应用于多视点视频编码的基于深度特征的图像区域分割与区域视差估计

基于多视点视图深度特征,提出一种通过简单块匹配运算划分多视点视图区域并估计区域视差的算法.首先基于深度对象的概念确定图像中具有不同深度的区域数量以及这些区域对应的区域视差,再根据误差最小化准则初步确定每个图像块所属区域.当区域中图像块数量小于某个阈值时,采用区域合并算法将该区域中的每个图像块合并到与它的视差最为接近的其它图像区域,通过迭代形成最终的有效图像区域划分.实验表明,该算法能够以图像块为基本单元有效地划分各深度层区域,并准确估计对应的区域视差.     

TeV gamma rays from distant BL Lacs and photon–paraphoton kinetic mixing

An effect of a proposed Beyond the Standard Model particle physics process on the energy spectra of distant astrophysical objects is presented. It is suggested that TeV gamma rays from distant BL Lacs may kinetically mix with hypothetical hidden sector paraphotons. The latter can traverse vast distances of space because of only very weak interactions with intervening objects or material. These paraphotons would then reconvert back into photons that have identical characteristics as their originators, and are detected on earth. Laser based photon regeneration experiments test this model of hidden sector physics in the range of parameters that would impact the TeV gamma ray energy spectra arriving from distant BL Lacs.     

STUDY ON THE EQUILIBRIUM ATTITUDE AND STABILITY OF BUOYANT CYLINDERS

浮心相当于浮体的支承点,其位置随物体的姿态而变化;物体稳定平衡时重心与浮心的距离达到极小。圆柱浮体有2~5个平衡位置,其中相间的1~2个稳定,其姿态与比重和长径比相关,在系统参数变化时可能突变。     

面向真彩色三维显示的分层角谱算法和Gerchberg-Saxton算法研究

分层角谱算法可用于三维物体全息图的生成计算,在计算时间、占用内存和算法复杂度方面占有优势.本文基于强度图和深度图的分层角谱算法,对多波长采样进行分析,有效抑制混频现象,实现了真彩色物体的计算全息图生成.针对分层角谱算法重建时产生的散斑噪声问题,引入Gerchberg-Saxton(GS)算法,提出基于分层角谱算法的GS算法.通过计算比较引入GS算法前后再现像的均方根误差和峰值信噪比,证明该算法可有效地抑制散斑噪声,提高全息图的再现像质量,更适用于复杂形貌三维物体的计算.     

高能闪光照相中钨球散射规律研究

 用解析法和Monte Carlo法研究了钨球闪光X光照相的散射照射量空间分布随钨球半径、系统中元件构成以及保护器件与记录平面间距的变化规律。指出散射照射量的空间分布是不均匀的，它随球半径增加而变大；保护器件是散射照射量的主要来源，增加从保护器件到记录平面的距离，可以降低散射分布不均匀性，以及有效地减少散射。     

Resolution deterioration in emission electron microscopy due to object roughness

In the present paper a general analytic expression has been obtained and confirmed by a computer simulation which links the surface roughness of an object under study in an emission electron microscope and it's resolution. A quantitative derivation was made for the model case when there is a step on the object surface. It was shown that the resolution is deteriorated asymmetrically relative to the step. The effect sets a practical limit to the ultimate lateral resolution obtainable in an emission electron microscope.     

Continuous Viewpoint Planning in Conjunction with Dynamic Exploration for Active Object Recognition

Active object recognition (AOR) aims at collecting additional information to improve recognition performance by purposefully adjusting the viewpoint of an agent. How to determine the next best viewpoint of the agent, i.e., viewpoint planning (VP), is a research focus. Most existing VP methods perform viewpoint exploration in the discrete viewpoint space, which have to sample viewpoint space and may bring in significant quantization error. To address this challenge, a continuous VP approach for AOR based on reinforcement learning is proposed. Specifically, we use two separate neural networks to model the VP policy as a parameterized Gaussian distribution and resort the proximal policy optimization framework to learn the policy. Furthermore, an adaptive entropy regularization based dynamic exploration scheme is presented to automatically adjust the viewpoint exploration ability in the learning process. To the end, experimental results on the public dataset GERMS well demonstrate the superiority of our proposed VP method.     

NUMERICAL STUDY OF HIGH-REYNOLDS-NUMBER FLOW PAST A BLUFF OBJECT

A semi-explicit finite difference scheme is proposed to study unsteady two-dimensional, incompressible flow past a bluff object at high Reynolds number. The bluff object comes from a class of elliptical cylinders in which the aspect ratio and the angle of attack are two controlled parameters. Associated with the streamfunction–vorticity formulation, the interior vorticity, streamfunction and wall vorticity are updated in turn for each time step. The streamfunction and wall vorticity are solved by means of a multigrid method and a projection method respectively. In regard to the vorticity transport equation, implicitness is merely associated with the diffusion operator, which can be made semi-explicit via approximate factorization. Low-diffusive upwinding is devised to handle the convection part. Numerical results are reported for Reynolds numbers up to 40,000. Comparisons with other numerical or physical experiments are included.