CN101894274A - Complex sequence graphics matching system based on topological relation and realization method thereof - Google Patents

Complex sequence graphics matching system based on topological relation and realization method thereof Download PDF

Info

Publication number
CN101894274A
CN101894274A CN 201010193054 CN201010193054A CN101894274A CN 101894274 A CN101894274 A CN 101894274A CN 201010193054 CN201010193054 CN 201010193054 CN 201010193054 A CN201010193054 A CN 201010193054A CN 101894274 A CN101894274 A CN 101894274A
Authority
CN
China
Prior art keywords
set
matching
pattern
polygon
flow
Prior art date
Application number
CN 201010193054
Other languages
Chinese (zh)
Inventor
张宝印
沈学如
Original Assignee
江苏如意通动漫产业有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 江苏如意通动漫产业有限公司 filed Critical 江苏如意通动漫产业有限公司
Priority to CN 201010193054 priority Critical patent/CN101894274A/en
Publication of CN101894274A publication Critical patent/CN101894274A/en

Links

Abstract

The invention provides a complex sequence graphics matching system based on a topological relation, which is characterized in that the system comprises an identification and extraction module, a directed module, a matching network, a maximum flow rate solving module, a relevance matching module and an integral matching output module. The invention resolves a polygon registration problem into a matching problem of elements in two sets. The essence of the problem lies in solving maximum matching problem and also solves the maximum matching of the polygon registration problem on the premise of meeting the topological relation. The invention resolves the difficult problem of establishment of internal relations for complex sequence graphics and conducts thoroughly technical transformation aiming at the existing manufacturing method of two dimension animations, namely, manual labor is replaced by a computer, which is a fundamental technical breakthrough for improving production efficiency of animation products.

Description

基于拓扑关系的复杂序列图形匹配系统及其实现方法 Implementation based on a complex sequence pattern matching system topology and

技术领域 FIELD

[0001] 本发明涉及数字图形图像处理以及数字娱乐领域,尤其是一种基于拓扑关系的复杂序列图形匹配方法。 [0001] The present invention relates to digital image processing, and graphic digital entertainment, particularly a complex sequence pattern matching method based on topology.

背景技术 Background technique

[0002] 目前,序列图形图像的匹配技术是建立图形之间内在关系的关键技术,对于复杂图形还涉及分层技术。 [0002] At present, the sequence of graphic images matching technology is a key technology to establish the intrinsic relationship between the graphics for complex graphics also relates layering technology. 这些技术是实现目标自动匹配、拓扑关系自动建立、自动高质量上色以及计算机自动内插的关键。 These technologies are automatically matched to achieve the goal, topology automatically established, high-quality color and automatic key computer automatically interpolated. 也是实现二维动画制作自动化、智能化的核心技术。 But also to achieve the two-dimensional animation production automation, intelligent core technology.

[0003] 序列图形图像配准问题的研究是数字视频、模式识别、计算机视觉以及动画自动处理的一个主要内容,与静止图像处理不同的是:序列图形图像包含很多帧间瞬时的相关因素(冗余度),当然也可以把序列图形图像当作一个一个静态图像序列进行处理,其间处理是相对独立的,但是,借助于多帧处理技术所产生的冗余度就可以使得我们去开发更有效的算法,大多数情况下,相对独立的处理序列图形图像的方法,根本无法满足序列图形图像所要求的海量数据高速运转等要求,也不能满足二维动画向自动化、智能化发展的需要。 [0003] Sequence graphic image registration problem is that digital video, one of the main contents of pattern recognition, computer vision and automatic processing of animation, and still image processing is different: graphic image sequence contains a lot of inter-related factors instantaneous (redundant redundancy), of course, also possible to sequence a graphic image as a still image sequence processing, during processing are relatively independent, but, by means of redundancy generated by processing a plurality of frames can be enabled us to develop more effective the algorithm, in most cases, relatively independent of the method for processing a sequence of graphic images, simply can not meet the required sequence of graphic images of huge amounts of data at high speed, etc., can not meet the needs of two-dimensional animation to automation, intelligent development.

[0004] 对于序列图形图像在动画领域的应用,主要分为两大类:一是基于栅格数据结构的序列图像配准;另一个是基于矢量的序列图形匹配。 [0004] For applications in the field of animation sequences of graphic images, divided into two categories: one is based on image registration sequence of raster data structure; the other is based on the sequence match pattern vector. 对于序列图像配准,主要方法如光流分析法、相位相关法、块匹配法、像素逆归法、贝斯法等;对于序列图形匹配,主要方法如轮廓匹配、质心匹配以及模式匹配等。 For the sequence of image registration, the main method of optical flow analysis method, phase correlation method, a block matching method, the pixels inverse normalization, Bayesian method and the like; for the sequence of pattern matching, the primary method such contour matching, centroid matching and pattern matching.

[0005] 对于基于栅格数据结构的序列图像配准,在动画方面的应用目前主要集中在图像变形(Image Morphing)领域,典型的如加拿大Toon Boom公司的二维动画制作系统,就是采用这个技术进行中间帧的自动内插。 [0005] For the sequence of image registration based raster data structures, application in animation is currently focused on the image distortion (Image Morphing) areas, such as the typical Canadian company Toon Boom 2D animation production system, is the use of this technology the intermediate frame is automatically inserted. 另外,Flash网络动画制作系统中也应用了这个技术。 In addition, Flash animation network systems also use this technology.

[0006] 对于基于矢量的序列图形匹配,在动画方面的应用还处在研究初期,主要研究集中在简单多边形的相似性判断上,目前这类研究开展比较广泛和深入,研究成果颇多。 [0006] For sequences based on pattern matching vector used in the animation still in the early research, mainly research has focused on simple polygon similarity judgment, this type of research currently carried out more extensive and in-depth, research a lot.

[0007] 利用两个图形的轮廓形状进行相似性判断也是序列图形配准的一个重要手段,判断两个轮廓形状是否相似的方法很多,对特殊形状,如多边形一般可用计算角或边的比例的方法;对于规则的曲线可用参数判决法进行判断。 [0007] With the pattern shape of the two profile is an important means of sequence similarity determination pattern is registered, determines whether the contour shape of two similar methods are many special shapes, such as polygonal ratio generally available computing angles or edges method; versus rule judgment method determines the available parameters.

[0008] 在轮廓匹配方面,主要分为两大类:一类是基于线的匹配算法,即以多边形的轮廓为主要特征的匹配。 [0008] In the contour matching, mainly divided into two categories: one is based on the line matching algorithm, i.e., matches the outline of the polygon as the main feature. 典型算法是Yamamoto提出的基于多边形轮廓的松弛迭代匹配法。 A typical algorithm is based on iterative relaxation polygonal profile matching Yamamoto proposed. 其基本思想是:在多边形轮廓上按照一定算法选取一些特征点,由这些特征点组成近似多边形, 由近似多边形进行松弛迭代匹配达到多边形的匹配;另一类是基于面的匹配算法,即以多边形的面积和“质量”为主要特征进行匹配。 The basic idea is: selecting the polygonal profile in a number of feature points according to a certain algorithm, by such feature points approximate composition polygon, polygon approximation polygon matched by matching iterative relaxation reached; the other is based surface matching algorithm, i.e. polygons area and "quality" match as the main feature. 这类算法中Montgomery等人提出的方法具有一定的代表性,其基本思想是:将序列图形中的任意图形中各个轮廓中心向相邻图形作投影,若相邻图形上各轮廓中心落在距投影点一定阈值范围内,即认为此两轮廓有匹配关系。 Such algorithms Montgomery et al method has certain representation, the basic idea is: an arbitrary sequence pattern in each pattern contour as projected in the center of the adjacent pattern, if the pattern of each adjacent contour falling away from the center the projection point a certain threshold, i.e., a matching contour that these two relationships.

[0009] 基于多边形质心的多边形匹配是充分利用序列图形的内部信息(各多边形之间相互位置信息),根据各多边形的质心位置变化具有连续性这一规律,在相邻帧内寻找各多边形之间的匹配关系。 [0009] Polygon-based matching polygon centroid full use of the internal information (position information between each other polygons) sequential pattern, the continuity of each centroid position according to the law of change of the polygon, each polygon to find the adjacent frame the relationship between match.

[0010] 模式匹配是研究应用最多的匹配方式,是图形相似性判断最常用的手段。 [0010] Pattern matching is the most studied applications matching method is most commonly used pattern similarity judging means. 其主要思想是通过大量的学习和训练,提取出某类相似多边形的一些固定模式,然后利用这些固定模式与未知多边形进行匹配。 The main idea is through a lot of learning and training, to extract certain types of polygons similar to some fixed pattern, and then use these models to match fixing and unknown polygons.

[0011] 基于栅格数据结构的序列图像配准由于数据结构的限制,对于图形变化较大,出现遮挡等情况时无法完成配准工作,基于此方法不能很好的表达序列图形图像的内在联系。 [0011] Based on the sequence data structure of raster image registration due to limitations of the data structure, the registration could not be completed for the graphic work vary widely, where occlusion occurs, based on this method does not well expressed inner link graphic image sequence .

[0012] 基于轮廓匹配的算法由于理论本身问题存在一定缺陷。 [0012] there are some defects due to theoretical problems of its own contour matching algorithm based on. 第一类算法在多边形形状发生较大变化时,难以达到满意效果;第二类方法只孤立地依靠轮廓所在局部的位置进行匹配,很可能产生较大的误差,因此该方法只能达到局部最优,而非全局最优。 The first algorithm is a polygonal shape when large changes occur, it is difficult to achieve satisfactory results; the second isolation methods only rely on the local location where the matching profile, are likely to cause large errors, this method can only achieve a partial most excellent, rather than the global optimum. 另外由于各轮廓尺度大小不同,上述阈值并不是单一的,因而不好确定。 Also due to the different size scales each contour, the threshold value is not a single, and thus well established.

[0013] 基于质心的多边形匹配由于多边形产生叠加分割导致多边形质心产生较大变化, 多数情况下难以完成匹配,只适合变化不大的序列图形。 [0013] Based on the centroid of a polygon match since the superimposition polygon-generating division result in a greater change in the centroid of the polygon, in many cases difficult to complete the match, not only for sequence variation pattern.

[0014] 模式匹配方法由于前提条件是要预先知道一些固定模式,对于千变万化的序列图形在实用性上都受到一些限制。 [0014] Since the pattern matching method with the proviso that previously known to some fixed pattern, for changing a sequence of pattern in the application are subject to some restrictions.

[0015] 拓扑关系是指满足拓扑几何学原理的各空间数据间的相互关系。 [0015] satisfy the topological relationship is the relationship between the spaces of the principles of geometry data topology. 即用结点、弧段和多边形所表示的实体之间的邻接、关联、包含和连通关系。 I.e., with nodes and arcs between adjacent polygons entities represented by correlation, and comprising a communication relationship. 如:点与点的邻接关系、点与面的包含关系、线与面的相离关系、面与面的重合关系等。 Such as: the dots abutting relationship, with the surface including the point of phase relationship, with the wire surface from the relationship, the relationship between surface and the surface coincident the like.

[0016] 拓扑关系是指图形元素之间相互空间上的连接、邻接关系并不考虑具体位置.这种拓扑关系是由数字化的点、线、面数据形成的以用户的查询或应用分析要求进行图形选取、叠合、合并等操作。 [0016] The topology is a connection to each other in the space between the graphic elements, does not consider the specific position abutting relationship. This topology is required to query the user or Application digitized by point, line, surface data formed pattern selection, overlap, merger operation.

发明内容 SUMMARY

[0017] 本发明的目的在于克服上述缺陷,提供一种基于拓扑关系的复杂序列图形匹配方法,该方法建立复杂序列图形的内在联系。 [0017] The object of the present invention is to overcome the above drawbacks, there is provided a complex sequence pattern matching method based on the topological relationship, the method establishes a complex sequence pattern intrinsically linked. 这是智能无纸卡通的核心技术。 This is a smart paperless cartoon core technology.

[0018] 本发明的技术方案是: [0018] aspect of the present invention is:

[0019] 本发明把多边形配准问题归结为两个集合中元素的匹配问题。 [0019] The present invention is a polygonal registration problem is reduced to two matching elements in a set. 问题的实质在于求解最大匹配问题,也即以满足拓扑关系为条件对多边形配准问题求解,求出最大匹配。 Fact of the matter is that the maximum matching problem solving, namely to satisfy the condition of topological polygon registration problem solving, find the maximum matching.

[0020] 一种基于拓扑关系的复杂序列图形匹配系统,包括: [0020] A complex sequence of topological pattern matching system, comprising:

[0021] 识别、提取模块,用于奖前一帧图形中各多边形,得到具有相互分离的顶点集合V, 以及识别、提取与前一帧相邻的后一帧图形中各多边形,得到具有相互分离的顶点集合W; [0021] identification, extraction module, a graphical front prize for each polygon have mutually separated to obtain a set of vertices V, and identifying, after the previous frame adjacent to extract a pattern of each polygon have mutually give W is an isolated set of vertices;

[0022] 有向化模块,用于将顶点集合V中各元素指向顶点集合W连线,进行有向化,得到连接边和有向偶图,获得前一帧图形中各多边形的集合与后一帧图形中各多边形的集合中每一个多边形的拓扑关系; [0022] directional module, for each vertex set V elements in the set of vertices W point connection for each polygon before and after a set of a directed graph, thereby obtaining the connecting side and FIG even have to obtain each polygon of each polygon set topology in a pattern;

[0023] 匹配网络,用于将有向偶图形成匹配网络; [0023] The matching network, the matching network for there to form a bipartite graph;

[0024] 最大流量求解模块,利用最大流量法获得匹配网络中各个连接边的最大结果流量; [0024] The maximum flow solver module, maximum flow results in the matching network connected to the respective side of the flow method using a maximum;

[0025] 关联匹配模块,将获得的最大结果流量与匹配网络进行关联,对顶点集合V和顶点集合W中各元素进行匹配:[0026] 完整匹配输出模块,根据关联结果进行前一帧图形中各多边形与后一帧图形中各多边形的匹配,最终完成前一帧图形与后一帧图形的匹配。 [0025] The correlation matching module, the maximum of flow matching network obtained associating, vertex set V and vertex set W each element matching: [0026] Full match output modules for the previous frame pattern according to the correlation result each polygon match each polygon after a pattern, a pattern matching the final completion of the previous frame and the rear pattern.

[0027] 所述有匹配网络中包括超级源s和超级汇z,超级源s到超级汇ζ之间有若干条通路P。 [0027] The matching network includes a source s and a super super sinks z, super super source s to sink several pathways ζ P.

[0028] 所述通路P上每条边是一致定向的并且每条边上的流量小于通路P流量。 [0028] On each side of the path P is the same orientation and each edge is smaller than the flow passage P traffic.

[0029] 还包括查找模块,用于在匹配网络中查找通路P ; [0029] further comprises a searching module configured to search in the matching network path P;

[0030] 还包括流量增加模块,用于沿查找到的通路P增加流量。 [0030] flow rate is increased further comprising means for passage along the searched flow P increases.

[0031] 一种基于拓扑关系的复杂序列图形匹配方法,包括以下步骤: [0031] The pattern matching method of a complex sequence of topological, comprising the steps of:

[0032] A)识别、提取前一帧图形中各多边形,得到具有相互分离的顶点集合V,识别、提取与前一帧相邻的后一帧图形中各多边形,得到具有相互分离的顶点集合W ;顶点集合V中的各顶点对应前一帧图形中的每个多边形,顶点集合W中的各顶点对应后一帧图形中的每个多边形J1集合中为前一帧图形的所有多边形的集合(A,B,C,D,夂),々、8、(:、0、…为顶点集合V中各分离顶点J2集合中为后一帧图形的所有多边形的集合(a,b,c, d,…), a、b、c、d、…为顶点集合W中各分离顶点; [0032] A) identifying, extracting a pattern before each polygon have mutually separated to obtain a set of vertices V, identification, extracts the previous frame adjacent to a rear drawing polygons, having been separated from each other set of vertices W; vertex corresponding to each vertex set V of a front pattern of each polygon, each polygon J1 set a pattern corresponding to the respective vertices in the vertex in the set W is set of all polygons in the previous frame pattern (a, B, C, D, Wen), 々, 8, (:, 0, ... each vertex set V of vertices separated J2 set for the set of all polygons after a pattern (a, b, c, d, ...), a, b, c, d, ... W is the set of vertices each vertex separation;

[0033] B)将顶点集合V中各元素指向顶点集合W连线,得到连接边,获得前一帧图形中各多边形的集合与后一帧图形中各多边形的集合中每一个多边形的拓扑关系;该连接边即为每一个多边形的拓扑关系; [0033] B) The set V of vertices of each element in the connection point to the set of vertices W, to give connection side, each polygon obtained topological relationship collection set each polygon before a pattern with a pattern after each polygon ; the connecting edge of each polygon is the topology;

[0034] C)利用顶点集合V和顶点集合W形成匹配网络Gt ; [0034] C) using the set of vertices V and W form a set of vertices Gt matching network;

[0035] D)利用最大流量法获得步骤C)中匹配网络Gt中各个连接边的最大结果流量; [0035] D) obtained by the maximum flow method step C) results in the maximum flow rate to match Gt sides of each network connection;

[0036] E)将步骤C获得的最大结果流量与匹配网络进行关联,对顶点集合V和顶点集合W中各元素进行匹配: [0036] E) the maximum matching network traffic results obtained associating step C, the set of vertices V and W the set of vertices of each matching elements:

[0037] 若从顶点集合V的子集ν指向顶点集合W的子集w的边(v,w)的流量为1,则顶点集合V的子集ν与顶点集合W的子集w匹配,则得到匹配网络的流量给出的一个匹配; [0037] If the set of the subset V from the apex ν pointing vertices set W subset w flow side (V, w) is 1, the vertex set sub-V set sub ν vertex set W set w match, matching is obtained a given matching network traffic;

[0038] 若获得最大结果流量,则得到最大匹配; [0038] If the result of maximum flow, then the maximum matching;

[0039] 若最大结果流量的值为IVI,则得到完全匹配; [0039] If the maximum value of flow IVI, exact match is obtained;

[0040] 一个匹配即为顶点集合V中一个多边形与顶点集合W中一个多边形配准; [0040] A match is the vertex set V a set of vertices in a polygon in a polygonal registration W;

[0041] 最大匹配即为顶点集合V中有最多数量的多边形与顶点集合W中相应的多个多边形配准; [0041] vertex set V is the maximum matching in a maximum number of polygon vertices and a corresponding plurality of set W polygonal registration;

[0042] 完全匹配即为顶点集合V中有所有多边形与顶点集合W中相应的所有多边形配准; [0042] The set V of vertices is the exact match for all polygon vertices have set W of all polygons in the respective registration;

[0043] F)根据步骤E的关联结果进行前一帧图形中各多边形与后一帧图形中各多边形的匹配,最终完成前一帧图形与后一帧图形的匹配。 [0043] F) for the previous frame matches the pattern with an each polygon after drawing, before the completion of a final pattern matching with the pattern of a polygon in accordance with the respective correlation result of step E.

[0044] 所述步骤C利用顶点集合V和顶点集合W形成匹配网络Gt包括以下步骤: [0044] Step C using the set of vertices V and W form a set of vertices Gt matching network comprising the steps of:

[0045] 设G是具有相互分离的顶点集合V和W的有向偶图,其中每条边的方向从V中的顶点指向W中的顶点,在有向偶图G中引入超级源s和超级汇ζ ; [0045] Let G be a bipartite graph with a directed away from each other set of vertices V and W wherein the direction of each edge pointing vertices V and W from the apex, the introduction of super source s in the directed graph G and even super sinks ζ;

[0046] 将超级源s指向顶点集合V中各元素的每条边赋值容量1,将顶点集合V中各元素指向顶点集合W中各元素的每条边赋值容量1,将顶点集合W中各元素指向超级汇ζ的每条边赋值容量1,形成匹配网络Gt。 [0046] The super source pointing vertices set V s of the elements in the assignment capacity of each edge 1, the vertex set V pointing vertices of the elements set W of the elements in the assignment capacity of each edge 1, the set of vertices each W super sink element ζ point assignment capacity of each edge 1, a matching network Gt.

[0047] 所述步骤D包括以下步骤:[0048] 从某个初始流量开始,重复地增加流量值直到不能再增加为止。 [0047] Step D comprises the steps of: [0048] from a certain initial flow rate, the flow rate value is repeated until the increased until no further increase. 则结果流量将是一个最大流量; The result will be a maximum flow of traffic;

[0049] a)从匹配网络Gt的一个流量开始,查找匹配网络Gt中超级源s到超级汇ζ的一条通路P,该通路P上每条边是一致定向的并且每条边上的流量小于通路P流量; [0049] a) starting a flow of the matching network Gt, Gt find matching network in a super-source path P s to the super exchange ζ, on each side of the passage P and the flow rate is consistent with the orientation of each edge is less than P flow passage;

[0050] b)若满足上述条件的通路P不存在,则结束查找,通路P上每条边的流量是最大结 [0050] b) When the path P satisfying the above condition does not exist, the search finishes, the flow passage on each side of the junction P is the maximum

果流量; If the flow;

[0051] c)若查找到满足上述条件的通路P,则沿该通路增加流量Δ,直到查找不到满足上述条件的通路P,则返回步骤b。 [0051] c) if it can find the path P satisfying the above conditions, the [Delta] along the flow path increases, until you find the path P above condition is not satisfied, returns to step b.

[0052] 所述流量Δ = minX,X由数Cij-Fij组成;其中Cij为通路P的流量,Fij为通路P中一致定向的边(i,j)的流量。 [0052] The flow rate Δ = minX, X the number Cij-Fij; wherein Cij is a flow passage P, Fij is consistent orientation of the flow passage P in the edge (i, j) of.

[0053] 本发明的有益效果是: [0053] Advantageous effects of the present invention are:

[0054] 本发明实现了复杂序列图形的内在联系建立的难题,是对目前二维动画制作方法进行彻底的技术改造——用计算机替代手工劳动,提高动画产品生产效率的一次根本性的技术突破。 [0054] The invention achieves the problems intrinsically linked complex sequence pattern established, is currently a two-dimensional animation method to conduct a thorough transformation - computer to replace manual labor, improve animation production efficiency a fundamental technology breakthrough .

[0055] 本发明在多边形形状发生较大变化时,仍然可以达到满意效果;克服了孤立地依靠轮廓所在局部的位置进行匹配的弊端,匹配较为精确,该方不仅能达到局部最优,更可以实现全局最优。 [0055] The present invention is a polygonal shape when large changes occur, can still achieve satisfactory results; rely partially overcome the isolation position where the matching profile of drawbacks, more accurate matching, that party can not only achieve a local optimum, but also achieve global optimum.

[0056] 在多边形产生叠加分割导致多边形质心产生较大变化,也可以成功完成匹配。 [0056] generating polygon overlay cause splitting polygon centroid changes significantly, matching can be done successfully. 本发明无需预先设定固定模式,满足序列图形匹配的实时性。 The present invention does not require a predetermined fixed pattern, to meet the real-time sequence of pattern matching. 由于本发明能很好的表达序列图形图像的内在联系,因此本发明对于图形变化较大,出现遮挡等情况也可完成配准工作。 Since the present invention can express inner link sequence pattern image, the present invention is thus to large changes in pattern, where occlusion occurs registration work can be completed.

附图说明 BRIEF DESCRIPTION

[0057] 图1是本发明的基于拓扑关系的复杂序列图形匹配系统的示意图。 [0057] FIG. 1 is a schematic diagram based on a complex sequence pattern matching system of the topology of the present invention.

[0058] 图2是一致定向的边与非一致定向的边的示意图。 [0058] FIG. 2 is a schematic side edge coincides with the orientation of the non-uniform orientation.

[0059] 图3是集合Jl (A,B,C,D)和集合J2 (a,b,c,d,e)匹配的示意图。 [0059] FIG. 3 is Jl (A, B, C, D), and a schematic diagram of J2 (a, b, c, d, e) a set of matching set.

[0060] 图4是最大匹配和一个完全匹配的示意图。 [0060] FIG. 4 is a schematic view of a maximum matching and exact match.

[0061] 图5是匹配问题及匹配网络的示意图。 [0061] FIG. 5 is a schematic diagram of matching and matching network.

[0062] 图6是相互匹配的相邻两帧复杂图形多边形个数没有发生变化时的示意图。 [0062] FIG. 6 is a schematic diagram of two adjacent complex number pattern polygon match each other does not change.

[0063] 图7是相互匹配的相邻两帧复杂图形多边形发生叠加切割的示意图。 [0063] FIG. 7 is a complex pattern of adjacent two mutually matching polygon superimposed schematic cut occurs.

[0064] 图8是相互匹配的相邻两帧复杂图形多边形发生叠加遮挡的示意图。 [0064] FIG. 8 is a complicated pattern of adjacent two mutually matching polygon superimposed schematic occlusion occurs.

具体实施方式 Detailed ways

[0065] 下面结合附图对本发明作进一步描述: [0065] DRAWINGS The present invention is further described:

[0066] 我们注意到,不论图形内部的各个多边形怎样变化,序列图形之间的相互关系基本保持不变或变化缓慢,这样就可以利用拓扑关系进行配准,而达到两帧图形内部各个多边形的配准,从而完成两帧图形的配准。 [0066] We note that, regardless of how each change inside the polygon graphics, graphic correlation between sequence remained unchanged or change slowly, so that you can be registered topological relationship, and achieve two graphics inside each polygon registration to complete two graphic registration. 本发明采用基于复杂图形拓扑关系的最大流量法解决序列图形的匹配问题。 The present invention uses to solve the problem of matching patterns based on the maximum sequence of complicated flow topology graph.

[0067] 一种基于拓扑关系的复杂序列图形匹配系统,包括: [0067] A complex sequence of topological pattern matching system, comprising:

[0068] 识别、提取模块,用于奖前一帧图形中各多边形,得到具有相互分离的顶点集合V,以及识别、提取与前一帧相邻的后一帧图形中各多边形,得到具有相互分离的顶点集合W; [0068] identification, extraction module, a graphical front prize for each polygon have mutually separated to obtain a set of vertices V, and identifying, after the previous frame adjacent to extract a pattern of each polygon have mutually give W is an isolated set of vertices;

[0069] 有向化模块,用于将顶点集合V中各元素指向顶点集合W连线,进行有向化,得到连接边和有向偶图,获得前一帧图形中各多边形的集合与后一帧图形中各多边形的集合中每一个多边形的拓扑关系; [0069] directional module, for each vertex set V elements in the set of vertices W point connection for each polygon before and after a set of a directed graph, thereby obtaining the connecting side and FIG even have to obtain each polygon of each polygon set topology in a pattern;

[0070] 匹配网络,用于将有向偶图形成匹配网络; [0070] The matching network, the matching network for there to form a bipartite graph;

[0071] 最大流量求解模块,利用最大流量法获得匹配网络中各个连接边的最大结果流量; [0071] The solution module maximum flow, maximum flow results in the matching network connected to the respective sides of the method by using the maximum flow rate;

[0072] 关联匹配模块,将获得的最大结果流量与匹配网络进行关联,对顶点集合V和顶点集合W中各元素进行匹配: [0072] module associated with the matching, the matching network and the maximum flow results obtained associating, set V of vertices and the set of vertices W matching each element:

[0073] 完整匹配输出模块,根据关联结果进行前一帧图形中各多边形与后一帧图形中各多边形的匹配,最终完成前一帧图形与后一帧图形的匹配。 [0073] Full match output module, a pre pattern matching polygons each polygon after a drawing, pattern matching a pattern of a previous frame in accordance with the completion of the final correlation result.

[0074] 有匹配网络中包括超级源s和超级汇z,超级源s到超级汇ζ之间有若干条通路P。 [0074] There are several pathways source matching network including super and super sinks s z, s between the super super source sink ζ P.

[0075] 通路P上每条边是一致定向的并且每条边上的流量小于通路P流量。 [0075] On each side of path P and the flow rate is the same orientation on each edge is less than the flow passage P.

[0076] 还包括查找模块,用于在匹配网络中查找通路P ; [0076] further comprises a searching module configured to search in the matching network path P;

[0077] 还包括流量增加模块,用于沿查找到的通路P增加流量。 [0077] flow rate is increased further comprising means for passage along the searched flow P increases.

[0078] —种基于拓扑关系的复杂序列图形匹配方法,包括以下步骤: [0078] - kind of topological complex sequence pattern matching method, comprising the steps of:

[0079] A)识别、提取前一帧图形中各多边形,得到具有相互分离的顶点集合V,识别、提取与前一帧相邻的后一帧图形中各多边形,得到具有相互分离的顶点集合W ;顶点集合V中的各顶点对应前一帧图形中的每个多边形,顶点集合W中的各顶点对应后一帧图形中的每个多边形J1集合中为前一帧图形的所有多边形的集合(A,B,C,D,夂),々、8、(:、0、…为顶点集合V中各分离顶点J2集合中为后一帧图形的所有多边形的集合(a,b,c, d,…), a、b、c、d、…为顶点集合W中各分离顶点; [0079] A) identifying, extracting a pattern before each polygon have mutually separated to obtain a set of vertices V, identification, extracts the previous frame adjacent to a rear drawing polygons, having been separated from each other set of vertices W; vertex corresponding to each vertex set V of a front pattern of each polygon, each polygon J1 set a pattern corresponding to the respective vertices in the vertex in the set W is set of all polygons in the previous frame pattern (a, B, C, D, Wen), 々, 8, (:, 0, ... each vertex set V of vertices separated J2 set for the set of all polygons after a pattern (a, b, c, d, ...), a, b, c, d, ... W is the set of vertices each vertex separation;

[0080] B)将顶点集合V中各元素指向顶点集合W连线,进行有向化,得到连接边,获得前一帧图形中各多边形的集合与后一帧图形中各多边形的集合中每一个多边形的拓扑关系; 该连接边即为每一个多边形的拓扑关系; [0080] B) The set V of vertices of each element in the connection point to the set of vertices W, has to be based, to give connection side, to obtain a set of the set of polygons each polygon before a pattern with a pattern after each a polygon topology; each polygon is the topology of the connection edge;

[0081] C)利用顶点集合V和顶点集合W形成匹配网络Gt ; [0081] C) using the set of vertices V and W form a set of vertices Gt matching network;

[0082] D)利用最大流量法获得步骤C)中匹配网络Gt中各个连接边的最大结果流量; [0082] D) obtained by the maximum flow method step C) results in the maximum flow rate to match Gt sides of each network connection;

[0083] E)将步骤C获得的最大结果流量与匹配网络进行关联,对顶点集合V和顶点集合W中各元素进行匹配: [0083] E) the maximum matching network traffic results obtained associating step C, the set of vertices V and W the set of vertices of each matching elements:

[0084] 若从顶点集合V的子集ν指向顶点集合W的子集w的边(v,w)的流量为1,则顶点集合V的子集ν与顶点集合W的子集w匹配,则得到匹配网络的流量给出的一个匹配; [0084] If the set of the subset V from the apex ν pointing vertices set W subset w flow side (V, w) is 1, the vertex set sub-V set sub ν vertex set W set w match, matching is obtained a given matching network traffic;

[0085] 若获得最大结果流量,则得到最大匹配; [0085] If the result of maximum flow, then the maximum matching;

[0086] 若最大结果流量的值为IVI,则得到完全匹配; [0086] If the maximum value of flow IVI, exact match is obtained;

[0087] 一个匹配即为顶点集合V中一个多边形与顶点集合W中一个多边形配准; [0087] A match is the vertex set V a set of vertices in a polygon in a polygonal registration W;

[0088] 最大匹配即为顶点集合V中有最多数量的多边形与顶点集合W中相应的多个多边形配准; [0088] vertex set V is the maximum matching in a maximum number of polygon vertices and a corresponding plurality of set W polygonal registration;

[0089] 完全匹配即为顶点集合V中有所有多边形与顶点集合W中相应的所有多边形配准;[0090] F)根据步骤E的关联结果进行前一帧图形中各多边形与后一帧图形中各多边形的匹配,最终完成前一帧图形与后一帧图形的匹配。 [0089] The set V of vertices is the exact match has all polygon vertices corresponding set W of all polygons registered; [0090] F) for a front and rear graphical each polygon according to a pattern correlation result of Step E matching each polygon, finally completed a pattern matching a pattern of the front and the rear.

[0091] 步骤C利用顶点集合V和顶点集合W形成匹配网络Gt包括以下步骤: [0091] Step C using the set of vertices V and W form a set of vertices Gt matching network comprising the steps of:

[0092] 设G是具有相互分离的顶点集合V和W的有向偶图,其中每条边的方向从V中的顶点指向W中的顶点,在有向偶图G中引入超级源s和超级汇ζ ; [0092] Let G be a bipartite graph with a directed away from each other set of vertices V and W wherein the direction of each edge pointing vertices V and W from the apex, the introduction of super source s in the directed graph G and even super sinks ζ;

[0093] 将超级源s指向顶点集合V中各元素的每条边赋值容量1,将顶点集合V中各元素指向顶点集合W中各元素的每条边赋值容量1,将顶点集合W中各元素指向超级汇Z的每条边赋值容量1,形成匹配网络Gt。 [0093] The super source pointing vertices set V s of the elements in the assignment capacity of each edge 1, the vertex set V pointing vertices of the elements set W of the elements in the assignment capacity of each edge 1, the set of vertices each W element refers to the super exchange capacity assignment Z 1 each side, forming a matching network Gt.

[0094] 步骤D包括以下步骤: [0094] Step D comprises the steps of:

[0095] 从某个初始流量开始,重复地增加流量值直到不能再增加为止。 [0095] from a certain initial flow rate, the flow rate value is repeated until the increased until no further increase. 则结果流量将是一个最大流量; The result will be a maximum flow of traffic;

[0096] a)从匹配网络Gt的一个流量开始,查找匹配网络Gt中超级源s到超级汇ζ的一条通路P,该通路P上每条边是一致定向的并且每条边上的流量小于通路P流量; [0096] a) starting a flow of the matching network Gt, Gt find matching network in a super-source path P s to the super exchange ζ, on each side of the passage P and the flow rate is consistent with the orientation of each edge is less than P flow passage;

[0097] b)若满足上述条件的通路P不存在,则结束查找,通路P上每条边的流量是最大结 [0097] b) When the path P satisfying the above condition does not exist, the search finishes, the flow passage on each side of the junction P is the maximum

果流量; If the flow;

[0098] c)若查找到满足上述条件的通路P,则沿该通路增加流量Δ,直到查找不到满足上述条件的通路P,则返回步骤b。 [0098] c) if it can find the path P satisfying the above conditions, the [Delta] along the flow path increases, until you find the path P above condition is not satisfied, returns to step b.

[0099] 流量Δ = minX, X由数Cij-Fij组成;其中Cij为通路P的流量,Fij为通路P中一致定向的边(i,j)的流量。 [0099] Flow Δ = minX, X the number Cij-Fij; wherein Cij is a flow passage P, Fij is consistent orientation of the flow passage P in the edge (i, j) of.

[0100] 具体技术方案如下: [0100] In particular the following technical solution:

[0101] 如果Gt是一个传输网络,G中的一个最大流量(maximal flow)是具有最大值的流量。 [0101] If a transmission network is Gt, a maximum flow rate (maximal flow) G is the flow rate has a maximum value. 一般地,可能存在几个具有相同最大值的流量。 Generally, there may be several have the same maximum flow rate. 给出求得最大流量的一个算法。 Give an algorithm to obtain maximum flow. 其基本概念是简单的,从某个初始流量开始,重复地增加流量值直到不能再增加为止。 The basic concept is simple, start with one of the initial flow rate, repeatedly until the value increases until the flow can not be increased. 则结果流量将是一个最大流量。 The result will be a maximum flow of traffic.

[0102] 将每条边上的流量设置为0作为初始流量。 [0102] The flow rate of each edge is set to 0 as the initial flow rate. 为了增加给定流量的值,必须找出从源到汇的一条通路并沿这条通路增加流量。 To increase the value of a given flow rate, you must find a path from the source to the sink and increases flow along this path.

[0103] 引入某些术语。 [0103] Certain terminology is introduced. 暂时地,考虑G的边是无向的并设 Temporarily, consider the edge G is undirected and is provided

[0104] P = (v0, V1,…,vn) , V0 = a, Vn = ζ [0104] P = (v0, V1, ..., vn), V0 = a, Vn = ζ

[0105] 是这个无向图中从a到ζ的一条通路。 [0105] This is a undirected graph from a path to the ζ. 如果在P中边e是从ν"指向Vi,则e是一致定向的(关于P);否则,则e是非一致定向的(关于P),如图2。 If edge e is from the P ν "points Vi, e is the same orientation (with respect to P); otherwise, the non-uniform orientation of e (P on), as shown in FIG 2.

[0106] 如果能够找到一条通路P,使通路P上每条边是一致定向的并且每条边上的流量小于其流量,就可能增加流量值。 [0106] If able to find a path P, so that each side of the path P is the same orientation and each edge is smaller than the flow rate of the flow, it is possible to increase the flow rate value.

[0107] 给出如下定理: [0107] the following theorem gives:

[0108]定理 2: [0108] Theorem 2:

[0109] 设P是网络Gt中从a到ζ的满足下列条件的通路: [0109] Let P be the Gt from a network to satisfy the following conditions ζ passage:

[0110] 对P中每个一致定向的边(i,j), [0110] Each of the P directional consistent edge (i, j),

[0111] Fij < Cij [0111] Fij <Cij

[0112] Cij为通路P的流量,Fij为通路P中一致定向的边(i,j)的流量(b)对P中每个非一致定向的边(i,j);[0113] 设0 < F' ,j ;F' ,j为通道P中每个非一致性定向的边的流量。 [0112] Cij is the flow passage P, Fij is consistent path P oriented edge (i, j) of the flow (b) each side of the non-uniform orientation of P in the (i, j); [0113] 0 provided <F ', j; F', j flow passage P side is directed to each of the non-uniformity.

[0114] Δ= minX [0114] Δ = minX

[0115] 这里X由数Cij-Fij (对于P中一致定向的边(i,j))和Fij (对于P中非一致定向的边(i,j))组成。 [0115] where X is the number Cij-Fij (P consistent orientation to edge (i, j)), and Fij (directed to the same side of Africa P (i, j)) composition. 定义 definition

[0116] [0116]

(i,j)不在P中 (I, j) is not in the P

(i,j)在P中一致定向 (I, j) in the same orientation in the P

(i,j)在P中非一致定向 (I, j) is oriented in the same P-Africa

[0117] 则F*是一个流量,它的值比F的流量值大Δ。 [0117] F * is the one flow, which is greater than the flow rate F value Δ. 如果不存在满足定理2条件的通路, 则流量是最大的。 If there is no path satisfying the condition of Theorem 2, the flow is greatest.

[0118] [01]

[0119] [0119]

[0120] [0120]

的; [0121] [0122] ; And [0121] [0122]

[0123] [0123]

[0124] [0124]

根据定理2我们就可以构造求最大流量的算法。 According to Theorem 2 we can construct algorithms to solve the maximum flow.

1.从一个流量(例如,每条边上的流量为0)开始; 1. From a flow (e.g., flow rate of 0 on each edge) begins;

2.查找满足定理2条件的一条通路,如果如此的通路不存在,结束;流量是最大 2. Find a path to meet the conditions of Theorem 2, if such a path does not exist, end; traffic is the biggest

3.如果此通路P存在,则沿增加流量Δ,其中Δ按定理2定义,并转向2。 3. If the path P is present, the increased flow along the [Delta], wherein [Delta] is defined by Theorem 2, 2 and the steering. 在形式化的算法中,我们查找满足定理2条件的一条通路,并同时追踪量Cij_Fij,Fij In formal algorithm, we find a path to meet the conditions of Theorem 2, and also track the amount of Cij_Fij, Fij

序列图形的匹配问题可以看作一个集合中的元素与另一个集合中的元素的匹配问题,我们可以把这个问题简化成求网络的最大流量,先从一个例子开始。 Matching sequences of graphic elements can be seen as a set with matching elements in the other set, we can simplify the problem into a maximum flow requirements of the network, start with an example of a start.

[0125] 假定集合J1由四个多边形A、B、C和D组成,集合J2由五个多边形a、、b、、c、d和e组成。 [0125] J1 is assumed that a set of four polygons A, B, C, and D, J2 set of five polygons a ,, b ,, c, d, and e components.

[0126]即 J1 = {A,B,C,D} [0126] i.e. J1 = {A, B, C, D}

[0127] J2 = {a, b, c, d, e} [0127] J2 = {a, b, c, d, e}

[0128] 假定按照某种条件,多边形A与多边形b和e配准,多边形B也与多边形b和e配准,多边形C与多边形a、c、、d和e配准,多边形D与多边形b和e配准。 [0128] according to certain assumed conditions, polygon A polygon registration b and e, and the polygon B also polygonal registration b and e, polygon polygon C a, c ,, d and e registration, D polygon polygon b and e registration. 两个集合J1和J2 中的每个多边形都能够相互配准吗? Two sets J1 and J2 each polygon can be co-registered it?

[0129] 这种情形能够用图2的图形刻划。 [0129] This situation can be scribed with a pattern 2 of FIG. 顶点代表多边形,一条边连接J1中的一个多边形和J2中的与之配准的多边形。 Representative polygon vertex, an edge is connected with the registration of a polygon J1 and J2 in the polygon. 考虑到多边形A、B和C,它们都与多边形b和e配准,从图2可以清楚地看出:不可能使每个多边形都能够相互配准。 Considering the polygons A, B and C, they can be clearly seen that the polygonal registration b and e in FIG. 2: impossible to each polygon can be co-registered. 如果多边形A和B各与J2中的一个多边形配准,则多边形D无法配准。 If polygons A and B each with a polygonal J2 in registration, registration is not D polygon. 因此,不存在对多边形A、B、C和D的完全配准。 Thus, complete with registration of the polygon A, B, C, and D does not exist.

[0130] 在此例中,一个匹配由J1的一个多边形与J2的一个多边形配准组成。 [0130] In this example, a match of a polygon with a polygonal J1 J2 registration composition. 一个最大匹配使最多的多边形得到匹配。 To make up to a maximum matching polygon matched. 对于图2图形,粗线表示最大的匹配。 For the graph in Figure 2, the thick line indicates the maximum matching. 完全匹配使J1的每个多边形得到配准。 Exact match for each polygon obtained J1 registration. 我们证明了图2图形没有完全匹配。 We proved that graph in Figure 2 does not match exactly.

[0131] [0131]

大个数。 Large number.

[0132] [0132]

[0133] [0133]

[0134] [0134]

图3中,边连接相配准的多边形。 In FIG 3, in registration polygon edge connector. 粗线表示最大的匹配,即能够配准的多边形的最 Thick line represents the maximum match, that is capable of registering the most polygons

下面给出完全配准的定义: 定义4 Give complete registration of the following definitions: Definition 4

设G是具有相互分离的顶点集合V和W的有向偶图,其中每条的方向从V中的顶 Let G be separated from each other having a vertex set V and W have the dual graph, where each direction from the top V

10点指向W中的顶点。 10:00 pointing vertices of W. G中的一个匹配时没有公共顶点的边集E。 No common edge when a matching set of vertices of G E. G的最大匹配时含有最大变数的匹配E。 E. matching variable containing the maximum when the maximum matching of G G的一个完全匹配是一个匹配E,它具有性质:如果ν e V,则对某个we W,存在(v,W) e E。 G is a match is a match E, which has the property: if ν e V, on the one we W, the presence of (v, W) e E.

[0135] 图4中,黑粗线表示的匹配是一个最大匹配且是一个完全匹配。 In [0135] FIG 4, matching a black bold line represents the maximum matching and is a match.

[0136] 下面讨论如何将匹配问题用网络模型表示。 How [0136] The following discussion will be represented by matching network model.

[0137] 首先对图3的图形中的每条边赋值容量1,如图5。 [0137] First, the pattern of FIG. 3 in each side of the assignment capacity of 1, 5 in FIG. 其后我们增加一个超级源s和从s到A、B、C和D中每个容量为1的边。 Then we added a super source from s to s and A, B, C and D each with a capacity of 1 side. 最后我们引入一个超级汇ζ和从a、b、c、d、e中每一个到ζ的容量为1的边。 Finally, we introduced a super sink ζ and the a, b, c, d, e to each side capacity of ζ 1. 我们称如图4的网络为匹配网络。 We call for the network in Figure 4 matching network.

[0138] 下面的定理关联匹配网络和流量。 [0138] The following theorem associated matching network and traffic.

[0139]定理 3 [0139] Theorem 3

[0140] 设G是具有相互分离的顶点集合V和W的有向偶图,其中每条边的方向从V中的顶点指向W中的顶点。 [0140] Let G be a set of vertices separated from each other to have a bipartite graph V and W wherein the direction of each edge pointing vertices V and W from the apex. G中的任意顶点在V中或在W中。 G is any vertex in V or W is.

[0141] (a)匹配网络的流量给出G的一个匹配。 [0141] (a) gives a matching network traffic matching of G. 顶点ν e V与W e W匹配当且仅当边(v, w)的流量为1。 Vertex ν e V and W e W if and only if traffic matching edge (v, w) is 1.

[0142] (b)最大流量对应于最大匹配。 [0142] (b) the maximum flow rate corresponding to the maximum matching.

[0143] (c)值为|V|的流量对应于完全匹配。 [0143] (c) the value of | V | flow rate corresponding to the exact match.

[0144] 证明:设a(z)是匹配网络的源(汇)并假定流量已知。 [0144] Proof: set a (z) is the matching network source (sink) and assuming the flow rate is known.

[0145] 假定边(v,w)有流量1,其中ν e V, we W。 [0145] assumed that edge (v, w) with a flow rate of 1, wherein ν e V, we W. 进入顶点ν的唯一的边是(a,v)。 The only side into the vertex is ν (a, v). 这个边必然有流量1,于是进入顶点ν的流量是1。 This flow must have an edge, then flow into the vertex is ν 1. 因为流出ν的流量也是1,流量为1的形式如(ν,χ)的唯一的边(ν,w)。 Because flow out of v is 1, in the form of a flow, such as (ν, χ) is the only edge (ν, w). 类似地,流量为1形式为(X,w)的唯一边是(V,w)。 Similarly, the flow rate is 1 in the form of (X, w) is the only edge (V, w). 如果E是流量为1的形式如(v,w)的边的集合,则E的成员没有公共顶点,于是给出G的一个匹配。 If such flow is E (v, w) in the form of a set of edges, the member is not a common vertex E, then G is given a match.

[0146] (b)和(C)两部分直接从如下事实得出:V中匹配的顶点数等于相应流量的值。 [0146] (b) and (C) two partially derived from the fact that: V matching the number of vertices equal to the value corresponding to the flow rate.

[0147] 因为最大流量给出最大匹配,最大流量算法应用于匹配网络生成最大网络。 [0147] Since the maximum flow rate that gives the maximum matching, the maximum flow algorithm to generate the maximum matching network network.

[0148] 这样基于拓扑关系的多边形配准问题,就是考虑一帧图形内的多边形与相邻帧图形所包含的多边形的配准问题。 [0148] Such polygonal registration problem of topological, registration problem is to consider a polygon polygon in an adjacent frame pattern included in a pattern. 实际上就是讨论一个集合中的元素匹配另一个集合中的元素的问题。 Actually discuss a set of matched elements to another set of elements.

[0149] J1集合中为前一帧图形的所有多边形的集合(A,B,C,D,…), [0149] J1 is set before the set of all polygons in a pattern (A, B, C, D, ...),

[0150] J2集合中为后一帧图形的所有多边形的集合(a,b,c,d,…), [0150] J2 is set in all the set of polygons after a pattern (a, b, c, d, ...),

[0151] 连接边为每一个多边形的拓扑关系。 [0151] connected to each side of a polygon topology.

[0152] 用该技术分三种情况进行配准实验,实验时,界面左侧窗口的卡通图形为前一帧图形,界面右侧窗口的卡通图形为后一帧图形。 [0152] The three cases using the registration technique experiment, experiment, cartoon graphics interface window on the left front is a cartoon graphics graphical interface right of the window after a pattern.

[0153] 第一种情况,如图6,是相邻帧多边形个数没有改变,拓扑关系也没有发生变化。 [0153] The first case, as shown in FIG 6, the number of polygons adjacent frames is not changed, there was no change topology.

[0154] 第二种情况,如图7,是多边形发生叠加切割现象,多边形个数以及拓扑关系均发生了变化。 [0154] The second case, as shown in FIG 7, is superimposed on the polygonal cutting occurs, the topology and the number of polygons are changed. 即卡通人物的左手分别位于其身体的不同位置,造成不同色块多边形的叠加切割。 That cartoon characters left are located in different locations in the body, resulting in different color overlay polygon cutting.

[0155] 第三种情况,如图8,是多边形发生叠加遮挡现象,导致多边形个数发生变化。 [0155] The third case, as shown in FIG 8, a polygon overlay blocking phenomenon, resulting in a change in the number of polygons. 即卡通人物的左手分别位于其身体的不同位置,造成不同色块多边形的叠加切割。 That cartoon characters left are located in different locations in the body, resulting in different color overlay polygon cutting. 即卡通人物的右手分别位于其身体的不同位置,造成不同色块多边形的叠加遮挡。 That's right hand cartoon characters are located in different locations in the body, resulting in different color overlay polygon blocked. [0156] 三种情况的实验结果如下:相同颜色的多边形表示成功匹配的多边形。 Experimental Results [0156] The following three cases: the same color polygon polygon representing a successful match.

Claims (8)

  1. 一种基于拓扑关系的复杂序列图形匹配系统,其特征是包括:识别、提取模块,用于将前一帧图形中各多边形,得到具有相互分离的顶点集合V,以及识别、提取与前一帧相邻的后一帧图形中各多边形,得到具有相互分离的顶点集合W;有向化模块,用于将顶点集合V中各元素指向顶点集合W连线,进行有向化,得到连接边和有向偶图,获得前一帧图形中各多边形的集合与后一帧图形中各多边形的集合中每一个多边形的拓扑关系;匹配网络,用于将有向偶图形成匹配网络;最大流量求解模块,利用最大流量法获得匹配网络中各个连接边的最大结果流量;关联匹配模块,将获得的最大结果流量与匹配网络进行关联,对顶点集合V和顶点集合W中各元素进行匹配:完整匹配输出模块,根据关联结果进行前一帧图形中各多边形与后一帧图形中各多边形的匹配,最 A complex sequence pattern matching based system topology which is characterized in comprising: identifying, extracting means for drawing a front polygons, having been separated from each other vertex set V, and identifying, extracting the previous frame after a pattern adjacent to each polygon vertex set W has been separated from each other; the module has, for each element in the vertex set V W vertex set point connection has to be oriented, and the connecting side to give directed bipartite graph, each polygon obtained before the topology graph a collection of polygons after a drawing each polygon set; matching network for matching network is formed to have the bipartite graph; maximum flow solving module, with the maximum flow method to obtain the maximum result of the flow matching network side of the respective connection; maximum of flow matching network associated with a matching module obtained associating, vertex set V and vertex set W each element matching: full match output modules, each polygon to match a pattern before each polygon after a pattern according to the correlation result, most 完成前一帧图形与后一帧图形的匹配。 Pattern matching is completed a previous frame and the rear pattern.
  2. 2.根据权利要求1所述的基于拓扑关系的复杂序列图形匹配系统,其特征是所述有匹配网络中包括超级源s和超级汇z,超级源s到超级汇ζ之间有若干条通路P。 The complex sequence of topological pattern matching system according to claim 1, wherein said matching network comprises a source of super and super sinks Z s, super source s to a number of paths between the super exchange ζ P.
  3. 3.根据权利要求2所述的基于拓扑关系的复杂序列图形匹配系统,其特征是所述通路P上每条边是一致定向的并且每条边上的流量小于通路P流量。 3. The complex sequence of topological pattern matching system according to claim 2, characterized in that on each side of the path P is the same orientation and each edge is smaller than the flow passage P traffic.
  4. 4.根据权利要求1或3所述的基于拓扑关系的复杂序列图形匹配系统,其特征是还包括查找模块,用于在匹配网络中查找通路P ;还包括流量增加模块,用于沿查找到的通路P增加流量。 The complex sequence of topological pattern matching system of claim 1 or claim 3, characterized by further comprising a searching module configured to search in the matching network path P; flow rate increases further comprising means for direction to find the flow passage P increases.
  5. 5. 一种基于拓扑关系的复杂序列图形匹配的实现方法,其特征是包括以下步骤:A)识别、提取前一帧图形中各多边形,得到具有相互分离的顶点集合V,识别、提取与前一帧相邻的后一帧图形中各多边形,得到具有相互分离的顶点集合W;B)将顶点集合V中各元素指向顶点集合W连线,得到连接边,获得前一帧图形中各多边形的集合与后一帧图形中各多边形的集合中每一个多边形的拓扑关系;C)利用顶点集合V和顶点集合W形成匹配网络Gt ;D)利用最大流量法获得步骤C)中匹配网络Gt中各个连接边的最大结果流量;E)将步骤C获得的最大结果流量与匹配网络进行关联,对顶点集合V和顶点集合W中各元素进行匹配:若从顶点集合V的子集ν指向顶点集合W的子集w的边(v,w)的流量为1,则顶点集合V的子集ν与顶点集合W的子集w匹配,则得到匹配网络的流量给出的一个匹配 A method to realize a complicated sequence of pattern matching based topology, characterized by comprising the following steps: A) identifying, extracting a pattern before each polygon have mutually separated to obtain a set of vertices V, identification, extraction and before a post adjacent a drawing each polygon, to give the set of vertices W have mutually separated; B) vertices set V of vertices of each element point connection set W, to obtain a connection side, a pattern obtained prior to each polygon collection of each polygon of the topology set each polygon after a pattern with; C) using a vertex set V and a set of vertices W form a matching network Gt; matching network Gt obtained in step C D) by using the maximum flow method) maximum of flow edge of each connection; E) the maximum of flow matching network in step C obtained associating, vertex set V and vertex set W each element matches: If the set of the subset V from the apex ν point to a set of vertices flow side (V, w) a subset of w is 1 W, the sub vertex set V w is set to match a subset of the vertex set W ν is obtained a matching network traffic matching analysis 若获得最大结果流量,则得到最大匹配;若最大结果流量的值为|V|,则得到完全匹配;F)根据步骤E的关联结果进行前一帧图形中各多边形与后一帧图形中各多边形的匹配,最终完成前一帧图形与后一帧图形的匹配。 If the result of maximum flow, then the maximum matching; if the result of the maximum flow rate value | V |, exact match is obtained; F.) For each pattern before a polygon after a pattern according to each correlation result of Step E matching polygonal, before the completion of a final match pattern and a rear pattern.
  6. 6.根据权利要求5所述的基于拓扑关系的复杂序列图形匹配的实现方法,其特征是所述步骤C利用顶点集合V和顶点集合W形成匹配网络Gt包括以下步骤:设G是具有相互分离的顶点集合V和W的有向偶图,其中每条边的方向从V中的顶点指向W中的顶点,在有向偶图G中引入超级源s和超级汇ζ ;将超级源s指向顶点集合V中各元素的每条边赋值容量1,将顶点集合V中各元素指向顶点集合W中各元素的每条边赋值容量1,将顶点集合W中各元素指向超级汇ζ的每条边赋值容量1,形成匹配网络Gt。 The implementation complexity of a sequence of topological pattern matching according to claim 5, wherein said step C using a set of vertices V and W form a set of vertices Gt matching network comprising the steps of: G is provided with mutually isolated the set of vertices V and W have the dual graph in which each edge direction pointing vertices V and W from the apex, there is incorporated in the super and super source s to sink ζ dual graph G; s pointing to the super source vertex set V of the elements in the assignment capacity of each edge 1, the set of vertices V of the elements in the set of vertices of each edge point W of the elements 1 assigned capacity, the point of each element ζ super exchange of each vertex set W 1 assigned capacity side, forming a matching network Gt.
  7. 7.根据权利要求5所述的基于拓扑关系的复杂序列图形匹配的实现方法,其特征是所述步骤D包括以下步骤:a)从匹配网络Gt的一个流量开始,查找匹配网络Gt中超级源s到超级汇ζ的一条通路P,该通路P上每条边是一致定向的并且每条边上的流量小于通路P流量;b)若满足上述条件的通路P不存在,则结束查找,通路P上每条边的流量是最大结果流量;c)若查找到满足上述条件的通路P,则沿该通路增加流量Δ,直到查找不到满足上述条件的通路P,则返回步骤b。 The implementation complexity of a sequence of topological pattern matching according to claim 5, characterized in that said step D comprises the steps of: a) starting a flow of Gt matching network, the matching network to find the super source Gt s to sink ζ of a super path P, on each side of the passage P and the flow rate is consistent with the orientation of each edge is smaller than the flow passage P; b) if the path P satisfying the above condition does not exist, the search finishes, passage traffic on each side of flow P is the maximum; c) If the path to find P satisfying the above conditions, the [Delta] along the flow path increases, until you find the path P above condition is not satisfied, returns to step b.
  8. 8.根据权利要求7所述的基于拓扑关系的复杂序列图形匹配的实现方法,其特征是所述流量Δ= minX,X由数Cij-Fij组成;其中Cij为通路P的流量,Fij为通路P中一致定向的边(i,j)的流量。 The implementation complexity of a sequence of topological pattern matching according to claim 7, wherein said flow rate Δ = minX, X the number Cij-Fij; wherein Cij is a flow passage P, Fij for the passage P consistent flow directed edge (i, j) of.
CN 201010193054 2010-06-03 2010-06-03 Complex sequence graphics matching system based on topological relation and realization method thereof CN101894274A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN 201010193054 CN101894274A (en) 2010-06-03 2010-06-03 Complex sequence graphics matching system based on topological relation and realization method thereof

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN 201010193054 CN101894274A (en) 2010-06-03 2010-06-03 Complex sequence graphics matching system based on topological relation and realization method thereof

Publications (1)

Publication Number Publication Date
CN101894274A true CN101894274A (en) 2010-11-24

Family

ID=43103462

Family Applications (1)

Application Number Title Priority Date Filing Date
CN 201010193054 CN101894274A (en) 2010-06-03 2010-06-03 Complex sequence graphics matching system based on topological relation and realization method thereof

Country Status (1)

Country Link
CN (1) CN101894274A (en)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1168035A (en) * 1996-04-30 1997-12-17 株式会社鹰山 Spectrum spread communications system for high-speed communications
US20070265531A1 (en) * 2001-09-28 2007-11-15 Xingbai He Ultrasound imaging system
US20080018672A1 (en) * 2006-07-21 2008-01-24 Chunghwa Picture Tubes, Ltd. Method for improving image stitch-in phenomenon
CN101540892A (en) * 2009-04-23 2009-09-23 上海中安电子信息科技有限公司 Method for people counting in doorway on DSP video gathering device
CN101561249A (en) * 2009-05-19 2009-10-21 上海理工大学 Device and method for automatically detecting fit dimension of surgical knife blade

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1168035A (en) * 1996-04-30 1997-12-17 株式会社鹰山 Spectrum spread communications system for high-speed communications
US20070265531A1 (en) * 2001-09-28 2007-11-15 Xingbai He Ultrasound imaging system
US20080018672A1 (en) * 2006-07-21 2008-01-24 Chunghwa Picture Tubes, Ltd. Method for improving image stitch-in phenomenon
CN101540892A (en) * 2009-04-23 2009-09-23 上海中安电子信息科技有限公司 Method for people counting in doorway on DSP video gathering device
CN101561249A (en) * 2009-05-19 2009-10-21 上海理工大学 Device and method for automatically detecting fit dimension of surgical knife blade

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
《测绘学院学报》 20011231 张宝印 等 时序图形相邻两帧的配准 , *

Similar Documents

Publication Publication Date Title
Labatut et al. Robust and efficient surface reconstruction from range data
Shen et al. Structure recovery by part assembly
Katartzis et al. A stochastic framework for the identification of building rooftops using a single remote sensing image
CN101872416B (en) Vehicle license plate recognition method and system of road image
CN101086788A (en) Method and device for generating a disparity map from stereo images and stereo matching method and device therefor
JP2005228131A (en) Drawing verification device
Martinovic et al. 3d all the way: Semantic segmentation of urban scenes from start to end in 3d
CN102142153A (en) Image-based remodeling method of three-dimensional model
CN101388115B (en) Depth image autoegistration method combined with texture information
Planitz et al. The correspondence framework for 3D surface matching algorithms
Gal et al. Surface reconstruction using local shape priors
Lee et al. An algorithm for automatic 2D quadrilateral mesh generation with line constraints
Budd et al. Global non-rigid alignment of surface sequences
Li et al. Locality preserving constraints for super-resolution with neighbor embedding
CN101055620A (en) Shape comparison device and method
Chang et al. Surface reconstruction from point clouds by transforming the medial scaffold
CN101408941A (en) Method for multi-dimension segmentation of remote sensing image and representation of segmentation result hierarchical structure
Wang et al. Projective analysis for 3D shape segmentation
CN101719140B (en) Figure retrieving method
GB2457215A (en) Automatic 3D Modelling
JP6395158B2 (en) How to semantically labeling the acquired image of the scene
Orbay et al. Beautification of design sketches using trainable stroke clustering and curve fitting
Concha et al. Using superpixels in monocular SLAM
CN101706950B (en) High-performance implementation method for multi-scale segmentation of remote sensing images
Zhang et al. Road extraction by deep residual u-net

Legal Events

Date Code Title Description
C06 Publication
C10 Request of examination as to substance