一种深度学习分割遥感影像道路的损失函数

传统的根据光谱特征或形态学算法来分割道路,存在精度低、阈值难确定等缺点,而深度学习中已有的方法并未考虑道路的特性,只是利用通用方法来分割道路.针对上述不足,提出了一种针对道路特有形态的深度学习损失函数——形态损失函数.首先使用连通性算法将预测结果划分为若干个相互分离的连通区域,分别计算这些区域的面积与外接圆面积的比值,...     

聚酰亚胺薄膜与碳网格共固化基板成型工艺研究

为满足未来深空探测对航天器太阳翼基板更高的耐温性需求,提出一种新型的太阳翼基板成型工艺方法：聚酰亚胺薄膜与碳网格面板共固化成型。重点研究了聚酰亚胺薄膜与网格面板的剥离性能、网格面板的拉伸性能、蜂窝夹层结构的弯曲性能、共固化基板的耐温性能,并与J-133胶粘贴聚酰亚胺薄膜的传统工艺进行对比。结果表明,共固化工艺的聚酰亚胺薄膜剥离强度约为传统工艺的2倍,且在140℃下的各项力学性能保持率均大于87%,具有很好的耐温性;共固化基板在经历-100~140℃热真空试验后,外观质量及各项性能均合格,满足太阳翼基板耐140℃及以下空间环境的使用需求。     

基于MADDPG的多无人机协同任务决策

针对多无人机任务决策方法研究中传统优化算法难以在短时间内得到期望结果的问题,基于深度强化学习提出一种无人机多智能体深度确定性策略梯度(MADDPG)算法,通过允许无人机在学习时使用全局信息,在应用决策的时候只使用局部信息的方法,从网络结构、状态空间、动作空间和奖励函数设计了 MADDPG算法的模型结构.最后通过仿真实验...     

Improved specular point prediction precision using gradient descent algorithm

Global Navigation Satellite Systems Reflectometry (GNSS-R) utilizes GNSS signals reflected off the Earth surface for remote sensing applications. Due to weak power of reflected signals, GNSS-R receiver needs to track reflected signals by open loop. The first step is to calculate the position of specular point. The specular point position error of the existing algorithm—Quasi-Spherical Earth (QSE) Approach—is about 3 km which may cause troubles in data post-processing. In this paper, gradient descent algorithm is applied to calculate position of specular point and the calculation is based on World Geodetic System 1984 (WGS 84) ellipsoid in geodetic coordinate. The benefit of this coordinate is that it is easy to investigate the effect of real surface’s altitude. Learning rate—the key parameter of the algorithm—is adaptively adjusted according to initial error, latitude and gradient descent rate. With self-adaptive learning rate strategy, the algorithm converges fast. Through simulation and test on Global Navigation Satellite System Occultation Sounder II (GNOS II), the performances of the algorithm are validated. The specular point position error of the proposed algorithm is about 10 m. The speed of the proposed algorithm is competitive compared with the existing algorithm. The test on GNOS II shows that the proposed algorithm has good real-time performance.     

一种新型滑模控制算法在挠性多体卫星姿态控制中的应用

针对一类挠性多体卫星的复合控制问题，提出一种新型滑模变结构控制算法。新型算法利用闭环系统Lyapunov函数的一阶导数估计值设计控制器，且控制器采用了递归学习控制结构，能够有效解决传统滑模控制技术的颤振问题。随后根据Lyapunov稳定性理论证明闭环系统轨迹可以快速收敛到滑模面，并且系统状态误差可在滑模面上渐近收敛到零。此外，设计的控制器能够有效抑制外部干扰，而且控制器只需要控制输入矩阵信息而不需受控系统和未知参数的其他先验信息，使得算法具有较强鲁棒性。最后通过数值仿真与现有文献中控制算法进行对比，结果充分验证了本文设计控制算法的有效性和实用性。     

Retrieval and validation of aerosol optical depth using GF-1 WFV cameras data

Conventional AOD (Aerosol Optical Depth) retrieval is restricted to the global and regional scale due to the limited spatial resolution of satellites. This does not allow for aerosol monitoring at the city level. The Chinese GF-1 Wide Field of View (WFV) sensors have sufficiently fine resolution as a data source for AOD retrieval with fine spatial resolution and a 4-day revisit time. In this study, principles similar to those in the Deep Blue (DB) and Dark Target (DT) algorithms were used to retrieve AOD at 100 m spatial resolution from GF-1 WFV images supported by Moderate Resolution Imaging Spectraradiometer (MODIS) surface reflectance (SR) products (MOD09A1). The derived GF-1 WFV AOD were compared with a combination of MOD04_3K DT AOD and MOD04_L2 DB AOD (MODIS AOD) to find that they yield reasonable Spearman correlations (R_S > 0.82) over Taiwan and Beijing. The derived GF-1 WFV AOD were also validated against Aerosol Robotic Network (AERONET) AOD; the Spearman correlation values were R_S = 0.911 in Beijing and R_S = 0.858 in Taiwan.     

基于机器视觉的万用表数字识别技术研究

为提高手持式数字万用表校准系统的自动化水平，本文通过研究图像灰度阈值法、传统方法的目标分类和深度学习三种不同类型的字符识别技术，提出了两种基于机器视觉的数字识别方案。测试结果显示，两种方案的字符识别准确率均可达到99.8%，但其在硬件资源占比、编程难易程度上二者还存在显著差异。该机器视觉字符识别功能的成功开发与应用，可为更多无程控通信接口的计量测试设备，及一些不适于人工作业的危险计量工作环境进行类似的数字识别提供借鉴。     

Reliability and reliability sensitivity analysis of structure by combining adaptive linked importance sampling and Kriging reliability method

The application of reliability analysis and reliability sensitivity analysis methods to complicated structures faces two main challenges: small failure probability (typical less than 10⁻⁵) and time-demanding mechanical models. This paper proposes an improved active learning surrogate model method, which combines the advantages of the classical Active Kriging – Monte Carlo Simulation (AK-MCS) procedure and the Adaptive Linked Importance Sampling (ALIS) procedure. The proposed procedure can, on the one hand, adaptively produce a series of intermediate sampling density approaching the quasi-optimal Importance Sampling (IS) density, on the other hand, adaptively generate a set of intermediate surrogate models approaching the true failure surface of the rare failure event. Then, the small failure probability and the corresponding reliability sensitivity indices are efficiently estimated by their IS estimators based on the quasi-optimal IS density and the surrogate models. Compared with the classical AK-MCS and Active Kriging – Importance Sampling (AK-IS) procedure, the proposed method neither need to build very large sample pool even when the failure probability is extremely small, nor need to estimate the Most Probable Points (MPPs), thus it is computationally more efficient and more applicable especially for problems with multiple MPPs. The effectiveness and engineering applicability of the proposed method are demonstrated by one numerical test example and two engineering applications.     

高校思想政治教育工作队伍整体优化的内涵与进路

基于辩证系统思维引领,科学界定高校思想政治教育工作队伍整体优化的内涵,创新开启立足于微观、中观和宏观三个不同层面的多维研究视角,将思想政治教育工作队伍理论概括为三个不同层面上的"三位一体"化大军,并通过确定建构理想目标、把握发展趋势、整合资源力量、追求最高境界的进路,将高校思想政治教育工作队伍整体优化的理论研究引向深入,促使高校思想政治教育工作实效性进一步增强。     

强化词块训练 提高英语听说能力

以词块教学理论为依据,分析在听说教学中进行词块训练的可行性。研究词块训练在听说课堂中的具体实施过程,并通过问卷调查和访谈结果,论证了听说课上引入词块理念能提高英语学习者听说能力从而提出词块学习对语言习得的有效性。