiz, O. Alvarez-Cardenas, A. Iñiguez-Carrillo
The COVID-19 pandemic has changed the traditional teaching and learning process, moving students' educational hands-on activities carried out in the classroom to home activities that include the use of online tools. Here we describe an after-class online coding club conducted for a month, where elementary (primary) school students programmed and tested games running on the BBC Micro: bit microcontroller board. By developing gaming mini-projects, the students learned computing topics such as logic, sensors, random number generation, game development and programming, and how small physical computing mini-projects were conducted. In this paper, we describe how students and an instructor developed and tested games made for the Micro: bit, running an online simulator and physically using the Micro: bit at home. The paper shows lessons learned on developing games with the BBC Micro: bit in the coding club, and challenges that students encountered in the game development with the BBC Micro: bit. Future work includes multisensory educational gaming projects with the BBC Micro: bit.
{"title":"Experiences in Developing and Testing BBC Micro: bit Games in a K-12 Coding Club during the COVID-19 Pandemic","authors":"M. Garcia-Ruiz, O. Alvarez-Cardenas, A. Iñiguez-Carrillo","doi":"10.1109/icisfall51598.2021.9627364","DOIUrl":"https://doi.org/10.1109/icisfall51598.2021.9627364","url":null,"abstract":"The COVID-19 pandemic has changed the traditional teaching and learning process, moving students' educational hands-on activities carried out in the classroom to home activities that include the use of online tools. Here we describe an after-class online coding club conducted for a month, where elementary (primary) school students programmed and tested games running on the BBC Micro: bit microcontroller board. By developing gaming mini-projects, the students learned computing topics such as logic, sensors, random number generation, game development and programming, and how small physical computing mini-projects were conducted. In this paper, we describe how students and an instructor developed and tested games made for the Micro: bit, running an online simulator and physically using the Micro: bit at home. The paper shows lessons learned on developing games with the BBC Micro: bit in the coding club, and challenges that students encountered in the game development with the BBC Micro: bit. Future work includes multisensory educational gaming projects with the BBC Micro: bit.","PeriodicalId":240142,"journal":{"name":"2021 IEEE/ACIS 20th International Fall Conference on Computer and Information Science (ICIS Fall)","volume":"194 1-2","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114027204","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 : 2021-10-13DOI: 10.1109/icisfall51598.2021.9627486
Xianghao Meng, Dongmei Li, Qichen Han
The semantic relationship in thesaurus is introduced into the current network information retrieval tool, which can realize semantic retrieval. Using a statistical language model to express query statements and return results in the form of probability distribution can more effectively complete the construction of user model and realize personalized retrieval. Firstly, this paper proposes a similarity calculation method based on the relationship between words in the thesaurus. On the basis of this method, combined with the idea of query expansion and weighted sorting, this paper proposes a semantic retrieval method of forestry information based on the thesaurus. Secondly, this paper uses a statistical language model to propose personalized retrieval methods based on three different user models: topic model, historical model and mixed model. Finally, a forestry information personalized semantic retrieval system is realized by using semantic retrieval and personalized retrieval method. Experimental results indicate that the proposed personalized semantic retrieval method can effectively improve the retrieval performance.
{"title":"Personalized Semantic Retrieval System based on Statistical Language Model","authors":"Xianghao Meng, Dongmei Li, Qichen Han","doi":"10.1109/icisfall51598.2021.9627486","DOIUrl":"https://doi.org/10.1109/icisfall51598.2021.9627486","url":null,"abstract":"The semantic relationship in thesaurus is introduced into the current network information retrieval tool, which can realize semantic retrieval. Using a statistical language model to express query statements and return results in the form of probability distribution can more effectively complete the construction of user model and realize personalized retrieval. Firstly, this paper proposes a similarity calculation method based on the relationship between words in the thesaurus. On the basis of this method, combined with the idea of query expansion and weighted sorting, this paper proposes a semantic retrieval method of forestry information based on the thesaurus. Secondly, this paper uses a statistical language model to propose personalized retrieval methods based on three different user models: topic model, historical model and mixed model. Finally, a forestry information personalized semantic retrieval system is realized by using semantic retrieval and personalized retrieval method. Experimental results indicate that the proposed personalized semantic retrieval method can effectively improve the retrieval performance.","PeriodicalId":240142,"journal":{"name":"2021 IEEE/ACIS 20th International Fall Conference on Computer and Information Science (ICIS Fall)","volume":"29 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129296736","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 : 2021-10-13DOI: 10.1109/icisfall51598.2021.9627390
Yan Qi, Li Huachun, He Yifeng
High reliability is a key feature of the manned spacecraft's 1553B bus communication software, and it is the basis for the normal operation of the manned spacecraft's electronic system. To realize the high reliability of the 1553B bus communication software, five robust design methods, named “optimized initialization procedure”, “complete initialization action”, “reasonable data response modes”, “correct data access steps” and “defensive illegal data measures”, are proposed. These methods are applicable to a variety of manned spacecraft and can greatly improve the stability and reliability of the on-board electronic systems.
{"title":"High-Reliability Design Methods of the 1553B Bus Communication Software for Manned Spacecraft","authors":"Yan Qi, Li Huachun, He Yifeng","doi":"10.1109/icisfall51598.2021.9627390","DOIUrl":"https://doi.org/10.1109/icisfall51598.2021.9627390","url":null,"abstract":"High reliability is a key feature of the manned spacecraft's 1553B bus communication software, and it is the basis for the normal operation of the manned spacecraft's electronic system. To realize the high reliability of the 1553B bus communication software, five robust design methods, named “optimized initialization procedure”, “complete initialization action”, “reasonable data response modes”, “correct data access steps” and “defensive illegal data measures”, are proposed. These methods are applicable to a variety of manned spacecraft and can greatly improve the stability and reliability of the on-board electronic systems.","PeriodicalId":240142,"journal":{"name":"2021 IEEE/ACIS 20th International Fall Conference on Computer and Information Science (ICIS Fall)","volume":"44 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114576072","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 COVID-19 that emerged at the end of 2019 is the biggest public health emergency encountered by human in the past 100 years. In the face of COVID-19, people need to get correct, comprehensive and clear information. However, traditional information retrieval methods only return a collection of related web pages, and users need to distinguish the authenticity from redundant and complicated information. Therefore, such information acquisition methods are inefficient and cannot serve users well. To meet the needs of users for related information, it is necessary to study the question answering system for the COVID-19. This paper studies and builds a COVID-19 question answering system based on knowledge graph. In the System, the question answering function is realized by template matching, which based on the Naive Bayes algorithm. For the input questions, the system firstly performs entity recognition, using entity type labeling combined with entity similarity matching to identify entities in the user's questions. Then the system predicts the user's question intention and use the trained question classifier to predict the category number. Finally Cypher is utilized to query graph database to generate and output the answer. The system implemented in this paper can help users quickly obtain the information they want and improve the user's information acquisition efficiency. The system can provide people convenient and fast ways of obtaining information about COVID-19, such as medical treatment, health, materials, prevention and control, scientific research, so as to help people take precautions against diseases and decrease the incidence of COVID-19.
{"title":"The COVID-19 Question Answering System Based on Knowledge Graph","authors":"Yuze Sun, Yifei Cai, Yunkai Shen, Qian-cai Zhang, Xiaolong Feng, Mengmeng Yin, Dongmei Li","doi":"10.1109/icisfall51598.2021.9627485","DOIUrl":"https://doi.org/10.1109/icisfall51598.2021.9627485","url":null,"abstract":"The COVID-19 that emerged at the end of 2019 is the biggest public health emergency encountered by human in the past 100 years. In the face of COVID-19, people need to get correct, comprehensive and clear information. However, traditional information retrieval methods only return a collection of related web pages, and users need to distinguish the authenticity from redundant and complicated information. Therefore, such information acquisition methods are inefficient and cannot serve users well. To meet the needs of users for related information, it is necessary to study the question answering system for the COVID-19. This paper studies and builds a COVID-19 question answering system based on knowledge graph. In the System, the question answering function is realized by template matching, which based on the Naive Bayes algorithm. For the input questions, the system firstly performs entity recognition, using entity type labeling combined with entity similarity matching to identify entities in the user's questions. Then the system predicts the user's question intention and use the trained question classifier to predict the category number. Finally Cypher is utilized to query graph database to generate and output the answer. The system implemented in this paper can help users quickly obtain the information they want and improve the user's information acquisition efficiency. The system can provide people convenient and fast ways of obtaining information about COVID-19, such as medical treatment, health, materials, prevention and control, scientific research, so as to help people take precautions against diseases and decrease the incidence of COVID-19.","PeriodicalId":240142,"journal":{"name":"2021 IEEE/ACIS 20th International Fall Conference on Computer and Information Science (ICIS Fall)","volume":"18 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114435462","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 : 2021-10-13DOI: 10.1109/icisfall51598.2021.9627386
Zeyu Chen, Yana Zhang, Lianyi Zhang, Cheng Yang
Video auto-editing is a new application of artificial intelligence technology in the media industry. Camera motion is one of the critical characteristics of videos and is vital for shot arrangement. Rules based camera motion classification algorithms generalize poorly from one dataset to another. Machine learning or deep learning based algorithms have better cross-task performance. However, in order to remove the motion information in the foreground area, researchers have to use semantic segmentation neural networks that are computationally expensive. Existing foreground segmentation algorithms are only effective for samples with clear foreground areas. The salient areas generated by the saliency segmentation algorithms are not the same as the foreground areas in many cases. This paper proposes a novel deep learning based camera motion classification framework MUL-MOVE-Net, which is composed of multiple instantaneous motion networks MOVE-Net. In MOVE-Net, a lightweight RO- TextCNN module is proposed to learn multi-scale templates in the 1D angle histogram of optical flow information. Without using semantic segmentation network, the algorithm is capable of foreground fault tolerance while ensuring efficiency. For the experiments, a dataset MOVE-SET is constructed with more than 100, 000 pairs of instantaneous camera motion samples. On the testing set, our algorithm achieves an accuracy of 95.3% and a Macro-F1 value of 0.9385. In the shot-level motion classification task, the accuracy of MUL-MOVE-Net gets 4% higher than that of SGNet, and Macro-F1 0.3 higher. As a result, MUL-MOVE-Net could efficiently classify the camera motion in real time and is helpful for video auto-editing.
{"title":"RO-TextCNN Based MUL-MOVE-Net for Camera Motion Classification","authors":"Zeyu Chen, Yana Zhang, Lianyi Zhang, Cheng Yang","doi":"10.1109/icisfall51598.2021.9627386","DOIUrl":"https://doi.org/10.1109/icisfall51598.2021.9627386","url":null,"abstract":"Video auto-editing is a new application of artificial intelligence technology in the media industry. Camera motion is one of the critical characteristics of videos and is vital for shot arrangement. Rules based camera motion classification algorithms generalize poorly from one dataset to another. Machine learning or deep learning based algorithms have better cross-task performance. However, in order to remove the motion information in the foreground area, researchers have to use semantic segmentation neural networks that are computationally expensive. Existing foreground segmentation algorithms are only effective for samples with clear foreground areas. The salient areas generated by the saliency segmentation algorithms are not the same as the foreground areas in many cases. This paper proposes a novel deep learning based camera motion classification framework MUL-MOVE-Net, which is composed of multiple instantaneous motion networks MOVE-Net. In MOVE-Net, a lightweight RO- TextCNN module is proposed to learn multi-scale templates in the 1D angle histogram of optical flow information. Without using semantic segmentation network, the algorithm is capable of foreground fault tolerance while ensuring efficiency. For the experiments, a dataset MOVE-SET is constructed with more than 100, 000 pairs of instantaneous camera motion samples. On the testing set, our algorithm achieves an accuracy of 95.3% and a Macro-F1 value of 0.9385. In the shot-level motion classification task, the accuracy of MUL-MOVE-Net gets 4% higher than that of SGNet, and Macro-F1 0.3 higher. As a result, MUL-MOVE-Net could efficiently classify the camera motion in real time and is helpful for video auto-editing.","PeriodicalId":240142,"journal":{"name":"2021 IEEE/ACIS 20th International Fall Conference on Computer and Information Science (ICIS Fall)","volume":"2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126126146","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：