Alternative branching rules for some nonconvex problems

Following Linderoth (A simplicial branch-and-bound algorithm for solving quadratically constrained quadratic programs, Math. Program. 103 (2005), pp. 251–282), in this paper we show how a preliminary theoretical analysis about the quality of convex and concave envelopes over different regions may suggest branching rules alternative to the traditional ones based on boxes. For two nonconvex problems (minimization of the sum of products of affine functions and maximization of the sum of ratios of affine functions, in both cases over polytopes), we show how the indications of the theoretical analysis lead to computational results which confirm the superiority of the alternative branching rules with respect to those based on boxes.     

Adaptive Motion/Force Tracking Control for a Class of Mobile Manipulators

In this paper, modeling and adaptive motion/force tracking control is considered for a class of mobile manipulators under the holonomic and affine constraints with the presence of uncertainties and disturbances. Based on a suitable reduced dynamic model, adaptive controllers are proposed to ensure that the states of a closed‐loop system asymptotically track desired trajectories while the constraint force remains bounded by tuning design parameters. Detailed simulation results confirm the effectiveness of the control strategy.     

Topological entropy and data rate for practical stability: a scalar case

A special class of nonlinear systems, i.e. expanding piecewise affine discrete‐time scalar systems with limited data rate, is used to investigate the role of topological entropy and date rate in a practical stability problem. This special class of nonlinear system as an abstract model is an extension of discrete‐time unstable scalar systems, a well‐known model for quantized feedback design. For such systems with finite quantization levels, how to design a quantized feedback controller to achieve practical stability is considered as a boundability problem. Unlike the existing results about topological entropy for nonlinear stabilization and optimal control for linear systems, for the boundability problem under consideration, the feedback topological entropy defined in this paper is not equal to the minimum number of the quantization interval (i.e. the minimal information rate) and only provides a necessary condition for the boundability of our system under some condition. A necessary and sufficient condition for the boundability is also presented in terms of an inequality related to data rate and the minimum number of quantization intervals. Copyright © 2009 John Wiley and Sons Asia Pte Ltd and Chinese Automatic Control Society     

一种新的基于Zernike正交矩亚像素边缘定位的直径测量方法

提出了一种新的基于亚像素边缘定位技术的棒状物体直径测量方法;该方法利用两个CCD摄像头,分别获得两张棒状物体横截面的椭圆特征图像,利用仿射变换将灰度椭圆特征图像变换成圆形特征图像,然后利用亚像素中的Zernike正交矩对边缘进行精确定位,进而分别计算出棒状物体的直径,再将两个测量数据加以平均融合,完成对直径的测量;实验结果表明,与像素级方法相比较,该方法定位精度高,测量误差小,能够达到提高测量精度的目的。     

仿射子空间最近点分类算法

为了取得更好的识别效果,受支持向量机的几何解释和最近点问题启发,提出了一种新的模式分类算法——仿射子空间最近点算法。该算法是将支持向量机最近点法的最近点搜索区域由两类训练集凸包推广到两类训练样本各自张成的仿射子空间,并以仿射子空间作为样本分布的粗略估计,通过仿射子空间中的最近点对来构造平分仿射子空间间隔的最优分类超平面。该算法在ORL人脸识别数据库上进行的比较实验中取得了较好的识别效果,从而证实了该方法的可行性和有效性。     

基于仿射变换的图像置乱改进新算法

传统的仿射变换的图像置乱算法是将图像中像素的位置重新排列,从而使原始图像变换成一个杂乱无章的新图像,但其本身的像素值并未发生改变。所以置乱后图像的灰度直方图都不发生改变,而灰度直方图又是衡量一幅图像的重要特征。对于那些对安全性要求比较高的领域,如军事或牵涉到国家安全的机构,这类置乱方法的安全性可能达不到我们期望的要求。针对以上问题,该文提出了一种基于仿射变换的图像置乱改进新算法。该算法首先对原始图像进行仿射变换,然后对变换后的图像进行Zig-Zag扫描形成一个一维向量,再对一维向量中每个像素和它前面相邻的像素进行按位异或运算,同时对其异或运算的结果进行像素值的交叉换位,再次改变像素点的灰度值,从而提高图像置乱效果和安全性。实验结果表明,该算法与传统的图像置乱算法相比,置乱后的图像不仅直方图发生了显著改变,而且置乱图像的统计特征更像白噪声,更加适用于图像文件的加密传输和秘密图像信息在数字水印技术中的置乱预处理。     

Observer‐based controller design of discrete‐time piecewise affine systems

This paper presents a novel observer‐based controller design method for discrete‐time piecewise affine (PWA) systems. The basic idea is as follows: at first, a piecewise linear (without affine terms) state feedback controller and a PWA observer are designed separately, and then it is proved that the output feedback controller constructed by the resulting observer and state feedback controller gains can guarantee the stability of the closed‐loop system. During the controller design, the piecewise‐quadratic Lyapunov function technique is used. Moreover, the region information is taken into account to treat the affine terms, so the controller gains can be obtained by solving a set of linear matrix inequalities, which are numerically feasible with commercially available software. Three simulation examples are given finally to verify the proposed theoretical results. Copyright © 2010 John Wiley and Sons Asia Pte Ltd and Chinese Automatic Control Society     

基于主动漂移矫正的运动目标跟踪算法

针对传统的图像仿射配准(Affine image alignment, AIA)算法无法有效兼顾算法效率和鲁棒性的问题, 提出了一种快速鲁棒的新仿射配准算法---主动漂移矫正(Active drift correction, ADC)算法用于跟踪视频运动目标. 该算法的基本思想是: 通过引入一个漂移矫正项, 来改进传统算法的目标能量函数, 使算法具备抗漂移的能力, 从而提高算法的鲁棒性. 改进后的算法不需要传统算法中为增强鲁棒性而采用的许多复杂措施(如: 被动漂移矫正). 实验结果表明: 本文提出的算法简单、有效, 在不必牺牲算法效率和复杂度的条件下, 可以获得比传统算法更好的鲁棒性.     

仿射重投影混合虚实注册方法

针对特征图形结合自然纹理特征的虚实注册误差较大、运行速度慢的问题,提出仿射重投影混合虚实注册方法.通过建立摄像机、真实场景以及特征坐标系之间对应关系的仿射坐标系,且采用ORB算法进行特征检测及后续特征模板的构造,并利用仿射重投影矩阵来恢复特征模板在当前图像上的投影位置信息以计算虚实注册矩阵.采用OpenGL生成的立方体模型与真实场景之间进行虚实注册仿真实验.仿真结果表明:在特征图形被遮挡或不可见时,本文方法在精度、健壮性以及速率方面表现优异,提高了增强现实(AR)系统的实时性和鲁棒性.