基于拼接缝自适应消除和全景图矫直的快速图像拼接算法

针对图像拼接存在色差过渡不均匀、图像倾斜扭曲及拼接效率低的现象，提出一种基于拼接缝自适应消除和全景图矫直的快速图像拼接算法。首先，用尺度不变特征变换（SIFT）提取图像指定区域特征点并用双向K近邻（KNN）算法进行图像配准，有效提高算法效率；其次，利用动态规划思想提出自适应公式找到最优拼接缝并用图像融合算法对其自适应消除，解决拼接缝色差过渡不均匀问题；最后，针对累积拼接误差形成全景图倾斜的现象，利用边缘检测算法提出自适应拟合四边形矫直模型，把原始全景图矫直为一个全新的全景图。所提算法与分块图像拼接和二叉树图像拼接算法相比，图像质量提升了5.84%~7.83%，拼接时间仅为原来的50%~70%。实验结果表明，该算法不仅通过自适应更新机制减少不同图像背景下拼接缝色差过渡不均匀的现象，从而提高了图像质量；而且提高了拼接效率，降低了全景图倾斜扭曲程度。     

基于CNN的人体姿态识别

姿态识别是人机交互中重要的研究课题之一，随着机器学习与神经网络的发展，研究的方式和成果趋于多样化，姿态识别的应用价值也日趋广泛。本文通过构建卷积神经网络模型，该模型共有11层，在对采样的数据集中5种人体姿态进行卷积与池化操作，最后进入全连接层进行分类，从而完成对数据集的训练和识别。结果显示，相较于机器学习方法，该模型的识别性能更加优秀，且免去了复杂的特征提取方式设计，让网络自身提取特征进行识别分类，效果更好。     

基于结构光的星箭对接环相对位姿测量方法

空间非合作目标位姿测量是空间在轨维护的前提。面向空间机械手对空间载体自动抓捕的应用需求，提出了基于结构光的空间非合作目标视觉测量方法。该方法以双套双线结构光测量装置作为测量传感器，以空间载体上普遍存在的星箭对接环作为抓捕目标。根据对接环的共同特点，选择直线特征和点特征相结合作为相对位姿的求解特征；基于直线特征求解圆环平面法向量，基于点特征求解圆环圆心坐标，进而得到机械手工具坐标系与对接环坐标系间的相对位姿关系；基于多重几何约束实现了不同光照条件下图像上目标直线的鲁棒识别。建立了演示验证实验系统，在大量演示实验的基础上，进行了实验结果分析。     

优化特征提取的互动式人脸活体检测研究

针对人脸识别系统中出现的通过照片或视频“欺诈”解锁问题，提出一种活体检测方法，通过随机指令判断被检测对象是否为真人。利用HOG特征和随机森林算法对摄像头采集的图像进行人脸检测，预测脸部68个特征点位置，把68个特征点位置和脸部位置的相对位置归一化后作为姿态的特征，提取的姿态特征与SVM分类器相结合，训练出分类效果较好的头部姿态估计分类器。通过多次实验，对特征提取方法进行优化，进一步提高检测准确率。最后，使用随机互动式命令序列对被检测人发出指令，实现活体检测。该方法对头部姿态估计具有较高的鲁棒性，并且可以有效地防范照片和视频认证攻击。     

Learning a Deep Predictive Coding Network for a Semi-Supervised 3D-Hand Pose Estimation

In this paper we present a CNN based approach for a real time 3D-hand pose estimation from the depth sequence. Prior discriminative approaches have achieved remarkable success but are facing two main challenges: Firstly, the methods are fully supervised hence require large numbers of annotated training data to extract the dynamic information from a hand representation. Secondly, unreliable hand detectors based on strong assumptions or a weak detector which often fail in several situations like complex environment and multiple hands. In contrast to these methods, this paper presents an approach that can be considered as semi-supervised by performing predictive coding of image sequences of hand poses in order to capture latent features underlying a given image without supervision. The hand is modelled using a novel latent tree dependency model (LDTM) which transforms internal joint location to an explicit representation. Then the modeled hand topology is integrated with the pose estimator using data dependent method to jointly learn latent variables of the posterior pose appearance and the pose configuration respectively. Finally, an unsupervised error term which is a part of the recurrent architecture ensures smooth estimations of the final pose. Experiments on three challenging public datasets, ICVL, MSRA, and NYU demonstrate the significant performance of the proposed method which is comparable or better than state-of-the-art approaches.      

Semiconductor Fe-doped SrTiO3-δ perovskite electrolyte for low-temperature solid oxide fuel cell (LT-SOFC) operating below 520 °C

High-temperature operation of solid oxide fuel cells causes several degradation and material issues. Lowering the operating temperature results in reduced fuel cell performance primarily due to the limited ionic conductivity of the electrolyte. Here we introduce the Fe-doped SrTiO_3-δ (SFT) pure perovskite material as an electrolyte, which shows good ionic conduction even at lower temperatures, but has low electronic conduction avoiding short-circuiting. Fuel cell fabricated using this electrolyte exhibits a maximum power density of 540 mW/cm² at 520 °C with Ni-NCAL electrodes. It was found that the Fe-doping into the SrTiO_3-δ facilitates the creation of oxygen vacancies enhancing ionic conductivity and transport of oxygen ions. Such high performance can be attributed to band-bending at the interface of electrolyte/electrode, which suppresses electron flow, but enhances ionic flow.     

纱条中纤维排列状态与纱条不匀的关系

为研究短纤维纱条中纤维的排列状态与纱条截面纤维根数分布不匀的关系,采用几何概率方法建立了等分区间内纤维左头端数分布与纱条截面纤维根数的期望之间的数学模型,并定义了表征纱条中纤维排列状态的参数。运用蒙特卡罗方法模拟了纱条截面根数不匀随排列参数变化的关系曲线。结果表明:短纤维纱条截面纤维根数的CV值与纱条中短纤维排列参数呈负线性关系,与纤维长度分布和模拟区间大小的划分无关;且当排列参数接近1时,即纤维左头端数在所划分区间中为固定值时,纱条截面纤维根数不匀最低,而当排列参数趋向于0时,即纤维左头端数在所划分区间中呈泊松分布时则不匀最大。     

基于深度学习的人体姿态估计方法综述

基于深度学习的人体姿态估计方法旨在通过构建合适的神经网络，直接从二维的图像特征中回归出人体姿态信息。主要按照2D人体姿态估计到3D人体姿态估计的顺序，并从单人检测与多人检测、稀疏的关节点检测与密集的模型构建等方面，对近年来基于深度学习的人体姿态估计方法进行系统介绍，从而初步了解如何通过深度学习的方法得到人体姿态的各个要素，包括肢体部件的相对朝向和比例尺度、骨骼关节点的位置坐标和连接关系，甚至更为复杂的人体蒙皮模型信息。最后,对当前研究面临的挑战以及未来的热点动向进行概述，清晰地呈现出该领域的发展脉络。     

Effect of TiB2 addition on grain orientation of porous Si3N4-TiB2 composites by magnetic field alignment technology

Textured porous Si₃N₄-TiB₂ composites were prepared by gel-casting–assisted magnetic field alignment technology and subsequent pressureless sintering in the N₂ atmosphere. The effects of TiB₂ content on grain orientation and properties of composites were studied. Results showed that using the magnetic field of 6T, Si₃N₄ grains exhibited a high a- or b-axis orientation and TiB₂ grains exhibited a high c-axis orientation. In textured green bodies, the increase in TiB₂ content from 0 to 20 wt%, the degree of texture of Si₃N₄ grains showed a slight decrease from 0.78 to 0.71. However, the degree of texture of TiB₂ grains was maintained at about 0.98 regardless of TiB₂ content. After sintering, except for the Si₃N₄ and TiB₂ phases, new TiN and BN phases also appeared in the samples. Additionally, the sintering process promotes grain orientation of Si₃N₄ and TiB₂. Herein, the degree of texture of TiB₂ reached to about 1 regardless of TiB₂ content, meaning that the complete grain orientation formed for TiB₂. Owing to the reconstruction reaction between TiB₂ and N₂, new TiN and BN phases did not form the orientation. The increase in TiB₂ contents gradually increased the apparent porosity of the textured samples from 27.9% to 39.4%, thereby leading to the decrease in bending strength from 126.8 ± 18.8 MPa to 60.3 ± 11.2 MPa and from 105.3 ± 19.8 MPa to 53.6 ± 13.0 MPa in two directions, respectively.