为解决在训练物体六自由度位姿估计神经网络时,人工标注真实场景数据集困难的问题,提出一种自动生成大量单目六自由度位姿估计数据集的方法,可提高数据集标注效率和精度。考虑采集图象环境的光照、物体遮挡等条件,以单目RGB相机、物体...为解决在训练物体六自由度位姿估计神经网络时,人工标注真实场景数据集困难的问题,提出一种自动生成大量单目六自由度位姿估计数据集的方法,可提高数据集标注效率和精度。考虑采集图象环境的光照、物体遮挡等条件,以单目RGB相机、物体三维模型作为输入,在运动恢复结构(structure form motion,SfM)算法框架中添加尺度先验信息约束,实现在真实场景快速生成大量用于六自由度位姿估计训练的数据集。以生活用品为例,分别制作无遮挡、有遮挡数据集,与现有六自由度位姿估计数据集作对比,使用神经网络算法验证根据该方法制作出数据集的可行性与有效性。展开更多
With the advancement of technology and the increase in user demands, gesture recognition played a pivotal role in the field of human-computer interaction. Among various sensing devices, Time-of-Flight (ToF) sensors we...With the advancement of technology and the increase in user demands, gesture recognition played a pivotal role in the field of human-computer interaction. Among various sensing devices, Time-of-Flight (ToF) sensors were widely applied due to their low cost. This paper explored the implementation of a human hand posture recognition system using ToF sensors and residual neural networks. Firstly, this paper reviewed the typical applications of human hand recognition. Secondly, this paper designed a hand gesture recognition system using a ToF sensor VL53L5. Subsequently, data preprocessing was conducted, followed by training the constructed residual neural network. Then, the recognition results were analyzed, indicating that gesture recognition based on the residual neural network achieved an accuracy of 98.5% in a 5-class classification scenario. Finally, the paper discussed existing issues and future research directions.展开更多
Two novel spline adaptive filtering(SAF)algorithms are proposed by combining different iterative gradient methods,i.e.,Adagrad and RMSProp,named SAF-Adagrad and SAF-RMSProp,in this paper.Detailed convergence performan...Two novel spline adaptive filtering(SAF)algorithms are proposed by combining different iterative gradient methods,i.e.,Adagrad and RMSProp,named SAF-Adagrad and SAF-RMSProp,in this paper.Detailed convergence performance and computational complexity analyses are carried out also.Furthermore,compared with existing SAF algorithms,the influence of step-size and noise types on SAF algorithms are explored for nonlinear system identification under artificial datasets.Numerical results show that the SAF-Adagrad and SAFRMSProp algorithms have better convergence performance than some existing SAF algorithms(i.e.,SAF-SGD,SAF-ARC-MMSGD,and SAF-LHC-MNAG).The analysis results of various measured real datasets also verify this conclusion.Overall,the effectiveness of SAF-Adagrad and SAF-RMSProp are confirmed for the accurate identification of nonlinear systems.展开更多
A real-time,long-term surface meteorological blended forcing dataset(SMBFD)has been developed based on station observations,satellite retrievals,and reanalysis products in China.The observations are collected at natio...A real-time,long-term surface meteorological blended forcing dataset(SMBFD)has been developed based on station observations,satellite retrievals,and reanalysis products in China.The observations are collected at national and regional automatic weather stations,satellite data are obtained from the Fengyun(FY)series satellites retrievals,and the reanalysis products are obtained from the ECMWF.The 90-m resolution digital terrain elevation data in China are obtained from the Shuttle Radar Topographic Mission(SRTM)for temperature and humidity elevation adjustment.The dataset includes 2-m air temperature and humidity,10-m zonal and meridional winds,downward shortwave radiation,surface pressure,and precipitation.The spatial resolution is 1 km,and the temporal resolution is 1 h.During the data processing procedure,various data fusion techniques including the space–time multiscale variational analysis,the discrete ordinates radiative transfer(DISORT)model,the hybrid radiation estimation model,and a terrain correction algorithm are employed.Dependent and independent evaluations of the dataset are performed against observations.The SMBFD dataset is also compared with similar datasets produced in other major meteorological operational centers in the world.The results are as follows.(1)All variables show reasonable geographic distribution features and realistic spatial and temporal variations.(2)Dependent and independent evaluations both indicate that the gridded SMBFD dataset is close to the observations,while the dependent evaluation yields better results than the independent evaluation.(3)Compared with similar datasets produced in other meteorological operational centers,the real-time and retrospective surface meteorological fusion data obviously have higher quality.The dataset introduced in the present study is in general stable and accurate,and can be applied in various practice such as meteorology,agriculture,ecology,environmental protection,etc.Meanwhile,this dataset has been used as the atmospheric forcing data to drive the operational High-resolution Land Data Assimilation System of China Meteorological Administration.The dataset with the network Common Data Form(NETCDF)can be decoded by various programming languages,and it is freely available to non-commercial users.展开更多
文摘为解决在训练物体六自由度位姿估计神经网络时,人工标注真实场景数据集困难的问题,提出一种自动生成大量单目六自由度位姿估计数据集的方法,可提高数据集标注效率和精度。考虑采集图象环境的光照、物体遮挡等条件,以单目RGB相机、物体三维模型作为输入,在运动恢复结构(structure form motion,SfM)算法框架中添加尺度先验信息约束,实现在真实场景快速生成大量用于六自由度位姿估计训练的数据集。以生活用品为例,分别制作无遮挡、有遮挡数据集,与现有六自由度位姿估计数据集作对比,使用神经网络算法验证根据该方法制作出数据集的可行性与有效性。
文摘With the advancement of technology and the increase in user demands, gesture recognition played a pivotal role in the field of human-computer interaction. Among various sensing devices, Time-of-Flight (ToF) sensors were widely applied due to their low cost. This paper explored the implementation of a human hand posture recognition system using ToF sensors and residual neural networks. Firstly, this paper reviewed the typical applications of human hand recognition. Secondly, this paper designed a hand gesture recognition system using a ToF sensor VL53L5. Subsequently, data preprocessing was conducted, followed by training the constructed residual neural network. Then, the recognition results were analyzed, indicating that gesture recognition based on the residual neural network achieved an accuracy of 98.5% in a 5-class classification scenario. Finally, the paper discussed existing issues and future research directions.
基金supported by the National Natural Science Foundation of China(61871420)the Natural Science Foundation of Sichuan Province,China(23NSFSC2916)the introduction of talent,Southwest MinZu University,China,funding research projects start(RQD2021064).
文摘Two novel spline adaptive filtering(SAF)algorithms are proposed by combining different iterative gradient methods,i.e.,Adagrad and RMSProp,named SAF-Adagrad and SAF-RMSProp,in this paper.Detailed convergence performance and computational complexity analyses are carried out also.Furthermore,compared with existing SAF algorithms,the influence of step-size and noise types on SAF algorithms are explored for nonlinear system identification under artificial datasets.Numerical results show that the SAF-Adagrad and SAFRMSProp algorithms have better convergence performance than some existing SAF algorithms(i.e.,SAF-SGD,SAF-ARC-MMSGD,and SAF-LHC-MNAG).The analysis results of various measured real datasets also verify this conclusion.Overall,the effectiveness of SAF-Adagrad and SAF-RMSProp are confirmed for the accurate identification of nonlinear systems.
基金Supported by the National Key Research and Development Program of China(2018YFC1506601)National Natural Science Foundation of China(91437220)China Meteorological Administration Special Public Welfare Research Fund(GYHY201306045 and GYHY201506002).
文摘A real-time,long-term surface meteorological blended forcing dataset(SMBFD)has been developed based on station observations,satellite retrievals,and reanalysis products in China.The observations are collected at national and regional automatic weather stations,satellite data are obtained from the Fengyun(FY)series satellites retrievals,and the reanalysis products are obtained from the ECMWF.The 90-m resolution digital terrain elevation data in China are obtained from the Shuttle Radar Topographic Mission(SRTM)for temperature and humidity elevation adjustment.The dataset includes 2-m air temperature and humidity,10-m zonal and meridional winds,downward shortwave radiation,surface pressure,and precipitation.The spatial resolution is 1 km,and the temporal resolution is 1 h.During the data processing procedure,various data fusion techniques including the space–time multiscale variational analysis,the discrete ordinates radiative transfer(DISORT)model,the hybrid radiation estimation model,and a terrain correction algorithm are employed.Dependent and independent evaluations of the dataset are performed against observations.The SMBFD dataset is also compared with similar datasets produced in other major meteorological operational centers in the world.The results are as follows.(1)All variables show reasonable geographic distribution features and realistic spatial and temporal variations.(2)Dependent and independent evaluations both indicate that the gridded SMBFD dataset is close to the observations,while the dependent evaluation yields better results than the independent evaluation.(3)Compared with similar datasets produced in other meteorological operational centers,the real-time and retrospective surface meteorological fusion data obviously have higher quality.The dataset introduced in the present study is in general stable and accurate,and can be applied in various practice such as meteorology,agriculture,ecology,environmental protection,etc.Meanwhile,this dataset has been used as the atmospheric forcing data to drive the operational High-resolution Land Data Assimilation System of China Meteorological Administration.The dataset with the network Common Data Form(NETCDF)can be decoded by various programming languages,and it is freely available to non-commercial users.
