核空间聚类在图像纹理分类中的简化算法

模糊c均值聚类已广泛应用于模糊模式识别领域,但对于线性不可分数据并不适用.在核方法中通过将输入数据经过非线性映射投影到高维特征空间来解决非线性分类的问题.将传统的模糊c均值聚类算法应用于核空间中,对线性不可分的样本进行了核空间聚类的分类实验,得到了正确的分类结果.由于图像分类中分类样本(对应图像像素)数目庞大,造成了核空间聚类算法中特征距离的计算量过大.因此,在核空间聚类的基础上,提出了对图像先进行过分割,再对过分割的图像块进行核空间聚类的方法,大大降低了高维空间特征距离计算的运算成本,并取得了良好的分类效果.      

Long-wavelength lunar gravity field recovery from simulated orbit and inter-satellite tracking data

As has been demonstrated recently, inter-satellite Ka-band tracking data collected by the GRAIL (Gravity Recovery And Interior Laboratory) spacecraft have the potential to improve the resolution and accuracy of the lunar gravity field by several orders of magnitude compared to previous models. By means of a series of simulation studies, here we investigate the contribution of inter-satellite ranging for the recovery of the Moon’s gravitational features; the evaluation of results is made against findings from ground-based Doppler tracking. For this purpose we make use of classical dynamic orbit determination, supported by the analysis of satellite-to-satellite tracking observations. This study sheds particularly light on the influence of the angular distance between the two satellites, solar radiation modeling and the co-estimation of the lunar Love number k₂. The quality of the obtained results is assessed by gravity field power spectra, gravity anomalies and precision orbit determination. We expect our simulation results to be supportive for the processing of real GRAIL data.     

基于影像数据的冯·卡门撞击坑形貌分析

月球背面南极—艾肯（South Pole-Aitken,SPA）盆地内的冯·卡门（Von Kármán）撞击坑是“嫦娥4号”初步选定的着陆区,该区域的地形地貌分析是任务设计的一个重要环节。根据LOLA （Lunar Orbiter Laser Altimeter）高程数据和LROC（Lunar Reconnaissance Orbiter Camera）影像数据,建立了马尔科夫随机场模型,从聚类的角度对该区域的地形地貌进行了表示和分析。具体而言,利用模型中的似然函数对呈近似正态分布的观测数据进行了建模,模型的标记随机场则刻画了不同地物形貌间的空间关系,并通过概率推断求解出最终的聚类结果。实验结果表明,从聚类角度进行数据表示,可以更好地展示出冯·卡门撞击坑内低对比度区域的地形地貌;而通过设置不同的聚类数目,还有助于分析冯·卡门撞击坑内整体的地形分布与局部的典型地貌。     

基于改进蚁群算法的月面机器人的路径规划及虚拟仿真

针对月面机器人在复杂地形下的路径规划问题,提出了一种改进的蚁群算法。算法构建了栅格化地形图,基于人工势场法改进了蚁群算法的启发函数,加快了算法收敛速度;引入空间信息素划分方法,提高了蚁群在最短路径附近区域的搜索能力;实验证明,改进后的蚁群算法,路径规划成功率显著提高,收敛速度加快。在算法规划出月面机器人的最短路径后,采用虚拟仿真技术,基于unity3D构建虚拟月面环境和月球车,直观地展示了月面机器人在月面环境下的路径规划效果。     

嫦娥三号着陆区月表辐射特性测算

月球表面辐射特性研究是月球探索的一个重要领域,对于勘测月球表面形态及进行资源勘探等具有重要意义,但是目前在月球表面直接进行近距离测量的数据非常少.本文建立了基于嫦娥三号着陆器地形地貌相机和缓释电机在轨遥测数据研究月球表面辐射特性的方法,初步测算得到嫦娥三号着陆区月球表面的太阳辐射光谱反射率为0.105,发射率为0.866.所得结果能够为进一步空间探测任务提供原始测算数据,以降低设计难度及提高探测能力.      

Depth dependency of neutron density produced by cosmic rays in the lunar subsurface

Depth dependency of neutrons produced by cosmic rays (CRs) in the lunar subsurface was estimated using the three-dimensional Monte Carlo particle and heavy ion transport simulation code, PHITS, incorporating the latest high energy nuclear data, JENDL/HE-2007. The PHITS simulations of equilibrium neutron density profiles in the lunar subsurface were compared with the measurement by Apollo 17 Lunar Neutron Probe Experiment (LNPE). Our calculations reproduced the LNPE data except for the 350–400 mg/cm² region under the improved condition using the CR spectra model based on the latest observations, well-tested nuclear interaction models with systematic cross section data, and JENDL/HE-2007.     

Identification and characterization of science-rich landing sites for lunar lander missions using integrated remote sensing observations

Despite more than 52 years of lunar exploration, a wide range of first-order scientific questions remain about the Moon’s formation, temporal evolution, and current surface and interior properties. Addressing many of these questions requires obtaining new in situ analyses or return of lunar surface or shallow subsurface samples, and hence rely on the selection of optimal landing sites. Here, we present an approach to optimize science-rich lunar landing site selection studies based on the integration of remote sensing observations. Currently available remote sensing data, as well as features of interest published in the recent literature, were integrated in a Geographic Information System. This numerical database contains geographic information about all these findings, which can be consulted and used to simultaneously display multiple features and parameters of interest. To illustrate our approach, we identified the optimal landing sites to address the two top priorities (or goals) relative to Concept 3 of the National Research Council of the National Academies (2007), namely to ‘Determine the extent and composition of the primary feldspathic crust, (ur)KREEP layer, and other products of differentiation’ and to ‘Inventory the variety, age, distribution and origin of lunar rock types’. We review site requirements and propose possible landing sites for both these goals. We identified 29 sites that best fulfill both these goals and compare them with the landing sites of planned future lunar lander missions. Finally, we detail two of these science-rich sites (Aristarchus and Theophilus craters) which are particularly accessible through their location on the nearside.     

Analysis of data of “Clementine” and “KAGUYA” missions and “ULCN” and “KSC-1162” catalogues

In this paper an analysis of data coordinate systems from selenographic catalogues and space missions was carried out. The lunar macrorelief models were made on basis of the software package ASNI USTU using method of the spherical harmonic expansion. These models accurately describe the global features of the lunar figure. To construct these models the following sources of topographic information were used: “Clementine” and “KАGUYА” (Selena, Japan mission) missions, “KSC-1162” (Kazan selenocentric catalogue), “Kiev” (selenodesic catalogue), “SAI” (Chuikova (1975)), “Bills, Ferrari”, “ULCN” (The Unified Lunar Control Network 2005). Direct comparison hypsometric information “KSС-1162” catalogue data with “Clementine” mission was carried out. These researches confirmed a good agreement of the hypsometric information of compared systems. The normalized coefficients were obtained on basis of the hypsometric information expansion for eight sources. The displacement of the lunar center of mass (LCM) relatively to the lunar center of figure (LCF) was obtained by using topographic data selenodetical catalogues and space missions.     

基于核方法的聚类算法及其应用

在分析核方法的核心概念基础上,提出了一种基于核方法的聚类算法.通常,传统聚类算法只在数据特征差异较大时才有效,当数据特征差异较小时,很难取得较好的聚类效果.引入核函数,将原始数据由数据空间映射到特征空间,在特征空间中进行聚类.核函数的非线性映射使得原始数据的特征更完整地显现出来,从而能够更客观准确地聚类.与传统聚类方法相比,该方法聚类结果更客观有效.以16组实际数据为例,将该方法应用于数据分类研究中,聚类结果表明了该方法的可行性和有效性,从而为数据分类提供了一种新的可行方法.      

月面模拟环境的SIFT特征提取与匹配方法

为了满足月面巡视探测器的自主导航要求.使用一种新的基于尺度不变的特征点提取和匹配算法.首先根据尺度不变特征变换方法从图像中提取关键点作为特征点,然后进行左右双目图像的特征点匹配和视差的恢复.与传统特征算法相比,可以提高对不同光照环境图像匹配的鲁棒性和匹配精度.在模拟试验场的双目视觉图像匹配中,仿真实验取得较好的效果.     

基于快速模拟退火的组合聚类算法

应用模拟退火算法解决组合聚类问题有两方面,一是有效利用基础聚类作为先验信息,以获得尽可能好的组合聚类结果;二是降低模拟退火过程的随机性,提高算法收敛速度。针对这2个问题,提出了基于投票的快速模拟退火（BV-RSA）模型。该模型利用基础聚类对样本划分的完全或部分一致性作为启发信息,构建超点集合和超点投票箱,由超点取代其代表的样本子集参与退火过程,超点运动方向在投票箱范围内随机选择,降低了超点运动随机性,加速了组合聚类过程。数据集实验表明,BV-RSA模型在聚类精度和鲁棒性方面表现良好。      

Carbonaceous materials in cometary environments: the case of the 3-4 micrometers bands.

We present in this paper a simulation of cometary spectra between 3 and 4 micrometers performed by using the optical properties of various carbon-based materials measured at different temperatures in the range 300-520 K. In our computations we have used new laboratory data obtained for hydrogenated amorphous carbon (HAC) grains and three kinds of polycyclic aromatic hydrocarbons (PAHs). All these materials show significant features in the near IR region; however, only the synthetic spectra obtained from HAC grains show a satisfactory agreement with the profile of the cometary bands.     

月球探测器垂直软着陆4D轨道全局优化

提出一种包含动态规划法与共轭梯度法的组合优化算法,求解月球探测器垂直软着陆问题. 其以动态规划法中求得的次优控制变量作为共轭梯度法的初始控制变量,求得更为精确的最优控制变量和最优轨道. 月球垂直软着陆的轨道可以分为两段,即制动段与着陆段. 以燃料消耗量最小为性能指标,采用该组合优化算法分别对两段轨道进行了4D全局优化. 数值仿真结果表明,该组合算法优化精度较高,收敛速度快,稳定性好,可用于机载计算机实时生成垂直软着陆4D轨道,同时还可推广到其他终端时间自由型两点边值问题.      

An in situ laser induced breakdown spectroscope (LIBS) for Chandrayaan-2 rover: Ablation kinetics and emissivity estimations

A miniaturized in situ laser induced breakdown spectroscope-LIBS is one of the two lunar rover payloads to be flown in India’s next lunar mission Chandrayaan-2, with an objective to carry-out a precise qualitative and quantitative elemental analyses of lunar regolith at the proximity of the landing region. As per the imposed mission constraints and the executed design optimization studies, a compact and light-weight LIBS prototype model is developed at our premises. This paper mainly concerns with the estimation of theoretical aspects; especially on evaluation of elemental ablation parameters and signal-to-noise ratio (SNR) calculations for the designed instrument. Theoretical estimations and simulations yielded an incident laser power density of the order of 5 × 10¹⁰ W/cm² on the target surface at a defined lens-to-surface distance (LTSD) of 200 mm and revealed an SNR > 100 for most of the elements under consideration. This paper also addresses the impact of LTSD variation on detection capability. The estimation of plasma-temperatures was carried out utilizing the emission spectra obtained under high vacuum environments employing the LIBS laboratory model. Experimental investigations and the performed theoretical estimations asserted the successful operation of the configured LIBS instrument for in situ elemental analyses on lunar surface.     

月地双基SAR成像的快速后向投影算法研究

为了提高对全球变化的大尺度观测能力,月基对地观测的概念应运而生。由于月基单站合成孔径雷达（Synthetic Aperture Radar,SAR）的双程传播会造成回波信噪比衰减,为此针对月地双基SAR工作模式展开研究。考虑到月、地、星复杂的相对运动为成像带来困难,根据月、地、星的几何关系建立平面直角回波模型。同时考虑到月球和卫星运行轨迹的弯曲性,分析模型误差,推导了精确的回波二维频谱表达式。在此基础上提出了一种基于快速后向投影的SAR成像算法,补偿了方位空变的模型误差,利用频域子孔径成像和频谱融合实现成像算法的快速化。通过分析可知月地双基SAR的理论回波信噪比与地月距离的平方成反比,与月基单站SAR相比有显著提升,最后通过X波段SAR仿真数据验证了文章算法处理月地双基SAR成像问题的效性。     

月球软着陆的高精度自主导航与控制方法研究

提出了一种月球软着陆的高精度自主导航和控制方法.根据测距波束视线相对月面的几何关系,确定本体系的月心方向,组合利用测速仪获取的本体系三维速度确定本体系相对轨道系的姿态、速度及高度,并根据确定的参数性质选取适应的制导和控制方法.算法基于直接测量数据确定和控制姿态、高度和速度,不受惯性导航误差的影响,可以有效地提高姿态、速度与高度的确定精度和控制精度.数学仿真表明了算法的有效性.     

最小数据丢失量的地月中继卫星任务调度研究

月球背面的探测器必须依靠地月中继卫星进行数据传输与通信。地月中继任务包括实时性任务和延迟容忍类任务,如数传任务。当探测器等待传输的数据量超出用户存储容量时,延迟容忍类任务会由于探测器本地存储资源不足和地月中继卫星天线资源受限而无法完成,导致任务数据丢失,所以需要设计一种合理的地月中继任务调度策略,提高地月中继卫星的资源利用率,减少数据的丢失。对地月中继卫星任务调度进行了研究,在分析地月中继卫星数传任务的特点及用户的存储限制的基础上,以最小化数据丢失量为优化目标,建立了地月中继卫星任务调度模型,并设计了一种基于离散烟花算法（DFWA）的地月中继卫星任务调度算法。仿真数据分析表明,基于离散烟花算法的地月中继卫星任务调度算法在求解结果上优于遗传算法,是一种合理、有效的调度方法。     

一种基于PCA的面向对象多尺度分割优化算法

多尺度分割是图像面向对象分类的基础,针对不同区域特征最优分割尺度确定的主观性以及采用聚类算法时聚类中心确定的随机性,提出了一种联合降维与聚类算法的面向对象多尺度分割优化算法。该算法首先利用主成分分析法（PCA）降维排序后的结果产生初始聚类中心;然后采用K-means聚类和度量每一个像素点合并的概率,从而得到适应不同研究区域内不同尺度地物的分割结果。采用多个影像数据库,通过引入聚类评价指标（内部评价指标和外部评价指标）、分割评价指标（分割精度、过分割率和欠分割率）并结合现有的图像分割方法及原始的K-means算法、与PCA降维后的K-means聚类对比分析。研究结果表明:经过降维处理后进行的聚类算法稳定性更高;与传统的聚类算法相比,结合PCA降维更能自动识别最优分割尺度;降维技术和聚类算法联合之中,目视和定量评价指标表明经过降维预处理后的聚类能得到更高质量的分割结果。      

基于阈值的激光雷达K均值聚类算法

针对同一距离不同目标的激光雷达全波形回波数据聚类准确率低的问题,在分析K均值聚类算法原理的基础上,提出了一种基于阈值的K均值聚类算法。首先,利用强度信息对距离信息进行标定,使用强度信息作为特征进行聚类以区分同距离的不同目标。然后,利用阈值限定聚类中心间的最小距离,提高聚类准确率。最后,搭建了扫描验证平台进行平移和旋转成像,对算法有效性进行验证。通过不同颜色目标和模拟道路回波数据聚类实验表明,在不同阈值的情况下,提出的基于阈值的K均值聚类算法的聚类准确率均在90%以上,相比于无阈值的K均值聚类算法准确率提升10%以上,能够有效进行目标聚类和模拟道路提取。      

Galactic cosmic ray abundances and spectra behind defined shielding.

LET spectra have been measured for lunar missions and for several near Earth orbits ranging from 28 degrees to 83 degrees inclination. In some of the experiments the flux of GCR was determined separately from contributions caused by interactions in the detector material. Results of these experiments are compared to model calculations. The general agreement justifies the use of the model to calculate GCR fluxes. The magnitude of variations caused by solar modulation, geomagnetic shielding, and shielding by matter determined from calculated LET spectra is generally in agreement with experimental data. However, more detailed investigations show that there are some weak points in modeling solar modulation and shielding by material. These points are discussed in more detail.