基于深度学习的空间站舱内服务机器人视觉跟踪

为提升舱内跟随服务机器人的任务辅助能力,解决机器人对航天员的视觉跟踪问题,提出了一种基于深度学习和概率模型的人体视觉跟踪算法。利用深度卷积神经网络实现了对穿着多样、姿态任意人体的稳定检测。结合人体检测结果,设计了运动预测概率模型,实现了对指定人员准确、连续的跟踪。算法对包含大多数航天员活动的多个数据集进行了验证。实验结果表明:提出的跟踪算法实现了对穿着多样、姿态任意人体的稳定跟踪,并有效避免了由于穿着相似、遮挡可能造成的误跟踪问题。该算法为空间站舱内跟随服务机器人对航天员的视觉跟踪提供了有效的解决方法。算法基于融合的RGB-D图像,工程上易于构建和实现,也可拓展到其他跟随服务机器人视觉跟踪任务中。     

Orbit determination of BDS-3 satellite based on regional ground tracking station and inter-satellite link observations

The BeiDou global navigation satellite system (BDS-3) has established the Ka-band inter-satellite link (ISL) to realize a two-way ranging function between satellites, which provides a new observation technology for the orbit determination of BDS-3 satellites. Therefore, this study presents a BDS satellite orbit determination model based on ground tracking station (GTS) observations and ISL ranging observations firstly to analyze the impact of the ISL ranging observations on the orbit determination of BDS-3 satellites. Subsequently, considering the data fusion processing, the variance component estimation (VCE) algorithm is applied to the parameter estimation process of the satellite orbit determination. Finally, using the measured data from China’s regional GTS observations and BDS-3 ISL ranging observations, the effects of ISL ranging observations on the orbit determination accuracy of BDS-3 satellites are analyzed. Moreover, the impact of the VCE algorithm on the fusion data processing is evaluated from the aspects of orbit determination accuracy, Ka-band hardware delay parameter stability, and ISL ranging observation residuals. The results show that for China’s regional GTSs, the addition of BDS-3 ISL ranging observations can significantly improve the orbit determination accuracy of BDS-3 satellites. The observed orbit determination accuracy of satellite radial component is improved from 48 cm to 4.1 cm. In addition, when the initial weight ratio between GTS observations and ISL ranging observations is not appropriate, the various indicators which include orbit determination accuracy, ISL hardware delay, and ISL observation residuals were observed to have improved after the adjustment of the VCE algorithm. These results validate the effectiveness of the VCE algorithm for the fusion data processing of the GTS observations and ISL ranging observations.     

高能(450 kV)X射线参考辐射场附加过滤研究

利用MCNP程序对450kV X光机建立模型，计算各参考辐射质下的原级谱、平均光子能量。对ISO 4037-1中已有范围的管电压与附加过滤片厚度关系曲线进行二次拟合，确定模拟厚度范围。利用MCNP计算各附加过滤组合下的平均光子能量值，并与外推法得出的平均光子能量值进行对比，将相对偏差最小时的附加过滤作为最终确定值。从部分能量点的模拟与测量结果的对比可看出，模拟得出的平均光子能量在2%以内与PTB及中国计量院测量结果保持一致，附加过滤最大相对偏差为18%。     

捷联式航空重力测量系统的数据处理方法比较与分析

重力数据处理对获取高精度重力异常值有着重要作用，是重力测量的核心技术。重力仪在搭载运动载体进行重力测量时，载体的高频振动对重力测量数据和GPS数据均会产生不可避免的干扰，导致提取的重力异常粗值含有大量高频噪声。围绕重力数据的处理方法这一核心技术，介绍了FIR低通滤波、零相移滤波、标准Kalman滤波、正反Kalman滤波4种滤波方法的基本原理，运用这4种方法处理了SAG捷联式重力仪的某次实际飞行测量数据，比对了基于SINS/GPS组合导航和SINS/DGPS组合导航的重力测量数据处理结果。通过对本次试验重复测线内符合精度进行比对，验证了4种方法的可行性和优劣性，同时验证了SAG重力仪的测量精度。     

基于室内无线模型的穿墙损耗校正

K-M（Keenan-Motley）模型将单墙固定损耗值相加来计算室内无线信号穿透多墙的总损耗值，存在较大误差。针对该问题选取多种室内场景分别进行连续波（Continuous Wave，CW）测试，对无线信号穿墙损耗的影响因素和变化规律进行分析，提出了一种基于人工神经网络的室内无线模型穿墙损耗校正方法，对预处理后的测试数据进行训练并建立预测模型。经验证该预测模型符合校正判别准则，在实际场景下具有良好的预测准确度。     

混合式CRP偏转工况水动力和空泡性能试验研究

为了探究混合式CRP (混合式对转桨吊舱推进器)偏转工况下的水动力性能,在空泡水筒中开展了混合式CRP推进器偏转工况下的敞水试验和空泡试验。在其吊舱推进器偏转0°,±5°,±10°下测量了前、后桨的推力和扭矩,并对前后桨的空泡形态进行了观测。试验结果表明:前桨的推力、扭矩系数及空泡形态基本不受吊舱推进器偏转影响;后桨推力系数和扭矩系数随着偏转角度的增加而增加。偏转工况下,前桨的毂涡和梢涡会下泄至后桨桨叶,引起后桨叶面上产生大量空泡;在不同周向位置,后桨的空泡面积不同;后桨尾涡强度在吊舱体两侧呈现出差异性,较强的前桨尾涡会将较弱的后桨尾涡卷入,前后桨尾涡强度相当时,尾涡轨迹发生紊乱。     

基于IMU/FADS/无线电的火星上升器组合导航方案

以火星采样返回任务中火星表面上升为背景，研究了基于惯性测量单元（Inertial Measurement Unit, IMU）、嵌入式大气数据传感系统（Flush Air Data Sensing System, FADS）和无线电信标的组合导航方法。首先，在传统的IMU导航框架中加入由无线电测量获得的相对距离、速度信息，以及由FADS获取的动压、温度数据，建立了基于IMU、无线电和FADS的导航观测模型；然后，基于无迹卡尔曼滤波（Unscented Kalman Filter, UKF）技术对测量信息进行了融合，并压制了过程噪声和测量噪声，从而对上升器的状态进行了联合估计；最后，在数值仿真中，将UKF与自适应无迹卡尔曼滤波（Adaptive Unscented Kalman Filter, AUKF）技术进行了对比，在比较不同滤波器性能的同时，验证了组合导航方法的有效性。     

Electrochemical micromachining of micro-dimple arrays on cylindrical inner surfaces using a dry-film photoresist

The application of surface textures has been employed to improve the tribological performance of various mechanical components. Various techniques have been used for the application of surface textures such as micro-dimple arrays, but the fabrication of such arrays on cylindrical inner surfaces remains a challenge. In this study, a dry-film photoresist is used as a mask during through-mask electrochemical micromachining to successfully prepare micro-dimple arrays with dimples 94 lm in diameter and 22.7 lm deep on cylindrical inner surfaces, with a machining time of 9 s and an applied voltage of 8 V. The versatility of this method is demonstrated, as are its potential low cost and high efficiency. It is also shown that for a fixed dimple depth, a smaller dimple diameter can be obtained using a combination of lower current density and longer machining time in a passivating sodium nitrate electrolyte.