基于两视图的曲面物体自动建模方法

提出一种基于两视图的曲面物体自动重建方法. 采用基于Haudorff距离的特征匹配技术对图像的正面和侧面中的曲面物体进行匹配,自动获得图像中物体的大致形状及位置信息. 采用基于形状先验的图像分割算法自动地将曲面物体从图像的正面和侧面中分割出来. 利用这两幅图像中获得的侧影轮廓线,实现曲面自动重建和纹理自动提取. 通过真实曲面物体进行重建实验,验证了该方法的可行性.     

核废物桶放射性探测的层析γ扫描技术

介绍了用于非均匀放射性核废物无损测量层析γ扫描技术的原理和方法,并对标准200 L核废物钢桶样品进行了实际测量,采用152Eu作为透射源而实现了层析γ扫描重建的全过程,采用极坐标的体素划分形式和MCNP程序对系统进行效率刻度,分别采用代数迭代法和最大似然重建算法进行了透射和发射重建,得到了线性衰减系数与活度的分布图,并与分段γ扫描的测量结果加以对比.结果表明,层析γ扫描技术的重建结果明显好于分段γ扫描,其与真实值的偏差在10%以内,从而验证了极坐标划分和重建算法的可行性.     

小波变换应用于信号去噪研究

分别对小波分解与重构法去噪法、非线性小波变换阈值法去噪法以及小波变换模极大值去噪法的原理进行了阐述，并比较了这几种小波去噪方法的优缺点，通过计算机仿真结果表明有效可行。     

Millennial temperature reconstruction based on tree-ring widths of Qilian juniper from Wulan, Qinghai Province, China

The climate of the past 1000 years is an important context for evaluating the recent climate warming. However, there are few 1000-year-long climate reconstructions with annual resolution in the Qinghal-Tibet Plateau. In this paper, a dendroclimatic analysis was conducted for the radial growth of Qilian juniper from the upper forest limit in Wulan, Qinghai Province. The results of correlation analysis between the tree-ring widths and the climate variables indicate that the growth of junipers at the upper forest limit is mainly limited by low temperatures of September, November and February of the pregrowth season, and July of the current growth season. There is no significant correlation between the tree-ring widths and precipitation. A mean temperature from the previous year＇s September to the current year＇s April was reconstructed for the Wulan area since A.D. 1000. The reconstruction can explain 40.8% of the instrumental variance in the calibration period （1856-2002）. The reconstruction shows that the 20th century is the warmest 100 years, and the 1990s is the warmest decade during the past 1000 years, while the coldest 100 years and decade occur at 1600-1699 and 1642-1651, respectively. The variations are verified well by the temperature reconstruction of the middle Qilian Mountain （QL） and the total organic carbon （TOC） in the Qinghai Lake sediments. The comparison of our reconstruction with the annual temperature reconstruction of extra-tropical Northern Hemisphere suggests that the climate of Wulan during the Medieval Warm Period is of obvious regional specialty, but there was a good response to the climate of hemispheric scales during the recent 400 years.     

基于双CCD立体视觉系统的建模

介绍了一个由双CCD构成的立体视觉系统,对其工作过程和性能进行了分析,因为其性能不能满足要求,所以提出了一种通过两幅二维图像对应点的图像坐标获取其三维坐标的算法．该算法引入了摄像机的内外参数,包括摄像机的位置、取向以及镜头的畸变等,并利用Matlab进行了实现,结果表明该算法计算量小,而且能获得比较好的精度．     

美日欧重构稀土供应链战略对中国稀土产业的影响

 梳理了美日欧重构稀土供应链的行动,分析了重构的动因和可能性,识别了重构对中国稀土产业发展的影响,分析表明：（1）目前,通过重构稀土供应链,美日欧已初步建成了独立于中国的稀土资源供应链体系;（2）重构的理论依据在于保持美日欧在稀土高端应用领域的比较优势,现实依据在于保障美日欧等国稀土产业供应安全,因此重构稀土供应链是美日欧的一项长期战略;（3）在重构稀土供应链的影响下,美日欧陆续恢复对稀土矿的开采,稀土供给格局发生较大变化并逐渐走向多元化,中国稀土资源供应优势地位正逐渐降低;（4）根据稀土永磁材料的产量和新能源单车所耗钕铁硼磁材2种方式,预测到2030年稀土需求将达60多万t,而供给仅为35万t,远不及稀土的需求量,且随着时间的推移,该缺口会越来越大。因此建议：（1）充分利用国内国外2种资源、2个市场,建立多元化稳定的稀土供应格局;加大稀土供应满足市场需求,稳定稀土原材料价格,提高资源掌控力和资源保障力;（2）延伸产业链,塑造国际竞争新优势,规划稀土重点应用领域,加大稀土高端材料及其应用领域的研发力度,实现技术突破,破解稀土产业“低端锁定”困局;（3）加强稀土产业链、政策链、创新链的深度融合,提升中国稀土全产业链竞争力。     

医学CT三维重建

介绍了医学CT三维重建的过程 ,分析了目前所用的关键技术 ,并探讨了今后的发展趋势 ,为医学图像三维重建应用软件的开发打下了理论基础 .     

Blind reconstruction of linear scrambler

An algorithm based on eigenanalysis technique and Walsh-Hadamard transform （WriT） is proposed. The algorithm contains two steps. Firstly, the received sequence is divided into temporal windows, and a covariance matrix is computed. The linear feedback shift register （LFSR） sequence is reconstructed from the first eigenvector of this matrix. Secondly, equations according to the recovered LFSR sequence are constructed, and the Walsh spectrum corresponding to the equations is computed. The feedback polynomial of LFSR is estimated from the Walsh spectrum. The validity of the algorithm is verified by the simulation result. Finally, case studies are presented to illustrate the performance of the blind reconstruction method.     

基于深度学习的脑微观结构重建

为了研究脑运行机制以实现类脑智能,我们引入深度学习工具系统解决突触级脑微观重建中大数据自动分析的难题,包括:密集神经元重建、单根神经元追踪、关键细胞器检测和重建.其中我们使用带有候选区域(region proposal network,RPN)的全卷积网络(fully convolutional networks,FCN)检测线粒体和突触,结合SPPUnet(spatial pyramid pooling U-net)网络框架和MultiCut算法进行神经元的重建,各部分在量化分析和视觉上取得了较好的结果.重建工作为学术界开展高通量的突触尺度脑微观结构重建提供有效支持.     

脑微观重建系统研究

本文旨在研究如何利用硬件,软件,算法的系统工程的方案解决脑微观结构重建的高通量自动化数据分析的难题.通过实现样片制备,自动切片,显微成像,三维重建以及软件平台等分阶段工作,建立符合神经结构生物学特征的模式识别和机器学习分类方法,解决海量畸变图像拼接配准,密集神经结构识别,歧义神经联结,多源数据融合等限制分析效率的关键问题,为搭建大体量神经结构重建工程平台提供理论基础和解决方案,满足脑科学研究对高通量神经回路网络重建的需求.