Pub Date : 2020-09-01DOI: 10.1109/CACRE50138.2020.9230297
Yimo Liu, Di Bu, Guokai Zhang, Ye Luo, Jianwei Lu, Weigang Wang, Binghui Zhao
Prostate cancer has been a leading cause of death among males for a long time. Currently, with the help of computer-aided detection systems, prostate cancer can be detected in a relatively early stage, thus improving the patients’ survival rate. In this paper, we propose a computer-aided system based on deep learning method to help classify prostate cancer. Our model combines both convolutional neural network (CNN) extracted features and handcrafted features. In our model, the input data is sent into two subnets. One is a modified ResNet with an improved spatial transformer (ST) for high dimension feature extraction. The other subnet extracts three handcrafted features and processes them with a simple CNN. After those two subnets, the output features of the two subnets are concatenated and then sent into the final classifier for prostate cancer classification. Experimental results show that our model achieves an accuracy of 0.947, which is better than other state-of-the-art methods.
{"title":"Using CNN With Handcrafted Features for Prostate Cancer Classification","authors":"Yimo Liu, Di Bu, Guokai Zhang, Ye Luo, Jianwei Lu, Weigang Wang, Binghui Zhao","doi":"10.1109/CACRE50138.2020.9230297","DOIUrl":"https://doi.org/10.1109/CACRE50138.2020.9230297","url":null,"abstract":"Prostate cancer has been a leading cause of death among males for a long time. Currently, with the help of computer-aided detection systems, prostate cancer can be detected in a relatively early stage, thus improving the patients’ survival rate. In this paper, we propose a computer-aided system based on deep learning method to help classify prostate cancer. Our model combines both convolutional neural network (CNN) extracted features and handcrafted features. In our model, the input data is sent into two subnets. One is a modified ResNet with an improved spatial transformer (ST) for high dimension feature extraction. The other subnet extracts three handcrafted features and processes them with a simple CNN. After those two subnets, the output features of the two subnets are concatenated and then sent into the final classifier for prostate cancer classification. Experimental results show that our model achieves an accuracy of 0.947, which is better than other state-of-the-art methods.","PeriodicalId":325195,"journal":{"name":"2020 5th International Conference on Automation, Control and Robotics Engineering (CACRE)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128449229","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 navigational positioning accuracy of a seabed mining vehicle not only directly affects the efficiency of aggregation, but also affects the reliability and stability of mining operations. In order to determine the noise reduction filtering algorithm, Which is most suitable for state estimation and position estimation models in mining vehicle sea trials, an Adaptive Kalman Filter (AKF) based on linear self-navigation position estimation is first proposed under an ideal Gaussian noise model and compared With conventional Kalman filter (KF) and Innovation Kalman filter (IKF) algorithms. Secondly, based on the underwater projection observatory, a nonlinear position estimation model based on distance and angle is proposed, introducing Gaussian noise. Particle Filter (PF) and improved particle filtering algorithms such as the Unscented Kalman Particle Filter(UPF), the Extended kalman Filter(EPF) are used for state estimation. The simulation results show that under the nonlinear position estimation model, UPF not only solves the problem of conventional Particle Filter (PF) divergence, but also significantly improves the accuracy of position estimation compared to self-navigation position estimation. The UPF algorithm based on an underwater projection observatory is best suited for navigational positioning of polymetallic nodule seabed mining vehicles.
{"title":"Filters navigation and positioning based on mining vehicle motion model","authors":"Yuheng Chen, Hongyun Wu, Zhou Liu, Yongfeng Liu, Jingwei Li, Bolin Yin","doi":"10.1109/CACRE50138.2020.9230004","DOIUrl":"https://doi.org/10.1109/CACRE50138.2020.9230004","url":null,"abstract":"The navigational positioning accuracy of a seabed mining vehicle not only directly affects the efficiency of aggregation, but also affects the reliability and stability of mining operations. In order to determine the noise reduction filtering algorithm, Which is most suitable for state estimation and position estimation models in mining vehicle sea trials, an Adaptive Kalman Filter (AKF) based on linear self-navigation position estimation is first proposed under an ideal Gaussian noise model and compared With conventional Kalman filter (KF) and Innovation Kalman filter (IKF) algorithms. Secondly, based on the underwater projection observatory, a nonlinear position estimation model based on distance and angle is proposed, introducing Gaussian noise. Particle Filter (PF) and improved particle filtering algorithms such as the Unscented Kalman Particle Filter(UPF), the Extended kalman Filter(EPF) are used for state estimation. The simulation results show that under the nonlinear position estimation model, UPF not only solves the problem of conventional Particle Filter (PF) divergence, but also significantly improves the accuracy of position estimation compared to self-navigation position estimation. The UPF algorithm based on an underwater projection observatory is best suited for navigational positioning of polymetallic nodule seabed mining vehicles.","PeriodicalId":325195,"journal":{"name":"2020 5th International Conference on Automation, Control and Robotics Engineering (CACRE)","volume":"87 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114441245","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 : 2020-09-01DOI: 10.1109/CACRE50138.2020.9230274
Dongyang Zhao, Yuqing Chen, Shuanghe Yu
With the rapid technology development in autonomous navigation of Unmanned Aerial Vehicles (UAVs) and robust object detection based on deep neural networks, the field of traffic analysis through aerial video has attracted widespread attention. In this paper, we investigate the problems of ground vehicle tracking and speed estimation using aerial view videos. At the first stage, the vehicle detection is performed through the YOLOv3 network, which is the state-of-the-art object detector. Then, a tracking-by-detection method is designed to tracking the traffic vehicles. Furthermore, in order to estimate the vehicle speed in traffic while the UAV navigating in different heights, the least square algorithm is utilized to fit the measurement data and determine the power function mapping relationship between the vehicle pixel distance and the actual distance, which further improves the accuracy of speed estimation effectively.
{"title":"Tracking and Speed Estimation of Ground Vehicles Using Aerial-view Videos","authors":"Dongyang Zhao, Yuqing Chen, Shuanghe Yu","doi":"10.1109/CACRE50138.2020.9230274","DOIUrl":"https://doi.org/10.1109/CACRE50138.2020.9230274","url":null,"abstract":"With the rapid technology development in autonomous navigation of Unmanned Aerial Vehicles (UAVs) and robust object detection based on deep neural networks, the field of traffic analysis through aerial video has attracted widespread attention. In this paper, we investigate the problems of ground vehicle tracking and speed estimation using aerial view videos. At the first stage, the vehicle detection is performed through the YOLOv3 network, which is the state-of-the-art object detector. Then, a tracking-by-detection method is designed to tracking the traffic vehicles. Furthermore, in order to estimate the vehicle speed in traffic while the UAV navigating in different heights, the least square algorithm is utilized to fit the measurement data and determine the power function mapping relationship between the vehicle pixel distance and the actual distance, which further improves the accuracy of speed estimation effectively.","PeriodicalId":325195,"journal":{"name":"2020 5th International Conference on Automation, Control and Robotics Engineering (CACRE)","volume":"64 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114867710","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 : 2020-09-01DOI: 10.1109/CACRE50138.2020.9229963
Zhongnan Tang, Yujie Wang, Qing-yang Chen, Xixiang Yang
In order to achieve the cooperative attack of multi-UAV on targets (including static targets and moving targets), the process is divided into two stages, i.e. cruise stage and strike stage; the motion model of multi-UAV and targets and the observation model of targets are established based on the two-dimensional plane simplification assumption. In the cruise phase, multi-UAV cooperative target state estimation is realized based on Unscented Kalman filter (UKF), and cooperative attack guidance law is established under multiple constraints (including time, attack angle, seeker field angle, etc.) at the strike stage. In this paper, the system simulation is carried out for stationary target and moving target respectively, and the effectiveness of the proposed algorithm and scheme is verified. The results show that the Multi-UAV bearing-only state estimation can converge rapidly, the target positioning accuracy is about 10 m, and the estimation accuracy of the target line of sight angle is about 0.1°; the multi constraint guidance law can effectively improve the cooperative combat performance of the UAV cluster, the time cooperative accuracy is about 0.3s, the attack angle cooperative accuracy is about 0.5°, and the miss distance is less than 1m.
{"title":"Research on Target State Estimation and Terminal Guidance Algorithm in the Process of Multi-UAV Cooperative Attack","authors":"Zhongnan Tang, Yujie Wang, Qing-yang Chen, Xixiang Yang","doi":"10.1109/CACRE50138.2020.9229963","DOIUrl":"https://doi.org/10.1109/CACRE50138.2020.9229963","url":null,"abstract":"In order to achieve the cooperative attack of multi-UAV on targets (including static targets and moving targets), the process is divided into two stages, i.e. cruise stage and strike stage; the motion model of multi-UAV and targets and the observation model of targets are established based on the two-dimensional plane simplification assumption. In the cruise phase, multi-UAV cooperative target state estimation is realized based on Unscented Kalman filter (UKF), and cooperative attack guidance law is established under multiple constraints (including time, attack angle, seeker field angle, etc.) at the strike stage. In this paper, the system simulation is carried out for stationary target and moving target respectively, and the effectiveness of the proposed algorithm and scheme is verified. The results show that the Multi-UAV bearing-only state estimation can converge rapidly, the target positioning accuracy is about 10 m, and the estimation accuracy of the target line of sight angle is about 0.1°; the multi constraint guidance law can effectively improve the cooperative combat performance of the UAV cluster, the time cooperative accuracy is about 0.3s, the attack angle cooperative accuracy is about 0.5°, and the miss distance is less than 1m.","PeriodicalId":325195,"journal":{"name":"2020 5th International Conference on Automation, Control and Robotics Engineering (CACRE)","volume":"44 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127058294","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 : 2020-09-01DOI: 10.1109/cacre50138.2020.9229956
Shao-ting Yu, Cai-zhi Fan
Aiming at the problem of satellite attitude tracking with uncertain moment of inertia and external interference, an adaptive control method based on RBF neural network is proposed. First, based on the error quaternion and error angular velocity, the kinematics and dynamics equations of satellite attitude tracking are derived. Then, a direct controller based on RBF neural network is designed, and the Lyapunov stability theory is used to prove that the designed controller can ensure the progressive stability of the satellite attitude tracking system. Finally, the simulation of the designed control method was verified by MATLAB/SIMULINK software. The results show that the adaptive control based on RBF neural network can effectively overcome the influence of uncertain disturbances in the system, improve the accuracy of attitude control, and has a strong Robustness.
{"title":"Adaptive Control of Satellite Attitude Tracking Based on RBF Neural Network","authors":"Shao-ting Yu, Cai-zhi Fan","doi":"10.1109/cacre50138.2020.9229956","DOIUrl":"https://doi.org/10.1109/cacre50138.2020.9229956","url":null,"abstract":"Aiming at the problem of satellite attitude tracking with uncertain moment of inertia and external interference, an adaptive control method based on RBF neural network is proposed. First, based on the error quaternion and error angular velocity, the kinematics and dynamics equations of satellite attitude tracking are derived. Then, a direct controller based on RBF neural network is designed, and the Lyapunov stability theory is used to prove that the designed controller can ensure the progressive stability of the satellite attitude tracking system. Finally, the simulation of the designed control method was verified by MATLAB/SIMULINK software. The results show that the adaptive control based on RBF neural network can effectively overcome the influence of uncertain disturbances in the system, improve the accuracy of attitude control, and has a strong Robustness.","PeriodicalId":325195,"journal":{"name":"2020 5th International Conference on Automation, Control and Robotics Engineering (CACRE)","volume":"84 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125774747","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 : 2020-09-01DOI: 10.1109/CACRE50138.2020.9230271
Chenggang Lu, Jinxiang Wang, Xu Cui
The moving target detection and tracking of mobile robots have always been a difficulty and hot spot in the field of robot research. This paper focuses on the detection and tracking algorithm of the robot moving target based on the Laser Range Finder. Firstly, the algorithm uses a Laser Range Finder to obtain the relative distance and direction information between the robot and the moving object in real-time. Then, according to the current distance and azimuth information, the instantaneous movement speed and acceleration information of the robot and the moving target, and the physical parameters of the robot itself, the movement state of the movement target is calculated and predicted, and the control function is fitted. The control strategy enables the robot to track the moving target in real-time. Experiments were conducted on the AS-R robot platform. The experimental results show that the control algorithm can effectively track the moving target in real-time, and the algorithm has good robustness.
{"title":"Moving Target Tracking with Robot Based on Laser Range Finder","authors":"Chenggang Lu, Jinxiang Wang, Xu Cui","doi":"10.1109/CACRE50138.2020.9230271","DOIUrl":"https://doi.org/10.1109/CACRE50138.2020.9230271","url":null,"abstract":"The moving target detection and tracking of mobile robots have always been a difficulty and hot spot in the field of robot research. This paper focuses on the detection and tracking algorithm of the robot moving target based on the Laser Range Finder. Firstly, the algorithm uses a Laser Range Finder to obtain the relative distance and direction information between the robot and the moving object in real-time. Then, according to the current distance and azimuth information, the instantaneous movement speed and acceleration information of the robot and the moving target, and the physical parameters of the robot itself, the movement state of the movement target is calculated and predicted, and the control function is fitted. The control strategy enables the robot to track the moving target in real-time. Experiments were conducted on the AS-R robot platform. The experimental results show that the control algorithm can effectively track the moving target in real-time, and the algorithm has good robustness.","PeriodicalId":325195,"journal":{"name":"2020 5th International Conference on Automation, Control and Robotics Engineering (CACRE)","volume":"79 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125096096","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 : 2020-09-01DOI: 10.1109/CACRE50138.2020.9230075
B. Hu, Jianbo Xie, Yusheng He, Yinsong Wang, Pingyan Ma
As the main parameter of thermal power unit, the measurement and control of main steam temperature play an important role in the safe and economic operation of power plant. In order to improve the control quality of main steam temperature control system of thermal power unit, this paper presents a performance evaluation method of main steam temperature control system of thermal power unit based on Mahalanobis distance. This method is based on the output data of leading steam temperature and main steam temperature. Firstly, the data set representing the best performance of the system is selected according to Hurst index, and the benchmark of performance evaluation is established based on this data. Then the calculation process of traditional Mahalanobis distance is improved, and the specific calculation method of performance evaluation index based on the combination of Hurst index and improved Mahalanobis distance is given. Finally, the performance evaluation index is classified by membership function. The simulation results show that the evaluation method is effective and reasonable and the calculation is simple.
{"title":"Performance Evaluation of Main Steam Temperature Control System of Thermal Power Unit Based on Mahalanobis Distance","authors":"B. Hu, Jianbo Xie, Yusheng He, Yinsong Wang, Pingyan Ma","doi":"10.1109/CACRE50138.2020.9230075","DOIUrl":"https://doi.org/10.1109/CACRE50138.2020.9230075","url":null,"abstract":"As the main parameter of thermal power unit, the measurement and control of main steam temperature play an important role in the safe and economic operation of power plant. In order to improve the control quality of main steam temperature control system of thermal power unit, this paper presents a performance evaluation method of main steam temperature control system of thermal power unit based on Mahalanobis distance. This method is based on the output data of leading steam temperature and main steam temperature. Firstly, the data set representing the best performance of the system is selected according to Hurst index, and the benchmark of performance evaluation is established based on this data. Then the calculation process of traditional Mahalanobis distance is improved, and the specific calculation method of performance evaluation index based on the combination of Hurst index and improved Mahalanobis distance is given. Finally, the performance evaluation index is classified by membership function. The simulation results show that the evaluation method is effective and reasonable and the calculation is simple.","PeriodicalId":325195,"journal":{"name":"2020 5th International Conference on Automation, Control and Robotics Engineering (CACRE)","volume":"8 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124169037","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 : 2020-09-01DOI: 10.1109/CACRE50138.2020.9230063
Peng-fei Peng, Pu Li, Yu Liu
Aiming at the problem that the naval fleet air defense communication network survivability needs to be improved urgently, a network survivability evaluation function based on Natural Connectivity and Node Degree Uniformity Coefficient is established. This function not only inherits the advantages of simple expression of Natural Connectivity and fast calculation speed, but also has the ability to evaluate the network with uneven node distribution, which can significantly improve the accuracy of the evaluation of survivability. On this basis, combined with the basic idea of particle swarm optimization, the algorithm of air defense communication network of warship formation based on particle swarm optimization is proposed. The simulation results show that the improved network survivability evaluation function and particle swarm optimization based air defense communication networking algorithm can effectively improve the network survivability of real-time mobile communication networking, and have a good application in air defense communication networking optimization of naval ship formation in future complex battlefield environment.
{"title":"Research on Air Defense Communication Network of Warship Formation Based on Particle Swarm optimization","authors":"Peng-fei Peng, Pu Li, Yu Liu","doi":"10.1109/CACRE50138.2020.9230063","DOIUrl":"https://doi.org/10.1109/CACRE50138.2020.9230063","url":null,"abstract":"Aiming at the problem that the naval fleet air defense communication network survivability needs to be improved urgently, a network survivability evaluation function based on Natural Connectivity and Node Degree Uniformity Coefficient is established. This function not only inherits the advantages of simple expression of Natural Connectivity and fast calculation speed, but also has the ability to evaluate the network with uneven node distribution, which can significantly improve the accuracy of the evaluation of survivability. On this basis, combined with the basic idea of particle swarm optimization, the algorithm of air defense communication network of warship formation based on particle swarm optimization is proposed. The simulation results show that the improved network survivability evaluation function and particle swarm optimization based air defense communication networking algorithm can effectively improve the network survivability of real-time mobile communication networking, and have a good application in air defense communication networking optimization of naval ship formation in future complex battlefield environment.","PeriodicalId":325195,"journal":{"name":"2020 5th International Conference on Automation, Control and Robotics Engineering (CACRE)","volume":"23 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121827492","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}
Laser scanning confocal microscopy (LSCM) has become a common method for biological observation and medical science. Compared with traditional optical microscope, LSCM has the advantages of high contrast and three-dimensional (3D) imaging. However, with the increase of image depth, the resolution and contrast will be reduced due to the complexity of biological tissue aberrations. Adaptive optics system is an effective method to eliminate aberration. In this paper, a wavefront sensorless adaptive optics system is used to correct aberrations generated by complex refractive index of biological tissue. In order to increase the convergence speed and reduce the influence of photobleaching, an improved stochastic parallel gradient descent algorithm with adaptive coefficient is used to control the AO system. The optical path is simulated in ZEMAX and the feasibility of the proposed algorithm is verified in MATLAB. All simulation results demonstrate that the optimal algorithm can correct the aberration effectively with the designed optical system.
{"title":"Wavefront sensorless aberration correction utilizing SPGD algorithm with adaptive coefficient for laser scanning confocal microscopy","authors":"Tianyu Zhang, Zhizheng Wu, Xiang Wei, Feng Li, Jialiang Wu, Kongbin Zhu","doi":"10.1109/CACRE50138.2020.9229951","DOIUrl":"https://doi.org/10.1109/CACRE50138.2020.9229951","url":null,"abstract":"Laser scanning confocal microscopy (LSCM) has become a common method for biological observation and medical science. Compared with traditional optical microscope, LSCM has the advantages of high contrast and three-dimensional (3D) imaging. However, with the increase of image depth, the resolution and contrast will be reduced due to the complexity of biological tissue aberrations. Adaptive optics system is an effective method to eliminate aberration. In this paper, a wavefront sensorless adaptive optics system is used to correct aberrations generated by complex refractive index of biological tissue. In order to increase the convergence speed and reduce the influence of photobleaching, an improved stochastic parallel gradient descent algorithm with adaptive coefficient is used to control the AO system. The optical path is simulated in ZEMAX and the feasibility of the proposed algorithm is verified in MATLAB. All simulation results demonstrate that the optimal algorithm can correct the aberration effectively with the designed optical system.","PeriodicalId":325195,"journal":{"name":"2020 5th International Conference on Automation, Control and Robotics Engineering (CACRE)","volume":"13 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122167515","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 : 2020-09-01DOI: 10.1109/CACRE50138.2020.9230294
Hezhi Cao, Yanxin Ma, Ronghui Zhan, Chao Ma, Jun Zhang
Traditional Convolutional Neural Networks (CNN) are limited to extract informative local features of point clouds due to the fixed geometric structures in convolution kernel against irregular and unstructured point clouds. It usually requires data transformation such as voxelization or projection, inducing a possible loss of information. Instead of fitting the input points to the kernel by regularization, we choose to fit the kernel to input points to conduct convolution. In this paper, we present a new method to define and compute convolution directly on 3D point clouds by Adaptive Surface Fitting Convolution (ASFConv). The key idea is to utilize a set of kernel points distributed on the tangent plane and project them back to point cloud surface. After adapting to the distribution of input points, ASFConv kernel can better capture local neighborhood geometry and benefit the feature extraction. In the experiments, we evaluate our network on two public datasets: ModelNet40 and ShapeNet for classification and segmentation. The experimental results show that our method obtain competitive performances compared to the state-of-the-art.
{"title":"Adaptive Surface Fitting Convolution for 3D Point Cloud Analysis","authors":"Hezhi Cao, Yanxin Ma, Ronghui Zhan, Chao Ma, Jun Zhang","doi":"10.1109/CACRE50138.2020.9230294","DOIUrl":"https://doi.org/10.1109/CACRE50138.2020.9230294","url":null,"abstract":"Traditional Convolutional Neural Networks (CNN) are limited to extract informative local features of point clouds due to the fixed geometric structures in convolution kernel against irregular and unstructured point clouds. It usually requires data transformation such as voxelization or projection, inducing a possible loss of information. Instead of fitting the input points to the kernel by regularization, we choose to fit the kernel to input points to conduct convolution. In this paper, we present a new method to define and compute convolution directly on 3D point clouds by Adaptive Surface Fitting Convolution (ASFConv). The key idea is to utilize a set of kernel points distributed on the tangent plane and project them back to point cloud surface. After adapting to the distribution of input points, ASFConv kernel can better capture local neighborhood geometry and benefit the feature extraction. In the experiments, we evaluate our network on two public datasets: ModelNet40 and ShapeNet for classification and segmentation. The experimental results show that our method obtain competitive performances compared to the state-of-the-art.","PeriodicalId":325195,"journal":{"name":"2020 5th International Conference on Automation, Control and Robotics Engineering (CACRE)","volume":"21 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114629126","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：