Automated driving has been developed rapidly in the past five years. Many automakers have popularized SAE Level 2 on self-driving cars, some even on regular vehicles. To implement Level 3 or Level 4, it’s necessary to rely on a good localization system when GPS signal is not available. For this need, this paper proposes two effective methods to increase the accuracy of the real-time SLAM: The first method: When the original point cloud is input, the point cloud is divided into the short-range group, the medium-range group and the long-range group. An adaptive parameter adjustment method is then used to obtain the optimal parameters for each of these point cloud groups. However, Lidar’s physical characteristics can cause the point cloud to be insufficient, making an important part of points misjudged as outliers for the medium-range and long-range cases. Thanks to the help of the adaptive parameters, these point clouds, which were originally misjudged as outliers can be preserved in this paper. The second method: In point clouds, the same object in different ranges is represented with different point clouds. Hence, features, such as the roughness and density, can dramatically change with the variation of distance even when it is the same object. To solve this problem, we have designed three different range point cloud feature extraction methods to get more accurate point cloud features, such as the planes or edges. By combining these two steps, the LeGO-LOAM accuracy can be effectively increased by more than 30% while achieving the performance of the real-time SLAM, which is more accurate and faster than the NDT-Mapping used in Autoware.
{"title":"Fast Point Cloud Feature Extraction for Real-time SLAM","authors":"Sheng-Wei Lee, Chih-Ming Hsu, Ming-Che Lee, Yuan-Ting Fu, Fetullah Atas, A. Tsai","doi":"10.1109/CACS47674.2019.9024355","DOIUrl":"https://doi.org/10.1109/CACS47674.2019.9024355","url":null,"abstract":"Automated driving has been developed rapidly in the past five years. Many automakers have popularized SAE Level 2 on self-driving cars, some even on regular vehicles. To implement Level 3 or Level 4, it’s necessary to rely on a good localization system when GPS signal is not available. For this need, this paper proposes two effective methods to increase the accuracy of the real-time SLAM: The first method: When the original point cloud is input, the point cloud is divided into the short-range group, the medium-range group and the long-range group. An adaptive parameter adjustment method is then used to obtain the optimal parameters for each of these point cloud groups. However, Lidar’s physical characteristics can cause the point cloud to be insufficient, making an important part of points misjudged as outliers for the medium-range and long-range cases. Thanks to the help of the adaptive parameters, these point clouds, which were originally misjudged as outliers can be preserved in this paper. The second method: In point clouds, the same object in different ranges is represented with different point clouds. Hence, features, such as the roughness and density, can dramatically change with the variation of distance even when it is the same object. To solve this problem, we have designed three different range point cloud feature extraction methods to get more accurate point cloud features, such as the planes or edges. By combining these two steps, the LeGO-LOAM accuracy can be effectively increased by more than 30% while achieving the performance of the real-time SLAM, which is more accurate and faster than the NDT-Mapping used in Autoware.","PeriodicalId":247039,"journal":{"name":"2019 International Automatic Control Conference (CACS)","volume":"18 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125204917","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 : 2019-11-01DOI: 10.1109/CACS47674.2019.9024736
Bo-Jie Hsu, Bor-Sen Chen
In this study, in terms of systems biology approaches and deep learning method, we proposed a series of strategies for systems medicine design toward TNBC and non-TNBC. For systems biology approach, we constructed candidate genome-wide genetic and epigenetic network (GWGEN) by big data mining technique and identified real GWGEN of TNBC and non-TNBC by corresponding microarray data via system identification and model order selection methods. Core GWGEN of TNBC and non-TNBC were constructed from their corresponding GWGENs and then denoted by KEGG pathways to obtain core signaling pathways of TNBC and non-TNBC, which were compared to find essential carcinogenic biomarkers to bring about multiple cellular dysfunctions including cell proliferation, autophagy, immune response, cell differentiation, apoptosis, metastasis, angiogenesis, and epithelial-mesenchymal transition (EMT). With the help of the drug-target interaction (DTI) model based on deep neural network trained through feature vectors of drug-target databases, we could select candidate drugs for these drug targets. These candidate drugs were still filtered for the toxicity by LD50 and for regulation ability by connectively Map (CMap) as potential drugs, and then these potential drugs are combined as potential multiple-molecule drugs, i.e., resveratrol, sirolimus, prednisolone for TNBC and resveratrol, sirolimus, carbamazepine, verapamil for non-TNBC.
{"title":"Systems Drug Discovery and Design for Triple-Negative Breast Cancer and Non-Triple-Negative Breast Cancer Based on Systems Carcinogenic Mechanism and Deep Learning Method","authors":"Bo-Jie Hsu, Bor-Sen Chen","doi":"10.1109/CACS47674.2019.9024736","DOIUrl":"https://doi.org/10.1109/CACS47674.2019.9024736","url":null,"abstract":"In this study, in terms of systems biology approaches and deep learning method, we proposed a series of strategies for systems medicine design toward TNBC and non-TNBC. For systems biology approach, we constructed candidate genome-wide genetic and epigenetic network (GWGEN) by big data mining technique and identified real GWGEN of TNBC and non-TNBC by corresponding microarray data via system identification and model order selection methods. Core GWGEN of TNBC and non-TNBC were constructed from their corresponding GWGENs and then denoted by KEGG pathways to obtain core signaling pathways of TNBC and non-TNBC, which were compared to find essential carcinogenic biomarkers to bring about multiple cellular dysfunctions including cell proliferation, autophagy, immune response, cell differentiation, apoptosis, metastasis, angiogenesis, and epithelial-mesenchymal transition (EMT). With the help of the drug-target interaction (DTI) model based on deep neural network trained through feature vectors of drug-target databases, we could select candidate drugs for these drug targets. These candidate drugs were still filtered for the toxicity by LD50 and for regulation ability by connectively Map (CMap) as potential drugs, and then these potential drugs are combined as potential multiple-molecule drugs, i.e., resveratrol, sirolimus, prednisolone for TNBC and resveratrol, sirolimus, carbamazepine, verapamil for non-TNBC.","PeriodicalId":247039,"journal":{"name":"2019 International Automatic Control Conference (CACS)","volume":"23 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116244052","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 : 2019-11-01DOI: 10.1109/CACS47674.2019.9024740
Jiravit Pratvittaya, S. Wongsa
This work is motivated by the modelling of sticky control valves via the piecewise linear Hammerstein (PWL-HMM) identification. In PWL-HMM, the nonlinear part of the valve, i.e. stiction, is described by a point-slope-based hysteresis model where a set of knots have to be defined for both the ascent and descent paths of the valve position signal. Traditionally, the knots are assumed to be uniformly distributed, as a result some irrelevant knots might be included in the model. We tackle this problem by proposing two methods, namely the constant threshold-based and BIC-based knot reductions, to identify such irrelevant knots and refine the PWL model of stiction. Numerical, experimental and industrial examples are provided to illustrate the effectiveness of the proposed knot refinement methods.
{"title":"Identification of Valve Stiction in Control Loops Using Refined Piecewise-Linear Hammerstein Models","authors":"Jiravit Pratvittaya, S. Wongsa","doi":"10.1109/CACS47674.2019.9024740","DOIUrl":"https://doi.org/10.1109/CACS47674.2019.9024740","url":null,"abstract":"This work is motivated by the modelling of sticky control valves via the piecewise linear Hammerstein (PWL-HMM) identification. In PWL-HMM, the nonlinear part of the valve, i.e. stiction, is described by a point-slope-based hysteresis model where a set of knots have to be defined for both the ascent and descent paths of the valve position signal. Traditionally, the knots are assumed to be uniformly distributed, as a result some irrelevant knots might be included in the model. We tackle this problem by proposing two methods, namely the constant threshold-based and BIC-based knot reductions, to identify such irrelevant knots and refine the PWL model of stiction. Numerical, experimental and industrial examples are provided to illustrate the effectiveness of the proposed knot refinement methods.","PeriodicalId":247039,"journal":{"name":"2019 International Automatic Control Conference (CACS)","volume":"39 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130132434","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 : 2019-11-01DOI: 10.1109/CACS47674.2019.9024741
Hong-Hsu Liou, Ming-Tzu Ho
The aim of this paper is to design and implement a single-legged vertical hopping robot with a pneumatic cylinder. The energy analysis is used to determine the hopping cycle and height for the vertical hopping robot. The mathematical model of the system is derived. Second order sliding mode control is used to design control laws to control the mass flow rate for the upper chamber and lower chamber of the cylinder. The numerical simulations are conducted. In experiments, proportional valves are used for control actuation. For feedback control, the strokes of the cylinder rod and the hopping height of the cylinder are measured by a potentiometer and a magnetic encoder, respectively. The control schemes are implemented on a digital signal processor. From simulation and experimental results, it is shown that the designed sliding mode controllers can achieve effective hopping control.
{"title":"Hopping Control of a Pneumatic Single-Legged Robot using Sliding Mode Control","authors":"Hong-Hsu Liou, Ming-Tzu Ho","doi":"10.1109/CACS47674.2019.9024741","DOIUrl":"https://doi.org/10.1109/CACS47674.2019.9024741","url":null,"abstract":"The aim of this paper is to design and implement a single-legged vertical hopping robot with a pneumatic cylinder. The energy analysis is used to determine the hopping cycle and height for the vertical hopping robot. The mathematical model of the system is derived. Second order sliding mode control is used to design control laws to control the mass flow rate for the upper chamber and lower chamber of the cylinder. The numerical simulations are conducted. In experiments, proportional valves are used for control actuation. For feedback control, the strokes of the cylinder rod and the hopping height of the cylinder are measured by a potentiometer and a magnetic encoder, respectively. The control schemes are implemented on a digital signal processor. From simulation and experimental results, it is shown that the designed sliding mode controllers can achieve effective hopping control.","PeriodicalId":247039,"journal":{"name":"2019 International Automatic Control Conference (CACS)","volume":"112 2","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"120920643","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 : 2019-11-01DOI: 10.1109/CACS47674.2019.9024365
Ejaz Ahmad, Y. Song, Muhammad Arshad Khan, I. Youn
Ride comfort and the road holding capability of the vehicles are two major conflicting problems in-vehicle system dynamics and control. In this work, the effectiveness of an aerodynamic control system is investigated to overcome the tradeoff between ride comfort and road holding to enhance vehicle handling capability. A preview optimal control strategy comprised of feedback and a feedforward control strategy is used to track the ideal roll motion during cornering. The feedback control deals with tracking error while the purpose of the feedforward control is to deal with the future circular or a lane change maneuvers. The proposed control strategy helps to eliminate the effect of the lateral forces acting on the vehicle body in order to track the ideal roll motion. The simulation results show that the proposed control strategy accurately track the desired roll angle and enhance the ride comfort and vehicle handling capability.
{"title":"Attitude Motion Control of a Half car Model with Tracking Controller Using Aerodynamic Surfaces","authors":"Ejaz Ahmad, Y. Song, Muhammad Arshad Khan, I. Youn","doi":"10.1109/CACS47674.2019.9024365","DOIUrl":"https://doi.org/10.1109/CACS47674.2019.9024365","url":null,"abstract":"Ride comfort and the road holding capability of the vehicles are two major conflicting problems in-vehicle system dynamics and control. In this work, the effectiveness of an aerodynamic control system is investigated to overcome the tradeoff between ride comfort and road holding to enhance vehicle handling capability. A preview optimal control strategy comprised of feedback and a feedforward control strategy is used to track the ideal roll motion during cornering. The feedback control deals with tracking error while the purpose of the feedforward control is to deal with the future circular or a lane change maneuvers. The proposed control strategy helps to eliminate the effect of the lateral forces acting on the vehicle body in order to track the ideal roll motion. The simulation results show that the proposed control strategy accurately track the desired roll angle and enhance the ride comfort and vehicle handling capability.","PeriodicalId":247039,"journal":{"name":"2019 International Automatic Control Conference (CACS)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115211531","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 : 2019-11-01DOI: 10.1109/CACS47674.2019.9024734
K. Song, Pei-Chun Lu, Shao-Huan Song
In this paper we propose a novel human-robot interaction system to find a specific person in public for providing service. The system combines indoor localization, face recognition and robot navigation. The indoor localization uses deep neural network (DNN) and particle filtering to estimate the user position. A face recognition module provides the user identification to the robot. The robot first uses localization data to navigate to the vicinity of the user and then uses the face recognition to move to the front of the user to provide service. To verify the effectiveness of the design, we implemented the system to a mobile robot and integrated the application through a smart phone. The integrated experiments demonstrated that a user can call the robot to come to his/her front by using the proposed design. One also can order the robot via a smart phone to find a specific person and interact with him/her.
{"title":"Human-Robot Interaction Design Based on Specific Person Finding and Localization of a Mobile Robot","authors":"K. Song, Pei-Chun Lu, Shao-Huan Song","doi":"10.1109/CACS47674.2019.9024734","DOIUrl":"https://doi.org/10.1109/CACS47674.2019.9024734","url":null,"abstract":"In this paper we propose a novel human-robot interaction system to find a specific person in public for providing service. The system combines indoor localization, face recognition and robot navigation. The indoor localization uses deep neural network (DNN) and particle filtering to estimate the user position. A face recognition module provides the user identification to the robot. The robot first uses localization data to navigate to the vicinity of the user and then uses the face recognition to move to the front of the user to provide service. To verify the effectiveness of the design, we implemented the system to a mobile robot and integrated the application through a smart phone. The integrated experiments demonstrated that a user can call the robot to come to his/her front by using the proposed design. One also can order the robot via a smart phone to find a specific person and interact with him/her.","PeriodicalId":247039,"journal":{"name":"2019 International Automatic Control Conference (CACS)","volume":"73 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114217061","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 : 2019-11-01DOI: 10.1109/CACS47674.2019.9024731
Guo-Sheng Cai, H. Lin, Shih-Fen Kao
In this work we present the localization and navigation for a mobile robot in the outdoor environment. It is based on fusing the data from IMU, differential GPS and visual odometry using the extended Kalman filter framework. First, the IMU provides the heading angle information from the magnetometer and angular velocity, and GPS provides the absolute position information of the mobile robot. The image-based visual odometry is adopted to derive the moving distance and additional localization information. Finally, the mobile robot position is further refined using the extended Kalman filter. The experiments are carried out in the outdoor environment. We compare the results with the original GPS raw data, and the performance of the presented method is evaluated.
{"title":"Mobile Robot Localization using GPS, IMU and Visual Odometry","authors":"Guo-Sheng Cai, H. Lin, Shih-Fen Kao","doi":"10.1109/CACS47674.2019.9024731","DOIUrl":"https://doi.org/10.1109/CACS47674.2019.9024731","url":null,"abstract":"In this work we present the localization and navigation for a mobile robot in the outdoor environment. It is based on fusing the data from IMU, differential GPS and visual odometry using the extended Kalman filter framework. First, the IMU provides the heading angle information from the magnetometer and angular velocity, and GPS provides the absolute position information of the mobile robot. The image-based visual odometry is adopted to derive the moving distance and additional localization information. Finally, the mobile robot position is further refined using the extended Kalman filter. The experiments are carried out in the outdoor environment. We compare the results with the original GPS raw data, and the performance of the presented method is evaluated.","PeriodicalId":247039,"journal":{"name":"2019 International Automatic Control Conference (CACS)","volume":"12 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129013486","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 : 2019-11-01DOI: 10.1109/CACS47674.2019.9024360
Yu-Cheng Chu, Horng-Horng Lin
We propose a new object packing approach, RePack, to arrange a series of identical image objects to a rectangular canvas densely by a deep CNN with reinforcement learning. In our approach, adding a new object to an image pack of existing objects is modeled as classification of possible pack configurations by a CNN. To iteratively reinforce the CNN, pack trees are built to identify object overlaps and to find denser pack configurations for reinforcement training. Such a reinforcement learning process for enhancing a CNN for dense object packing is rarely seen in previous literature. Preliminary experimental results show that the reinforced deep CNN can generate dense object packs in a sequential manner for circular, triangular and quadrilateral objects.
{"title":"RePack: Dense Object Packing Using Deep CNN with Reinforcement Learning","authors":"Yu-Cheng Chu, Horng-Horng Lin","doi":"10.1109/CACS47674.2019.9024360","DOIUrl":"https://doi.org/10.1109/CACS47674.2019.9024360","url":null,"abstract":"We propose a new object packing approach, RePack, to arrange a series of identical image objects to a rectangular canvas densely by a deep CNN with reinforcement learning. In our approach, adding a new object to an image pack of existing objects is modeled as classification of possible pack configurations by a CNN. To iteratively reinforce the CNN, pack trees are built to identify object overlaps and to find denser pack configurations for reinforcement training. Such a reinforcement learning process for enhancing a CNN for dense object packing is rarely seen in previous literature. Preliminary experimental results show that the reinforced deep CNN can generate dense object packs in a sequential manner for circular, triangular and quadrilateral objects.","PeriodicalId":247039,"journal":{"name":"2019 International Automatic Control Conference (CACS)","volume":"53 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124886588","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 : 2019-11-01DOI: 10.1109/CACS47674.2019.9024358
S. Huang, I-Hsuan Su, Wei-Chi Lee, Tz-Heng Lin
The aim of this paper is to improve logistics service quality of cross border e-commerce business. Cross border e-commerce is one of the most important industries for capturing fast growing digital consuming market. Famous Asian e-commerce giants such as Alibaba from China and Softbank from Japan also adopted aggressive market expansion strategies, and these global competitors deeply push lots of pressure on Taiwan’s e-commerce operators. To stay competitive in this market, excellent logistics service quality is indispensable for supporting the e-commerce business development for satisfying online shoppers. Therefore, this paper applies multilayer QFD method to evaluate logistics service quality of cross border e-commerce operator. The theoretical contribution of this project is to modify classic QFD framework by using sophisticated multilayer analysis method. Multilayer QFD framework may improve this shortcoming by evaluating the requirements from both service provider side (logistics service provider) and customer side (online shopper). The logistics service quality solutions based on the technical measures may be more precise under multilayer QFD framework. The result of this paper will provide insights and implication for e-commerce operator, logistics provider and platform operators with solutions to improve service quality.
{"title":"Logistics service quality evaluation of cross border e-commerce operators: a multilayer framework analysis in digital shopping market","authors":"S. Huang, I-Hsuan Su, Wei-Chi Lee, Tz-Heng Lin","doi":"10.1109/CACS47674.2019.9024358","DOIUrl":"https://doi.org/10.1109/CACS47674.2019.9024358","url":null,"abstract":"The aim of this paper is to improve logistics service quality of cross border e-commerce business. Cross border e-commerce is one of the most important industries for capturing fast growing digital consuming market. Famous Asian e-commerce giants such as Alibaba from China and Softbank from Japan also adopted aggressive market expansion strategies, and these global competitors deeply push lots of pressure on Taiwan’s e-commerce operators. To stay competitive in this market, excellent logistics service quality is indispensable for supporting the e-commerce business development for satisfying online shoppers. Therefore, this paper applies multilayer QFD method to evaluate logistics service quality of cross border e-commerce operator. The theoretical contribution of this project is to modify classic QFD framework by using sophisticated multilayer analysis method. Multilayer QFD framework may improve this shortcoming by evaluating the requirements from both service provider side (logistics service provider) and customer side (online shopper). The logistics service quality solutions based on the technical measures may be more precise under multilayer QFD framework. The result of this paper will provide insights and implication for e-commerce operator, logistics provider and platform operators with solutions to improve service quality.","PeriodicalId":247039,"journal":{"name":"2019 International Automatic Control Conference (CACS)","volume":"31 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131126607","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}
Face detection and recognition is an important issue and a difficult task in computer vision and human-computer interaction. Recently, with the development of deep learning, several related technologies have been proposed for face detection and facial expression recognition (FER), and the outstanding convolutional neural networks are the most common used in this field. This thesis applies the multi-task cascade convolutional neural network to face detection, and then designs the real-time FER system based on densely connected convolution network (DenseNet). The system first scales the input image to an image pyramid, and then uses the hierarchical network to determine whether a candidate window includes a human face. If a face exists, then send the candidate window to the FER system. Since DenseNet possesses the property of feature reuse, it can effectively reduce the amount of parameters and computation efforts, beneficial to develop the real-time system. In order to capture the variation of facial muscle in different expressions, this architecture adopts convolution operations with a stride 1 and tries different numbers of dense blocks. Through experiments, the proposed system can achieve real-time recognition in 30FPS and with recognition accuracy better than human eyes.
{"title":"Multi-task Cascaded and Densely Connected Convolutional Networks Applied to Human Face Detection and Facial Expression Recognition System","authors":"Kuan-Yu Chou, Yi-Wen Cheng, Wei-Ren Chen, Yon-Ping Chen","doi":"10.1109/CACS47674.2019.9024357","DOIUrl":"https://doi.org/10.1109/CACS47674.2019.9024357","url":null,"abstract":"Face detection and recognition is an important issue and a difficult task in computer vision and human-computer interaction. Recently, with the development of deep learning, several related technologies have been proposed for face detection and facial expression recognition (FER), and the outstanding convolutional neural networks are the most common used in this field. This thesis applies the multi-task cascade convolutional neural network to face detection, and then designs the real-time FER system based on densely connected convolution network (DenseNet). The system first scales the input image to an image pyramid, and then uses the hierarchical network to determine whether a candidate window includes a human face. If a face exists, then send the candidate window to the FER system. Since DenseNet possesses the property of feature reuse, it can effectively reduce the amount of parameters and computation efforts, beneficial to develop the real-time system. In order to capture the variation of facial muscle in different expressions, this architecture adopts convolution operations with a stride 1 and tries different numbers of dense blocks. Through experiments, the proposed system can achieve real-time recognition in 30FPS and with recognition accuracy better than human eyes.","PeriodicalId":247039,"journal":{"name":"2019 International Automatic Control Conference (CACS)","volume":"33 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128148118","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：