Efficient image retrieval through vantage objects

We describe a new indexing structure for general image retrieval that relies solely on a distance function giving the similarity between two images. For each image object in the database, its distance to a set of m predetermined vantage objects is calculated; the m-vector of these distances specifies a point in the m-dimensional vantage space. The database objects that are similar (in terms of the distance function) to a given query object can be determined by means of an efficient nearest-neighbor search on these points. We demonstrate the viability of our approach through experimental results obtained with two image databases, one consisting of about 5200 raster images of stamps, the other containing about 72,000 hieroglyphic polylines.     

基于概念格因子分解的零件三维CAD模型检索

针对影响概念格应用的重要问题—即使是一个小规模数据集也会产生大量的形式概念,文中提出了可以满足关系覆盖的用对象（属性）概念分解形式背景对应的布尔矩阵的新方法.用这种方法原对象属性间的二元关系可以用数量在对象（属性）概念个数以内的概念表达出来,成为概念格因子.文中给出了概念格因子生成的基本原理及其算法.通过分析三维CAD零件模型功能表面间的关系构建零件工程图结构模型,并将其映射为形式背景,从而完成概念格因子到零件关键结构的应用.最后,实例演示了概念格因子在基于零件工程图结构模型的零件CAD模型检索中的运用.     

时空数据库移动点轨迹建模

时空数据库的移动点轨迹建模技术是近年来是时空数据库研究热点之一.本文在给定一系列时空数据库移动点的信息,找出实际应用中时间更新阈值和空间距离阈值,结合移动点时空模型和数学上线性回归的建模思想,建立一种新的支持过去、现在、未来的移动点轨迹模型,并有效地预测了移动点在某一时间阈值内,移动点的运动趋势及范围.     

基于谱残差和多分辨率分析的显著目标检测

根据人类视觉系统的特点,提出一种融合谱残差和多分辨率分析的显著目标检测方法。该方法通过在不同尺度上计算图像的亮度、颜色以及方向特征的谱残差,构建多分辨率显著性图谱序列,然后用线性插值方法将不同分辨率的特征显著图叠加得到3个特征显著图,再利用k均值聚类算法将每个特征显著图聚为两类,选择聚类中心距离最大的特征显著图作为最终的显著图,最后经过动态阈值处理获得图像的显著目标区域。基于自然图像的显著目标检测实验结果表明,该方法具有较强的稳定性和实用性,得到较为满意的检测结果。     

基于QP_TR信任域的序列图像目标跟踪

本文将信任域算法和尺度空间理论相结合,提出了一种能够精确描述目标尺寸连续变化的新的序列图像目标跟踪算法;将信任域算法与灰度模板相结合,提出了一种新的实时目标跟踪算法。在第一种算法中,首先将序列图像按照颜色直方图转换成目标概率分布图,目标区域在概率分布图中呈现为灰度块。然后通过检测该图在尺度空间中微分滤波器输出的极值,来决定这些灰度块的尺度。最后我们使用QP_TR信任域算法在尺度空间里和图像平面内快速搜索概率分布图的多尺度规范化Laplacian滤波函数极值,实现了目标定位并同时决定了其尺度,从而完成了跟踪任务。在第二种算法中,首先记录目标初始模板,在随后每一帧中应用OP_TR信任域算法搜索与该模板最相似的区域,实现目标定位。和现有算法的比较以及在大量真实序列图像上的实验表明,两种算法分别在目标大小描述,跟踪精度上以及运算速度上有了显著提高。     

On higher structures

In this paper, we discuss various philosophical aspects of the hyperstructure concept extending networks and higher categories. By this discussion, we hope to pave the way for applications and further developments of the mathematical theory of hyperstructures.     

移动对象无冗余周期模式发现算法

技术的发展进步和需求的多样性,产生了大量的数据,数据背后隐含模式的发现对问题的深入研究起到关键性的作用.为了挖掘移动对象的周期运动模式,提出了一种无冗余周期模式(Non-redundant period patterns,NRPP)发现算法.为解决噪声因素的影响,在现有方法基础上,引入相似性因子,删减冗余周期模式,使挖掘出的模式精简、准确.算法为不同模式设置不同阈值,不但解决了稀有项问题和组合爆炸问题,而且也使得挖掘效率更高效.采用公开移动对象数据实验,结果表明算法能高效挖掘出移动对象的周期模式.     

具有形变的平面轮廓匹配问题

提出了一种线性弹性模型用以解决具有形变的平面轮廓匹配问题.这种匹配方法克服了弹性匹配方法存在的计算复杂或内部相互制约力不强的缺陷,表现出非常强的鲁棒性.整个匹配过程是一个由粗至精的动态迭代过程.在所有弹簧结点上都进行外力计算.在外力的作用下,匹配过程能以很快的速度进行收敛.     

对象族拓扑约束求解的研究

针对当前对象族模型在求解拓扑约束时存在的缺陷,提出一种求解拓扑约束的新方法,这种方法在求解拓扑约束时,把拓扑约束映射为布尔约束满足问题,通过用SAT求解器求解布尔约束来求解拓扑约束。实践证明,该方法不仅直接关联与拓扑约束指定的特征的语义,而且当模型中存在大量相交的特征时也是可行的,提高了拓扑约束求解的效率。     

Blob Metamorphosis based on Minkowski Sums

This paper addresses the metamorphosis of soft objects built from skeletons. We propose a new approach that may be split into three steps. The first step consists in an original splitting of the initial and the final shapes with a view to creating a bijective graph of correspondence. In the second step, we assume that the skeletons are convex polygonal shapes, and thus take advantage of the properties of Minkowski sums to characterize the skeletons of intermediate shapes. Eventually, we characterize the intermediate distance and field functions; we describe a set of interpolation methods and propose to use a restricted class of parametrized distance and field functions so as to preserve coherence and speed-up computations. We show that we can easily extend those results to achieve a Bézier like metamorphosis where control points are replaced by control soft objects; in this scope, we have adapted existing accelerated techniques that build a Bézier transformation from a set of convex polyhedra to any kind of convex polygonal shapes. Eventually, we point out that matching all components of the initial and the final shapes generates amorphous intermediate shapes based on an overwhelming number of intermediate sub-components. Thus, we propose heuristics with a view to preserving coherence during the transformation and accelerating computations. We have implemented and tested our techniques in an experimental ray-tracer.