Simultaneous localization and mapping(SLAM)has been applied across a wide range of areas from robotics to automatic pilot.Most of the SLAM algorithms are based on the assumption that the noise is timeinvariant Gaussia...Simultaneous localization and mapping(SLAM)has been applied across a wide range of areas from robotics to automatic pilot.Most of the SLAM algorithms are based on the assumption that the noise is timeinvariant Gaussian distribution.In some cases,this assumption no longer holds and the performance of the traditional SLAM algorithms declines.In this paper,we present a robust SLAM algorithm based on variational Bayes method by modelling the observation noise as inverse-Wishart distribution with "harmonic mean".Besides,cubature integration is utilized to solve the problem of nonlinear system.The proposed algorithm can effectively solve the problem of filtering divergence for traditional filtering algorithm when suffering the time-variant observation noise,especially for heavy-tai led noise.To validate the algorithm,we compare it with other t raditional filtering algorithms.The results show the effectiveness of the algorithm.展开更多
Tropical cyclone (TC) center locating is crucial because it lays the foundation for TC forecasting. Locating TC centers, usually by manual means, continues to present many difficulties. Not least is the problem of inc...Tropical cyclone (TC) center locating is crucial because it lays the foundation for TC forecasting. Locating TC centers, usually by manual means, continues to present many difficulties. Not least is the problem of inconsistency between TC center locations forecast by different agencies. In this paper, an objective TC center locating scheme is developed, using infrared satellite images. We introduce a pattern-matching concept, which we illustrate using a spiral curve model. A spiral band model, based on a spiral band region, is designed to extract the spiral cloud-rain bands (SCRBs) of TCs. We propose corresponding criteria on which to score the fitting value of a candidate template defined by our models. In the proposed scheme, TC location is an optimization problem, solved by an ant colony optimization algorithm. In numerical experiments, a minimal mean distance error of 17.9 km is obtained when the scheme is tested against best-track data. The scheme is suitable for TCs with distinct SCRBs or symmetrical central dense overcast, and for TCs both with and without clear eyes.展开更多
Domain adaptation and adversarial networks are two main approaches for transfer learning.Domain adaptation methods match the mean values of source and target domains,which requires a very large batch size during train...Domain adaptation and adversarial networks are two main approaches for transfer learning.Domain adaptation methods match the mean values of source and target domains,which requires a very large batch size during training.However,adversarial networks are usually unstable when training.In this paper,we propose a joint method of feature matching and adversarial networks to reduce domain discrepancy and mine domaininvariant features from the local and global aspects.At the same time,our method improves the stability of training.Moreover,the method is embedded into a unified convolutional neural network that can be easily optimized by gradient descent.Experimental results show that our joint method can yield the state-of-the-art results on three common public datasets.展开更多
基金the National Natural Science Foundation of China(No.61803260)。
文摘Simultaneous localization and mapping(SLAM)has been applied across a wide range of areas from robotics to automatic pilot.Most of the SLAM algorithms are based on the assumption that the noise is timeinvariant Gaussian distribution.In some cases,this assumption no longer holds and the performance of the traditional SLAM algorithms declines.In this paper,we present a robust SLAM algorithm based on variational Bayes method by modelling the observation noise as inverse-Wishart distribution with "harmonic mean".Besides,cubature integration is utilized to solve the problem of nonlinear system.The proposed algorithm can effectively solve the problem of filtering divergence for traditional filtering algorithm when suffering the time-variant observation noise,especially for heavy-tai led noise.To validate the algorithm,we compare it with other t raditional filtering algorithms.The results show the effectiveness of the algorithm.
基金supported by National Natural Science Foundation of China (Grant Nos. 60775022 and 60805005)Shanghai Municipal Natural Science Foundation (Grant Nos.09ZR1413700 and No.08ZR1410700)Ph.D. Programs Foundation of Ministry of Education of China (Grant No. 200802481119)
文摘Tropical cyclone (TC) center locating is crucial because it lays the foundation for TC forecasting. Locating TC centers, usually by manual means, continues to present many difficulties. Not least is the problem of inconsistency between TC center locations forecast by different agencies. In this paper, an objective TC center locating scheme is developed, using infrared satellite images. We introduce a pattern-matching concept, which we illustrate using a spiral curve model. A spiral band model, based on a spiral band region, is designed to extract the spiral cloud-rain bands (SCRBs) of TCs. We propose corresponding criteria on which to score the fitting value of a candidate template defined by our models. In the proposed scheme, TC location is an optimization problem, solved by an ant colony optimization algorithm. In numerical experiments, a minimal mean distance error of 17.9 km is obtained when the scheme is tested against best-track data. The scheme is suitable for TCs with distinct SCRBs or symmetrical central dense overcast, and for TCs both with and without clear eyes.
基金the Aerospace Science and Technology Foundation(No.20115557007)the National Natural Science Foundation of China(No.61673262)the Military Science and Technology Foundation of China(No.18-H863-03-ZT-001-006-06)
文摘Domain adaptation and adversarial networks are two main approaches for transfer learning.Domain adaptation methods match the mean values of source and target domains,which requires a very large batch size during training.However,adversarial networks are usually unstable when training.In this paper,we propose a joint method of feature matching and adversarial networks to reduce domain discrepancy and mine domaininvariant features from the local and global aspects.At the same time,our method improves the stability of training.Moreover,the method is embedded into a unified convolutional neural network that can be easily optimized by gradient descent.Experimental results show that our joint method can yield the state-of-the-art results on three common public datasets.
