复杂室内图像的灭点检测与箱体重建方法

近年来,随着家居虚拟展示应用的推广,针对图像的室内场景建模技术成为研究和应用的热点。在图像理解的基础上,针对单张复杂室内图像提出了一套箱体建模方法。首先进行代表房间主轴方向的灭点检测,并通过分析影响灭点检测的因素,对灭点检测算法提出相应的改进,从而显著改善算法的执行效率和准确性。然后对图像的场景布局信息进行自动还原,并通过加入布局优化步骤,提高布局还原结果的准确率。最后利用灭点得到相机的内、外部参数,并以此为基础实现图像到三维模型的转换。实验表明,单张复杂室内图像的箱体重建方法能够快速地对图像场景进行分析,并恢复场景的箱体布局,满足虚拟重建的需要。     

M2Tracker: A Multi-View Approach to Segmenting and Tracking People in a Cluttered Scene

When occlusion is minimal, a single camera is generally sufficient to detect and track objects. However, when the density of objects is high, the resulting occlusion and lack of visibility suggests the use of multiple cameras and collaboration between them so that an object is detected using information available from all the cameras in the scene.In this paper, we present a system that is capable of segmenting, detecting and tracking multiple people in a cluttered scene using multiple synchronized surveillance cameras located far from each other. The system is fully automatic, and takes decisions about object detection and tracking using evidence collected from many pairs of cameras. Innovations that help us tackle the problem include a region-based stereo algorithm capable of finding 3D points inside an object knowing only the projections of the object (as a whole) in two views, a segmentation algorithm using bayesian classification and the use of occlusion analysis to combine evidence from different camera pairs.The system has been tested using different densities of people in the scene. This helps us determine the number of cameras required for a particular density of people. Experiments have also been conducted to verify and quantify the efficacy of the occlusion analysis scheme.     

复杂景物中人面的定位

人面是常见的复杂模式.在复杂景物图片中自动找出人面是一个困难但有重要意义的课题.这是自动人面识别的第一个重要步骤.该文提出了一个在复杂背景中找出不同尺寸人面的方法,能在较大的尺寸范围内在复杂背景的黑白图片中找到人面.     

从视频中检测人脸

视频中人脸检测的应用领域广泛，近来受到了极大关注。文中提出一种在MPEG流中检测人脸的新方法，它可以从复杂背景中有效地检测方向、大小不同的人脸，还可以处理多个人脸交叠的情况，为适应视频检索的需要，该算法依据帧间冗余性，自适应地调整肤色检测器，利用MPEG流中的运动矢量在一个GOP内跟踪人脸，依据场景的变化更新分割码本等措施，有效地提高了计算速度。用算法测试多个视频序列，实验结果令人满意。     

复杂背景下人眼自动定位

针对复杂背景下静态人脸灰度图像，提出一种人眼自动定位方法。首先基于人脸几何特征先验知识，建立人眼位置的判定准则，对分割阈值范围进行粗估计；然后采用分割阈值递增法，并结合人眼位置判定准则，判定分割图像中双眼黑块是否出现；最后利用二维相关系数作为对称相似性测度，检验检测出的双眼黑块的真实性。     

复杂背景下人眼的快速定位

采用基于改进Adaboost算法的级联式人脸和人眼分类器检测人脸和眼睛的候选位置,再用支持向量机（SVM）分类器验证并确定人眼的最佳位置;实现了在复杂背景图像中快速、准确的眼睛定位.与传统的Adaboost算法相比,改进的Adaboost算法构建分类器所需的特征数目大大减少,提高了眼睛定位速度;同时利用人脸几何特征的SVM分类器验证,提高了定位精度.实验结果表明该算法具有很好的精确性和实时性.     

一种有效的人脸识别方法

人脸识别是模式识别领域中一个相当困难而又有重要理论价值的研究课题.本文首先引入表示人脸的特征图像和投影图像的概念,提出利用投影图像在特征图像上投影的坐标作为描述人像本质属性的特征矢量,并导出了基于投影图像的人脸特征抽取方法.最后构造了一个层次的距离分类器进行人脸的识别.实验结果表明基于投影图像的特征矢量具有很好的稳定性、鉴别能力和识别率.     

大容量人脸识别门禁系统的研究

人脸识别技术是当今的热门技术,将人脸识别技术应用于安防领域是未来的发展方向.但是,在很短的时间内,对成千上万人的人脸识别如何实现?强背光影响人脸图像如何解决?从系统的角度出发,给出了大容量人脸识别门禁系统的设计,综合运用了远程射频识别(RFID)技术,重构人脸识别标准数据库,缩小了人脸识别的范围,提高了识别速度和识别率,解决了大容量的人脸识别问题;采用补偿光的办法,解决强背光对人脸图像影响.为人脸识别门禁系统的推广应用提供了借鉴.     

人脸的民族特征抽取及其识别

人脸的民族特征是人脸信息描述的重要特征之一.首先构建了中国多民族人脸数据库,利用人脸识别技术提取民族面部特征和民族识别.在特征抽取方法中,采集人脸中的代数特征和几何特征,采用LDA(线性判别分析)算法提取人脸图像的代数特征.还构建了能够抽取人脸几何特征的弹性模板,并利用Gabor小波进行特征点定位.实验采用KNN和C5.0分别学习训练集中的代数特征和几何特征,并对测试集进行预测分类.实验结果表明,利用代数特征方法和几何特征方法对藏族、维吾尔族、壮族3个民族的平均识别准确率分别达到79%和90.95%.     

真实感人脸的形状与表情空间

提出一种建立三维人脸扫描模型参数空间的算法,其中的模板拟和算法基于能量最小的优化机制,通过非线性优化过程求解模板模型在每个顶点上的位移矢量,使之逼近目标模型.优化目标方程由以下测度组成：距离、平滑度以及人脸特征对应,如特征曲线、边界和特征点对等.使用该算法可用于不同人脸以及不同表情模型之间的对应建立,从而获取一致参数化的人脸形状和表情空间.在文中系统中,三维面部特征曲线通过Canny边检测算法自动获取,这样自动检测获取的特征曲线可用于降低三维形状描述的维数,同时完整的面部几何形状通过径向基函数插值得到.在中性人脸和表情人脸模型上所作的一致参数化为许多应用提供了平台,如形状渐变,纹理迁移和表情迁移.考虑到自动提取的特征曲线和二维线画卡通人脸的相似性,使用迭代优化算法实现二维线画卡通人脸姿态到三维真实人脸模型的迁移.     

面向虚拟交互展示的人脸合成方法

人脸合成由于会受到光线、环境背景、头的姿态等的影响,在实时互动合成时常常难以得到满意的结果。针对虚拟交互展示中人脸互动合成的需求,本文基于快速样本学习的人脸检测,采用基于人脸细化参数化模板的快速特征点定位,最后实现人脸快速合成算法,实时生成用于科技馆展示的不同参观者的完美脸型效果。     

人脸正面图像中眼睛的精确定位

提出了人脸图像中眼睛定位方法,首先对灰度图像进行灰度形态学操作,所得谷图中像素数总体上随灰度级的增加而显著减少;用位于高端的灰度值作为阈值,逐次对谷图进行二值化,从所得二值图像中获得眼睛的候选位置;用PCA方法进行眼睛位置的最佳选择．该方法解决了二值化阈值步长的自适应选取问题．用AR,Yale和自建人脸数据库进行实验,得到眼睛定位成功率大于93％,平均定位偏差小于眼球半径的1／3．     

多回波环境中多机动目标跟踪的新算法*

目标的状态估计与数据关联是机动多目标跟踪中的关键问题.针对杂波环境中多机动目标的跟踪问题,本文首先引入一种自适应滤波算法,并与快速概率数据关联算法结合,提出一种适于实际应用的密集回波环境下机动多目标跟踪的新算法-快速自适应概率数据关联(FAPDA)算法,利用近似概率数据关联(PDA)算法的计算量达到优于联合概率数据关联(JPDA)算法的跟踪效果,并能快速检测到机动.通过与JPDA算法的仿真结果进行对比,表明了该算法的有效性和快速性.     

Task Oriented Optimum Positioning of a Mobile Manipulator Base in a Cluttered Environment

A method for generating discrete optimal sequences of base locations for mobile manipulators is presented that considers the task capability of the workspace in a cluttered environment. In implementation, the obstacles and task trajectories are represented by 2ⁿ trees, so that a series of set operations are performed to characterize the manipulators configuration space into topological subspaces. By incorporating trajectory-motion-capable subspaces into the enumeration of the cost function, an optimum search technique is made applicable to the determination of a task feasible location. The method is then extended to a multiple positioning problem by concatenating the single optimization processes into a serial multistage decision making system, for which an optimal set of decisions can be found through a computationally efficient dynamic programming process.The computational paradigm of the present method is coherent with topological workspace analysis, and thus applicable to task trajectories of arbitrary dimensions and shapes. The effectiveness of the presented method is demonstrated through simulation studies performed for a 3-d.o.f. regional manipulator operating under various task conditions.     

基于差别特征的神经网络人脸识别

根据视觉识别的差别特征分辨特性,该文对自联想神经网络进行了改进,提出了基于差别特征的识别方法。文中采用ＯＲＬ人脸图象库进行的对比识别实验表明,改进后的差别特征神经网络对原人脸图象和加斯噪声的人脸图象,都较自联想神经网络识别高,证实了差别特征的有效性。     

Locating a tree in a phylogenetic network

Phylogenetic trees and networks are leaf-labelled graphs that are used to describe evolutionary histories of species. The Tree Containment problem asks whether a given phylogenetic tree is embedded in a given phylogenetic network. Given a phylogenetic network and a cluster of species, the Cluster Containment problem asks whether the given cluster is a cluster of some phylogenetic tree embedded in the network. Both problems are known to be NP-complete in general. In this article, we consider the restriction of these problems to several well-studied classes of phylogenetic networks. We show that Tree Containment is polynomial-time solvable for normal networks, for binary tree-child networks, and for level-k networks. On the other hand, we show that, even for tree-sibling, time-consistent, regular networks, both Tree Containment and Cluster Containment remain NP-complete.     

Makeup Lamps: Live Augmentation of Human Faces via Projection

We propose the first system for live dynamic augmentation of human faces. Using projector‐based illumination, we alter the appearance of human performers during novel performances. The key challenge of live augmentation is latency — an image is generated according to a specific pose, but is displayed on a different facial configuration by the time it is projected. Therefore, our system aims at reducing latency during every step of the process, from capture, through processing, to projection. Using infrared illumination, an optically and computationally aligned high‐speed camera detects facial orientation as well as expression. The estimated expression blendshapes are mapped onto a lower dimensional space, and the facial motion and non‐rigid deformation are estimated, smoothed and predicted through adaptive Kalman filtering. Finally, the desired appearance is generated interpolating precomputed offset textures according to time, global position, and expression. We have evaluated our system through an optimized CPU and GPU prototype, and demonstrated successful low latency augmentation for different performers and performances with varying facial play and motion speed. In contrast to existing methods, the presented system is the first method which fully supports dynamic facial projection mapping without the requirement of any physical tracking markers and incorporates facial expressions.     

Development and Evaluation of a Chase View for UAV Operations in Cluttered Environments

Civilian applications for UAVs will bring these vehicles into low flying areas cluttered with obstacles such as building, trees, power lines, and more importantly civilians. The high accident rate of UAVs means that civilian use will come at a huge risk unless we design systems and protocols that can prevent UAV accidents, better train operators and augment pilot performance. This paper presents two methods for generating a chase view to the pilot for UAV operations in cluttered environments. The chase view gives the operator a virtual view from behind the UAV during flight. This is done by generating a virtual representation of the vehicle and surrounding environment while integrating it with the real-time onboard camera images. Method I presents a real-time mapping approach toward generating the surrounding environment and Method II uses a prior model of the operating environment. Experimental results are presented from tests where subjects flew in a H0 scale environment using a 6 DOF gantry system. Results showed that the chase view improved UAV operator performance over using the traditional onboard camera view.