基于CPU+GPU的大视场物镜成像畸变实时校正

针对工业大视场物镜畸变成像的实时校正问题,提出一种校正算法和CPU+GPU并行加速方案.根据光学畸变理论和相机标定技术,建立非球面畸变校正模型.利用棋盘样板计算光学中心和估计畸变系数,设计校正算法.在CPU+GPU并行加速方案基础上,设计内核自适应维度算法并优化运行程序,结合OPENGL驱动进行实时校正和显示.实验结果表明,本文设计的实时校正系统对高分辨率的畸变成像校正率可以达到98.2%,单帧耗时0.026 s,平均综合加速比为29.1.该系统精度高,可移植性强,简单易行,能够广泛应用于成像畸变的实时校正.     

探究服从二项分布的随机变量概率最大值问题

二项分布是应用最广泛的离散型随机变量概率模型．对与二项分布有关的一些问题的探究是很有意义的．同学们会提出这样的问题：服从二项分布的随机变量取何值时概率最大？下面我们来探究一下．     

改善局部难冷区域的一种导叶层板端壁

受二次流的影响,气膜冷却端壁表面存在大量难冷区域。紧凑型层板结构可以实现端壁的双侧冷却,从而缩小难冷面积。本文通过改变气膜孔布局,设计了一种可以有效改善局部难冷区域的层板冷却端壁。在主冷流温比1.5下,测量了端壁表面的温降特性和综合冷却效率分布,通过与均匀气膜布局的层板端壁相比较,发现:采用改进后的层板结构,可以显著降低端壁表面温度,缩小难冷区域面积,获得更均匀的冷却效果。且压力侧根部区域与吸力面前缘区域冷却效果增幅尤甚。另外,此层板的改进优势随着质量流量比的增加变得更明显。     

阴性及中性丝氨酸与纯的金属掺杂石墨烯相互作用的密度泛函理论研究

We present a theoretical study of interactions of anionic and neutral serine (Ser) on pure or metal-doped graphene surfaces using density functional theory calculations. Interactions of both types of Ser with the pure graphene surface show weak non-covalent interactions due to the formation of -COOH…π, -COO^-…π, and -OH…π interactions. On metaldoped graphene, covalent interactions to the surface dominate, due to the formation of strong metal-O and O-metal-O interactions. Furthermore, the doped Fe, Cr, Mn, Al, or Ti enhances the ability of graphene to attract both types of Ser by a combination of the adsorption energy, the density of states, the Mulliken atomic charges, and differences of electron density. At the same time, the interaction strengths of anionic Ser on various graphene surfaces are stronger than those of neutral Ser. These results provide useful insights for the rational design and development of graphene-based sensors for the two forms of Ser by introducing appropriate doped atoms. Ti and Fe are suggested to be the best choices among all doped atoms for the anionic Ser and neutral Ser, respectively.