基于BP神经网络的连续动作识别在清淤设备中的应用
|
摘要
在清淤作业中清淤设备的工作状况和能耗等级与设备所完成的动作密切相关,为评估清淤设备的这些参数,将核心的连续作业动作准确地从总作业过程中识别出来是重要的手段之一。将多个三维混合传感器分别安装于机械臂的各主要部分上,采集大量清淤作业时机械臂的运动数据,并建立挖斗型机械臂的模型。通过对数据进行分析,提出一种用于动作分割的复合特征。在完成动作分割的基础上,通过分析分割出的时间间隔内的连续动作,提出一种位移特征,并采用数据作为特征参数的BP(Back Propagation)神经网络进行动作判定。经过对试验数据验证,该方法的对于挖-移-卸这一连续动作的识别率达到98. 5%。
The working condition and the energy consumption level of desilting equipment is closely related to its action performed in the period of desilting operation. In order to evaluate these parameters,it is necessary to identify and separate key continuing operational action from the process. The model of the Bucket shape mechanical arm is constructed and a number of 3 D hybrid sensors are installed on the main parts of the mechanical arm to monitor the motion of the arm during the desilting operation. A comprehensive feature for motion segmentation is identified through analyzing the data from the sensors. After motion segmentation,the continuing action within time interval of each segment is analyzed and the displacement feature of continuing action is identified through analyzing the motion segment-wise. The Back Propagation( BP) neural network is used to recognize the interested action according to the feature. The developed algorithm is verified through a number of experiments which demonstrate the recognition accuracy as high as 98. 5%.
The working condition and the energy consumption level of desilting equipment is closely related to its action performed in the period of desilting operation. In order to evaluate these parameters,it is necessary to identify and separate key continuing operational action from the process. The model of the Bucket shape mechanical arm is constructed and a number of 3 D hybrid sensors are installed on the main parts of the mechanical arm to monitor the motion of the arm during the desilting operation. A comprehensive feature for motion segmentation is identified through analyzing the data from the sensors. After motion segmentation,the continuing action within time interval of each segment is analyzed and the displacement feature of continuing action is identified through analyzing the motion segment-wise. The Back Propagation( BP) neural network is used to recognize the interested action according to the feature. The developed algorithm is verified through a number of experiments which demonstrate the recognition accuracy as high as 98. 5%.
引文
[1]成娟.基于表面肌电和加速度信号融合的动作识别和人体行为分析研究[D].安徽:中国科学技术大学,2013.
[2]张旭.基于表面肌电信号的人体动作识别与交互[D].安徽:中国科学技术大学,2010.
[3]尤波,李忠杰,黄玲.基于改进型BP神经网络的手部动作识别[J].智能系统学报,DOI:10. 11992/tis. 201703018.
[4]熊俊涛,刘梓健,孙宝霞,等.基于视觉技术的手势跟踪与动作识别算法[J].计算机与现代化,2014,227(7):75-79.
[5]王见,陈义,邓帅.基于改进SVM分类器的动作识别方法[J].重庆大学学报,2016,39(1):12-17.
[6]汪少初,刘昱,郝文飞,等.基于惯性传感器的人员行进动作识别方法[J].电子测量与仪器学报,2014,28(6):630-636.
[7] WANG J,LEE H,WANG C,et al. Robust Environmental Sound Recognition for Home Automation[J]IEEE Transaction on Automation Science and Engineering,2008,5(1):25-31.
[8] ZHANG Y,ROCKETT P. Application of multiobjective Genetic Programming to the Design of Robot Failure Recogniton Systems[J]. IEEE Transaction on Automation Science and Engineering,2009,6(2):372-376.
[9] ROB S,CHOI H R. 3-D Tag-Based RFID System for Recogniton of Object[J]. IEEE Transaction on Automation Science and Engineering,2009,6(1):55-65.
[10] HUANG J,HUO W,XU W,et al. Control of UpperLimb Power-Assist Exoskeleton Using a Human-Robot Interface Based on Motion Intention Recogniton[J].IEEE Transaction on Automation Science and Engineering,2015,12(4):1257-1270.
[11] HAMED J A,FARROKH J. Adaptive Neural Network Control of Cable-Driven Parallel Robots with Input Saturation[J]. Engineering Applications of Artificial Intelligence,2017,65:252-260.
[12] PEDRAM,MAHDI,ALI. Adaptive Recurrent Neural Network with Lyapunov Stability Learning Rules for Robot Dynamic Terms Identification[J]. Engineering Applications of Artificial Intelligence,2017,65:1-11.
[13] ZHOU X,GANDHI S,SURI S. et al. e BAY in the Sky:Strategy-Proof Wireless Spectrum Auctions[C]//Proceedings of the 14th ACM International Conference on Mobile Computing and Networking,2008:2-13.
[14] CHEN Y,ZHANG J,WU K,et al. TAMES:a Truthful Double Auction for Multi-Demand Reterogeneous Spectrums[C]∥IEEE Transaction on Parallel and Distributed System,2014,25(11):3012-3024.
[15]白栋天,张磊,黄华. RGB-D视频中连续动作识别[J].中国科技论文,2016,11(2):168-172.
[2]张旭.基于表面肌电信号的人体动作识别与交互[D].安徽:中国科学技术大学,2010.
[3]尤波,李忠杰,黄玲.基于改进型BP神经网络的手部动作识别[J].智能系统学报,DOI:10. 11992/tis. 201703018.
[4]熊俊涛,刘梓健,孙宝霞,等.基于视觉技术的手势跟踪与动作识别算法[J].计算机与现代化,2014,227(7):75-79.
[5]王见,陈义,邓帅.基于改进SVM分类器的动作识别方法[J].重庆大学学报,2016,39(1):12-17.
[6]汪少初,刘昱,郝文飞,等.基于惯性传感器的人员行进动作识别方法[J].电子测量与仪器学报,2014,28(6):630-636.
[7] WANG J,LEE H,WANG C,et al. Robust Environmental Sound Recognition for Home Automation[J]IEEE Transaction on Automation Science and Engineering,2008,5(1):25-31.
[8] ZHANG Y,ROCKETT P. Application of multiobjective Genetic Programming to the Design of Robot Failure Recogniton Systems[J]. IEEE Transaction on Automation Science and Engineering,2009,6(2):372-376.
[9] ROB S,CHOI H R. 3-D Tag-Based RFID System for Recogniton of Object[J]. IEEE Transaction on Automation Science and Engineering,2009,6(1):55-65.
[10] HUANG J,HUO W,XU W,et al. Control of UpperLimb Power-Assist Exoskeleton Using a Human-Robot Interface Based on Motion Intention Recogniton[J].IEEE Transaction on Automation Science and Engineering,2015,12(4):1257-1270.
[11] HAMED J A,FARROKH J. Adaptive Neural Network Control of Cable-Driven Parallel Robots with Input Saturation[J]. Engineering Applications of Artificial Intelligence,2017,65:252-260.
[12] PEDRAM,MAHDI,ALI. Adaptive Recurrent Neural Network with Lyapunov Stability Learning Rules for Robot Dynamic Terms Identification[J]. Engineering Applications of Artificial Intelligence,2017,65:1-11.
[13] ZHOU X,GANDHI S,SURI S. et al. e BAY in the Sky:Strategy-Proof Wireless Spectrum Auctions[C]//Proceedings of the 14th ACM International Conference on Mobile Computing and Networking,2008:2-13.
[14] CHEN Y,ZHANG J,WU K,et al. TAMES:a Truthful Double Auction for Multi-Demand Reterogeneous Spectrums[C]∥IEEE Transaction on Parallel and Distributed System,2014,25(11):3012-3024.
[15]白栋天,张磊,黄华. RGB-D视频中连续动作识别[J].中国科技论文,2016,11(2):168-172.