A Self-calibrátion Method for Projector-Camera Modeling System

自标定方法可以得到相机系统的内部参数和/或外部参数,而不对场景结构作假设或作很少假设.本文提出一种新的自标定方法,用于投影相机建模系统.一个关键点在于投影仪既可以作为主动光源使用,也可以当做一个虚拟相机使用.首先,使用结构光方法从不同角度投影并拍摄一组图像.然后,通过将空间约束关系转换为内参约束,完成对系统的自标定.最后,利用估计的参数进行3D建模.实验结果验证了本文方法的可行性.     

基于自标定的两视图三维重构研究

相机标定是三维重构中的关键步骤,本文对在自标定下的两视图三维几何模型恢复进行研究。先采用SURF算法得到图像匹配点对,在此基础上,利用RANSAC鲁棒估计基础矩阵;然后根据Kruppa方程所推导出的二次方程自标定出相机内参数,进一步地,通过分解本质矩阵求解出相机外参数;最后,根据三角化法求得空间三维坐标值。实验结果证实,该方法具有效性和正确性,并且能够真实地再现物体的三维模型。     

二维平面靶标的数码相机标定

使用一种圆形靶标的数码相机外部参数标定算法,结合数码相机成像原理、坐标的旋转变换与平移变换,得到数码相机的标定参数。使用阈值分割法实现了靶标的边缘检测,通过确定像的坐标和相机间的距离,可以算出物平面上的点对应的像点的坐标。实验结果证明了这种方法的实用性。     

交通场景中的相机标定方法研究

现有交通场景下交通参数分析主要是在三维空间下实现的,要想获得交通场景中的三维空间信息不可或缺的一个环节就是相机标定。同时相机标定也是目前计算机视觉领域研究的重点和热点问题。本文基于相机成像模型和复杂的交通应用场景,研究了几种主要的相机标定方法,并总结了它们各自的特点。     

CT系统参数标定

CT系统安装时往往存在误差,从而影响成像质量,因此需要对安装好的CT系统进行参数标定,即借助于已知结构的样品(称为模板)标定CT系统的参数,并据此对未知结构的样品进行成像。本文提出一种基于最小二乘法的研究方法:先根据标定模版给定的信息,反演出CT系统旋转中心在正方形托盘中的位置、探测器单元之间的距离以及该CT系统使用的X射线的180个方向,从而完成CT系统的参数标定。     

基于视觉识别的机床工件定位方法

在数控机床加工中,对工件定位寻找加工零点,是加工过程中一个必不可少的环节,传统的工件定位方法一般是通过接触式对刀仪触碰工件的多个边,数控系统综合获取的数据,计算出工件的零点,对工件的过程比较耗时,手动对工件如果操作工操作不当,甚至会损坏对刀仪。本文提出一种基于视觉识别的工件定位方法,在相机标定后,通过对识别特征建模,将工业相机快速移动到定位区域,对定位区域拍照,并对获取的图像分析处理,识别出加工特征,计算出加工区域特征点相对机床零点的位置,并自动补偿至数控系统的零点偏置参数中。该定位方法只用机床移动一次,相对于传统自动测量方法的多次移动具有明显的效率优势,并且可以识别出对刀仪无法接触的特征,在多款数控机床上装载使用,定位精度都能够满足要求。     

基于约束神经网络的三维织物模拟算法

针对在进行复杂织物建模过程中,会遇到织物组织无序移动,造成建模效果不好等问题,提出了基于约束神经网络的三维织物运动建模方法。通过分析织物运作中特征点运动的对应关系,构造了一个可用于对织物形变进行准确估计的约束神经网络,添加可以对三维织物运动参数进行控制的约束关系,完成复杂无序运动下的三维织物计算机模拟。实验结果表明,本文的方法能够较好地模拟三维织物,取得了较好的效果。     

基于最小二乘法的圆盘类锻件视觉测量方法

介绍了一种圆盘类锻件视觉测量方法,详细介绍了其结构组成、工作原理、相机标定方法,图像处理方法以及使用最小二乘法来计算圆盘类锻件关键内外轮廓直径,采用Matlab语言进行程序编写,并完成整套测量系统,提高了圆盘类锻件的测量效率和测量精度。     

基于单目视觉的移动机器人室内定位

阐述移动机器人定位技术在机器人技术中的重要性,提出在单目视觉的条件下实现对于移动机器人的定位,并在实际室内环境运用单目相机、智能小车以及halcon软件平台进行试验,先通过相机标定得到相机参数,再运用相机在实验现场实际采集图像,利用相机参数和采集到的图像信息来定位智能小车,得出了小车实际坐标位置,证明本文提出的方法有效。     

基于参数物理差优化的三维地图重建算法研究

提出一种新的方法,基于二维坐标内的物理连续性,在相机的拍摄速度很高的时候,前后帧的位移变化很小,寻找一组不变量的基点,利用这些二维地域特征点和基点距离的物理连续性作为参数优化的最小值约束,使用非线性优化方法训练出非刚体运动时的运动结构参数,重建成三维地图区域模型.实验结果表明,该方法优于传统的三维地图重建方法.     

一种基于立方体的摄像机内参数标定方法

摄像机标定方法是计算机视觉的重要内容之一。提出了一种基于立方体的标定方法。该方法的模型为立方体,从3个或多个不同的方位摄取图像,由直径的正交关系获得3组正交灭点,从而线性的求出摄像机的5个内参数。模拟图像和真实图像实验表明,文章的自标定方法有较好的灵活性和较高的鲁棒性。     

Fluid dynamical analysis of the distribution of ink jet printed biomolecules in microarray substrates for genotyping applications

Oligonucleotide microarrays are tools used to analyze samples for the presence of specific DNA sequences. In the system as presented here, specific DNA sequences are first amplified by a polymerase chain reaction (PCR) during which process they are labeled with fluorophores. The amplicons are subsequently hybridized onto an oligonucleotide microarray, which in our case is a porous nylon membrane with microscopic spots. Each spot on the membrane contains oligonucleotides with a sequence complementary to part of one specific target sequence. The solution containing the amplicons flows by external agitation many times up and down through the porous substrate, thereby reducing the time delaying effect of diffusion. By excitation of the fluorophores the emitted pattern of fluorophores can be detected by a charge-coupled device camera. The recorded pattern is a characteristic of the composition of the sample. The oligonucleotide capture probes have been deposited on the substrate by using noncontact piezo ink jet printing, which is the focus of our study. The objective of this study is to understand the mechanisms that determine the distribution of the ink jet printed capture probes inside the membrane. The membrane is a porous medium: the droplets placed on the membrane penetrate in the microstructure of it. The three-dimensional (3D) distribution of the capture probes inside the membrane determines the distribution of the hybridized fluorescent PCR products inside the membrane and thus the emission of light when exposed to the light source. As the 3D distribution of the capture probes inside the membrane eventually determines the detection efficiency, this parameter can be controlled for optimization of the sensitivity of the assay. The main issues addressed here are how are the capture probes distributed inside the membrane and how does this distribution depend on the printing parameters. We will use two model systems to study the influences of different parameters: a single nozzle print head jetting large droplets at a low frequency and a multinozzle print head emitting small droplets at a high frequency. In particular, we have investigated the effects when we change from usage of the first system to the second system. Furthermore, we will go into detail how we can obtain smaller spot sizes in order to increase the spot density without having overlapping spots, leading eventually to lower manufacturing costs of microarrays. By controlling the main print parameters influencing the 3D distribution inside the porous medium, the overall batch-to-batch variations can possibly be reduced.     

基于光场渲染的动态3D目标重构技术

提出了一种基于光场渲染的动态3D目标重构的方法.目前基于图像建模的方法对于复杂场景难以建模,而基于图像渲染的方法因数据量大不利于动态场景的实时渲染.因此采用了模型和图像相结合的方法,从多视点视频图像中重建动态3D模型,采用光场映射算法对重建的3D模型进行光场采样,然后对样本分解压缩.实验结果表明,在保证真实感的同时,减少了数据量,并可实现动态3D目标的重构.     

基于正交小波变换的三维人脸门禁系统设计

提出一种结合嵌入式技术的三维人脸门襟系统的设计及具体实现方法。通过搭建相应硬件配准、运用正交小波技术,配合双目立体视觉原理,完成人脸的三维采集和特征识别。该系统因为加入了三维的因素,可以有效地克服传统的门襟系统受到光照、角度等外界因素影响带来的弊端。实验表明,该方法能够建立真实有效的三维人脸模型,提高了门禁系统识别的准确性。     

湖北省技术创新效率评价——基于研发资本存量角度

通过构建以研发资本存量作为创新投入变量建立省级区域技术创新指标体系,运用数据包络分析方法(DEA)对我国30个省级区域2004—2015年的技术创新效率进行测算与比较分析。研究结果表明,考虑研发资本存量后,可以显著考察到前期的研发资本投入对当期的技术创新活动的影响。又以湖北省为例进行技术创新效率分析,发现湖北省技术创新效率不高的主要原因为纯技术效率低、创新投入资源未充分利用和创新产出不足。     

A review on 3D printing functional brain model

Modern neuroscience increasingly relies on 3D models to study neural circuitry, nerve regeneration, and neural disease. Several different biofabrication approaches have been explored to create 3D neural tissue model structures. Among them, 3D bioprinting has shown to have great potential to emerge as a high-throughput/high precision biofabrication strategy that can address the growing need for 3D neural models. Here, we have reviewed the design principles for neural tissue engineering. The main challenge to adapt printing technologies for biofabrication of neural tissue models is the development of neural bioink, i.e., a biomaterial with printability and gelation properties and also suitable for neural tissue culture. This review shines light on a vast range of biomaterials as well as the fundamentals of 3D neural tissue printing. Also, advances in 3D bioprinting technologies are reviewed especially for bioprinted neural models. Finally, the techniques used to evaluate the fabricated 2D and 3D neural models are discussed and compared in terms of feasibility and functionality.     

Time-frequency channel modeling and estimation of multi-carrier spread spectrum communication systems

In wireless communications, the channel is typically modeled as a random, linear, time-varying system that spreads the transmitted signal in both time and frequency due to multi-path propagation and Doppler effects. Estimated channel parameters allow system designers to develop coherent receivers that increase the system performance. In this paper, we show how time-frequency analysis can be used to model and estimate the time-varying channel of a multi-carrier spread spectrum (MCSS) system using a complex quadratic sequence as the spreading code. We will show that for this spreading code, the effects of time delays and Doppler frequency shifts, caused by the mobility of environment objects, can be combined and represented effectively as time shifts. The discrete evolutionary transform (DET), as a time-frequency analysis method, enables us to estimate the effective time shifts via a spreading function and to use them to equalize the channel. Using the effective time shifts, the time-varying channel can be represented simply as linear-time invariant system by embedding the Doppler shifts that characterize the time-varying channel into effective time shifts. The channel parameters are used to estimate the data bit sent. To illustrate the performance of the proposed method we perform several simulations with different levels of channel noise, jammer interference, and Doppler frequency shifts.     

On the integration of model-based feature information in Product Lifecycle Management systems

As CAD models continue to become more critical information sources in the product’s lifecycle, it is necessary to develop efficient mechanisms to store, retrieve, and manage larger volumes of increasingly complex data. Because of their unique characteristics, 3D annotations can be used to embed design and manufacturing information directly into a CAD model, which makes models effective vehicles to describe aspects of the geometry or provide additional information that can be connected to a particular geometric element. However, access to this information is often limited, difficult, and even unavailable to external applications. As model complexity and volume of information continue to increase, new and more powerful methods to interrogate these annotations are needed.In this paper, we demonstrate how 3D annotations can be effectively structured and integrated into a Product Lifecycle Management (PLM) system to provide a cohesive view of product-related information in a design environment. We present a strategy to organize and manage annotation information which is stored internally in a CAD model, and make it fully available through the PLM. Our method involves a dual representation of 3D annotations with enhanced data structures that provides shared and easy access to the information. We describe the architecture of a system which includes a software component for the CAD environment and a module that integrates with the PLM server. We validate our approach through a software prototype that uses a parametric modeling application and two commercial PLM packages with distinct data models.