面向航天测发任务的虚拟手交互技术研究

发布时间：2018-04-16 18:56

本文选题：MEMS惯性传感器 + 运动跟踪　；参考：《南京理工大学》2017年硕士论文

【摘要】：航天测发任务是一项高风险,高投资的复杂系统工程,在测试过程中对试验场操作人员的现场操作能力要求较高。虚拟现实技术作为信息传递与交流的手段,为测发任务提供虚拟环境,操作人员可以通过驱动虚拟手在虚拟环境中完成检测任务,以此提高操作效率。论文以航天测发任务为研究背景,主要研究基于MEMS惯性传感器的手部跟踪技术,实现手势识别以及语义交互的设计。为实现虚拟手的运动跟踪,开发了一套基于惯性传感器的手部运动跟踪系统。通过分析人手的骨骼结构和运动特性,构建了三维手部骨骼模型。对手部初始姿态进行标定,将传感器坐标系转换到手部骨骼坐标系,并提出了一种手部骨骼比例的推算方法,使虚拟手更加真实。通过基于运动学的虚拟手位置解算,实现了手部运动的实时跟踪。对设计的典型手势进行数据采集,提取行为特征来表示手势,通过主成分分析法对手势特征数据进行降维,应用网格搜索和5折交叉验证的SVM分类算法。针对分类算法,分别使用多项式核函数,径向基核函数以及组合核函数对比识别率,选取最优的核函数,实现手势识别。为实现航天测发任务中的交互研究,采用隐马尔可夫模型对手部动作序列识别。面向特定场景设计了虚拟手的语义交互过程,并在场景中进行验证。论文的研究成果在航天发射任务仿真虚拟环境中得到应用验证。
[Abstract]:The mission of spaceflight testing is a complex system engineering with high risk and high investment.Virtual reality technology, as a means of information transmission and communication, provides a virtual environment for testing and sending tasks, and the operator can improve the operation efficiency by driving virtual hands to complete the detection tasks in the virtual environment.Based on the research background of spaceflight hair testing task, this paper mainly studies the hand tracking technology based on MEMS inertial sensor, realizes the design of gesture recognition and semantic interaction.In order to realize the motion tracking of virtual hand, a hand motion tracking system based on inertial sensor is developed.By analyzing the skeletal structure and motion characteristics of human hand, a three-dimensional hand bone model was constructed.The initial attitude of the hand is calibrated, the sensor coordinate system is converted to the skeletal coordinate system of the hand, and a method of calculating the ratio of the hand bone is proposed to make the virtual hand more realistic.The real-time tracking of hand motion is realized by solving the position of virtual hand based on kinematics.The typical gestures are collected and the behavior features are extracted to represent the gestures. The dimensionality of gesture features is reduced by principal component analysis (PCA), and the SVM classification algorithm based on grid search and 5 fold cross-validation is applied.For the classification algorithm, the polynomial kernel function, the radial basis kernel function and the combined kernel function are used to compare the recognition rate, and the optimal kernel function is selected to realize gesture recognition.In order to realize the interactive research of spaceflight mission, Hidden Markov Model (hmm) is used to recognize the hand motion sequence.The semantic interaction process of virtual hand is designed for specific scene and verified in the scene.The research results of this paper have been applied to the simulation virtual environment of space launch mission.
【学位授予单位】：南京理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：V55;TP391.9

【参考文献】