基于对应分析的训练样本的选择

本文提出一种基于对应分析的训练样本的选择方法。它从训练样本中自动地选择有代表性的典型训练样本,使得在自动分类中充分利用所采集的样本信息,以便得到满意的分类结果。通过实验与分析证明,该方法是可行的,它明显优于人工随机选择训练样本的方式。与基于Q型因子分析的训练样本选择方法相比,可以更快地得到较少的典型样本,满意的分类精度。     

Scientific field training for human planetary exploration

Forthcoming human planetary exploration will require increased scientific return (both in real time and post-mission), longer surface stays, greater geographical coverage, longer and more frequent EVAs, and more operational complexities than during the Apollo missions. As such, there is a need to shift the nature of astronauts’ scientific capabilities to something akin to an experienced terrestrial field scientist. To achieve this aim, the authors present a case that astronaut training should include an Apollo-style curriculum based on traditional field school experiences, as well as full immersion in field science programs. Herein we propose four Learning Design Principles (LDPs) focused on optimizing astronaut learning in field science settings. The LDPs are as follows:

(1)

LDP#1: Provide multiple experiences: varied field science activities will hone astronauts’ abilities to adapt to novel scientific opportunities

(2)

LDP#2: Focus on the learner: fostering intrinsic motivation will orient astronauts towards continuous informal learning and a quest for mastery

(3)

LDP#3: Provide a relevant experience—the field site: field sites that share features with future planetary missions will increase the likelihood that astronauts will successfully transfer learning

(4)

LDP#4: Provide a social learning experience—the field science team and their activities: ensuring the field team includes members of varying levels of experience engaged in opportunities for discourse and joint problem solving will facilitate astronauts’ abilities to think and perform like a field scientist.

The proposed training program focuses on the intellectual and technical aspects of field science, as well as the cognitive manner in which field scientists experience, observe and synthesize their environment. The goal of the latter is to help astronauts develop the thought patterns and mechanics of an effective field scientist, thereby providing a broader base of experience and expertise than could be achieved from field school alone. This will enhance their ability to execute, explore and adapt as in-field situations require.     

框幅式数字航空摄影飞行质量检查方法

目前框幅式航空摄影飞行质量检查依旧采用先冲印纸质像片,然后人工作业的方法。该方法耗时较长,限制了数字航空摄影快速高效优势的发挥。为此,研究出一套全数字航空摄影飞行质量的检查方法。该方法的整个实现过程完全在计算机上进行:首先,利用数字航空摄影生成的浏览影像完成影像质量检查;然后,应用飞行质量检查软件和机载POS(position and orientation system)数据检查影像的重叠度、像片旋偏角、航线弯曲度以及航高保持等质量因子;最后,根据检查结果整理和编写各类文档。为验证该方法的科学性和可靠性,以湖南省衡阳市实验区影像的人工质量检查结果与其进行对比分析,结果表明,二者具有很好的一致性,说明该方法能较好地反映航空摄影飞行成果质量,并且操作简单、效率高、成本低。     

倾斜航空摄影斜视影像分辨率分析

通过分析倾斜航摄仪的特点,提出一种面向倾斜影像的三维影像分辨率,介绍了3个方向分辨率概念以及推导了斜视影像分辨率的计算公式。结合当前市场上主流的倾斜航摄仪配置,分析了其影像分辨率变化的特点。本文得出的结论对航空摄影设计有重要指导意义,为影像量测精度分析打下了基础。     

双波长卫星激光测距

双波长卫星激光测距不必借助大气改正模型，利用双波长往返时间差就可以进行大气折射改正，采用皮秒级事件计时器等技术测量往返时间差，修正精度可以达到毫米量级，介绍了这一领域的研究进展、双波长大气改正的基本原理，分析了双波长激光测距的精度，最后对双波长卫星激光测距前景进行了展望。     

从窗口看云南天文台

简要总结了云南天文台２０年来的科研工作及人才培养情况，并对未来的发展提出了设想。     

样本输入方式对极端学习机预报日长变化的影响

针对极端学习机(Extreme Learning Machine,ELM)用于日长(Length-Of-Day,LOD)变化预报过程中,样本输入方式对预报结果的影响进行了研究。采用跨度、连续和迭代3种样本输入方式对日长变化进行预报。结果表明,不同的样本输入方式对预报结果有很大影响,样本按跨度输入的预报精度最低;样本采用连续输入方式在短期和中长期预报中预报精度较高,但计算速度较慢,较适合中长期预报;样本按迭代输入方式的短期预报精度稍优于连续输入方式,而中长期预报精度则不如连续输入方式,但具有较高的预报效率。这对于日长变化的实时快速预报有着较高的现实意义。     

构架航线对无人机影像区域网平差精度影响分析

与三角翼相比,无人机航摄影像像幅小、模型连接性差、质量轻、航摄时姿态不稳定;这些因素导致无人机航摄影像成图精度较差。为了解决该问题,传统航测方法在每两张航摄影像间布设大量控制点,增强模型刚性;GPS辅助空中三角测量通过提高曝光点的点位精度,可以减少控制点的个数,但是由于无人机像幅小,控制点约束范围小,因而仍然需要大量的控制点。针对以上问题,文章在GPS辅助空中三角测量的基础上,通过布设架构航线,进而实现增加影像之间的重叠度,增强模型刚性的目的,从而实现在减少控制点个数的同时,保证无人机航摄影像区域网平差精度的提高。实验表明,该方法与单纯的依赖GPS辅助空三技术相比,在精度相近的基础上能够减少约50%的控制点布设数量,大大减少了外业工作量。     

基于模糊自整定PID的直升机飞行控制律设计

在经典PID控制律基础上,引入模糊控制的概念设计模糊自整定PID控制律改进无人直升机姿态回路控制器,实现了PID控制器的参数在线自整定,并改善了控制器性能.通过Matlab对闭环系统进行算法仿真,表明模糊自整定PID控制器能够发挥优于传统PID控制器的控制性能.最后对所设计的控制器进行半物理仿真验证,通过对比典型输入指令下经典PID和模糊自整定PID两种控制器的控制效果,验证了模糊自整定PID对系统性能的优化.