基于视频和人体姿态估计的老年人摔倒监测研究

人口老龄化所带来的养老服务问题是现代社会面临的严重问题。例如在很多国家跌倒是造成老年人因伤致死的最大原因,因此如何对老年人进行自动摔倒监测就成为养老服务亟待解决的问题。目前,在室内摔倒监测领域中,基于可穿戴设备和基于环境传感器等主流摔倒监测方法面临着设备复杂、成本较高等问题。鉴于此,将人体姿态估计引入摔倒监测领域,提出了一种基于2D视频的摔倒监测算法。首先利用OpenPose数据集提取原始数据中人体关节的位置;其次利用这些具有增强特征的数据构建静态分类模型和动态分类模型;最后,在3个公共摔倒数据集上进行模型训练和摔倒监测的测试,取得了较好的效果,可以为摔倒监测相关研究提供一定的参考。     

基于六轴MEMS器件姿态测量系统

为了解决噪声和漂移等原因造成的误差不断累积的问题,针对陀螺仪的静态性能以及加速度计的动态性能,提出了一种利用六轴MEMS器件对照相机三脚架的稳定测量系统,介绍了MEMS器件的工作原理,介绍硬件系统和软件系统,完成了基于ADXR450陀螺仪和ADXL355加速度计的检测硬件系统设计,通过传感器获取角速度加速度信息,采用不同的滤波方式对输出结果进行了分析,比较卡尔曼滤波和一阶RC数字滤波;经比较,卡尔曼滤波实时性更好,一阶RC滤波动态响应更好;实验证明,系统静态下更适用于一阶滤波,计算出姿态测量角度误差在0.104°以内,得到理想的姿态信息,能有效地提高检测目标姿态的精度。     

车用摄像头锁紧装置姿态调整系统设计

针对汽车倒车用摄像头在装配过程中对锁紧装置的自动化装配能够实现锁紧装置的姿态调整要求,设计了姿态自动调整系统。采用激光照明和PIN-FET （Field Effect Transistors）混合集成光电接收器配合定时器联合工作,对锁紧装置的姿态进行识别,并通过翻转装置对锁紧装置的姿态进行调整,送由机械手完成锁紧装置的拾取与安装,实现车用摄像头的自动锁紧装配。结果表明,基于激光照明和PIN-FET混合集成光电接收器配合定时器联合工作的姿态自动调整系统,正确率高,锁紧装置在运送过程中即可完成识别,有效提高安装效率。     

Modal analysis of human body vibration model for Indian subjects under sitting posture

Need and importance of modelling in human body vibration research studies are well established. The study of biodynamic responses of human beings can be classified into experimental and analytical methods. In the past few decades, plenty of mathematical models have been developed based on the diverse field measurements to describe the biodynamic responses of human beings. In this paper, a complete study on lumped parameter model derived from 50th percentile anthropometric data for a seated 54- kg Indian male subject without backrest support under free un-damped conditions has been carried out considering human body segments to be of ellipsoidal shape. Conventional lumped parameter modelling considers the human body as several rigid masses interconnected by springs and dampers. In this study, concept of mass of interconnecting springs has been incorporated and eigenvalues thus obtained are found to be closer to the values reported in the literature. Results obtained clearly establish decoupling of vertical and fore-and-aft oscillations.     

Macro repositioning movements of pelvis and trunk during simulated driving tasks

Low back pain (LBP)³ is one of the most important features of discomfort in prolonged sitting postures. This is evidenced by an increase in the number of postural changes called macro-movements. The focus of study was the frequencies of macro repositioning movements in prolonged sitting posture resulting from the perception of discomfort caused by low-back pain. Eighteen (18) drivers performed driving tasks for a period of 90 min. The results indicate that LBP participants performed more movements (10.52 rep/h), compared to without low back pain participants (8.52 rep/h). However, the latter moved more than expected due to joint macro movements of the pelvis and trunk. LBP subjects generally performed macro-repositioning movements of the trunk only, while without low back pain (WLBP)⁴ subjects moved pelvis and trunk simultaneously. Although the WLBP participants moved less than those with LBP symptoms, they applied different movement strategies which should be considered in further research. Finally, the authors also propose that future research should be directed towards determining the influence of these lumbo pelvic movements when provided in a controlled manner, to improve comfort in seats and to help manufactures to offer better ergonomic seats.Relevance to industryAccording to the findings of this study, control of lumbo pelvic movements could be used to improve the design and production of ergonomic seats for driving activities.     

基于动力学与控制统一模型的蛇形机器人速度跟踪控制方法研究

对带有被动轮的蛇形机器人进行速度跟踪控制时,利用传统的动力学建模方法得到的动力学方程复杂且不利于控制器的设计. 本文基于微分几何的方法将带有被动轮的蛇形机器人动力学投影到速度分布空间中, 得到了动力学与控制统一模型, 更有利于速度跟踪控制器的设计. 考虑到蛇形机器人在进行速度跟踪时容易出现奇异位形, 提出增加头部扰动速度的方法. 基于头部扰动速度和统一模型, 提出避免奇异位形的速度跟踪控制方法, 最后通过逆向动力学得到控制力矩. 文中对速度跟踪控制进行了数值仿真和实验验证. 仿真和实验结果表明, 提出的速度跟踪控制方法能够跟踪想要方向的速度, 并且在跟踪过程中可以有效地避免奇异位形.     

Biomechanics and Kinematic Responses of the Upper Extremity during Isotonic and Isokinetic Wrench‐Turning Tasks

The purpose of this study was to investigate the impact of using a wrench under isotonic (constant torque) and isokinetic (constant speed) task modes (TM) at three work surface inclinations (WSI) (0°, 45°, and 90°) on the biomechanical (muscle activity) and kinematic (joint posture) responses of the upper extremity. The muscle activity of seven muscles (trapezius posterior deltoid, anterior deltoid, triceps, biceps, brachioradialis, and flexor digitorium) and posture of four body segments (shoulder adduction/abduction, elbow flexion/extension, forearm supination/pronation and wrist flexion/extension) were obtained using surface electromyography and motion tracking, respectively. WSI showed a statistically significant effect on the muscle activity of the posterior deltoid (p = .038), triceps (p = .016), and biceps (p = .021). The least muscle activity was recorded at the 0° WSI in the isotonic TM. WSI had a significant impact on the supination (p = .017) and pronation (p = .011) of the forearm. The 45° WSI had the least impact on forearm postures. Wrenches are widely used in industries, including automobile service and maintenance, manufacturing, carpentry, and general repair work. Their usage poses risks for the development of musculoskeletal disorders in the upper extremity. In spite of this, knowledge of their physical demands and associated impact on the upper extremity has not been well documented. This study provides empirical evidence on the biomechanical and kinematic responses of selected upper extremity muscles and limb segments and highlights task performance and workstation design factors that elicit undue levels of these responses. The results of this work can provide guidance for ergonomic interventions such as optimized task design and/or improved workstation design when it comes to wrench‐turning tasks.     

视觉导航相关技术发展及其研究分析

为了全面了解航空器应用背景下的视觉导航相关技术的研究与进展,深入分析了视觉导航关键技术。分别对与传感器图像相关的载体姿态提取技术、基于图像匹配算法的定位技术以及基于序列图像的速度测量技术这三种技术进行了全面阐述,重点论述了姿态与参考特征的关系、景像匹配定位的核心算法及速度测量的两类方法,并总结了视觉导航发展过程中的相关难点与发展趋势。     

多点传感阵列的体压分布测量系统设计

坐姿不良与许多疾病密切相关,而坐姿体压分布不平衡和一些腰椎疾病的发生有着相当大的联系。应用新型导电橡胶为传感材料,利用其力—电阻特性,研制64点阵式压电传感器型坐姿体压分布测量系统。应用本系统对1例正常人,3例患腰椎疾病者进行体压分布测量。实验结果表明:患腰椎疾病者的体压分布不平衡。证明系统的设计方案合理可行,能实现人体与座椅之间压力分布的动态、实时测量。     

基于半监督聚类的3维肢体分割算法

行为分析已经成为计算机视觉研究领域的热点,行为主体的肢体部件分割是行为分析中很重要的一部分同时也是一个难点问题,为了对3维肢体进行有效分割,提出了一种基于半监督聚类的肢体分割算法。该算法首先利用前一帧姿势估计反馈的时域信息来对3维主体进行初始的肢体部件分割;然后根据人体结构信息进一步确定行为主体上各个点与肢体部件之间的关系来得到半监督聚类的初始值;之后基于各个肢体部件的形状信息进行半监督聚类,迭代求解肢体部件分割的最优解;最后利用分割后的各个肢体部件进行行为主体的姿势参数估计。通过对IXMAS数据库中6种行为下的6个行为主体进行的肢体部件的分割实验结果证明,该新算法具有主体适应性、视角适应性以及行为适应性。