New! View global litigation for patent families

CN101719140B - Figure retrieving method - Google Patents

Figure retrieving method Download PDF

Info

Publication number
CN101719140B
CN101719140B CN 200910214068 CN200910214068A CN101719140B CN 101719140 B CN101719140 B CN 101719140B CN 200910214068 CN200910214068 CN 200910214068 CN 200910214068 A CN200910214068 A CN 200910214068A CN 101719140 B CN101719140 B CN 101719140B
Authority
CN
Grant status
Grant
Patent type
Prior art keywords
dimensional
grid
models
retrieving
figure
Prior art date
Application number
CN 200910214068
Other languages
Chinese (zh)
Other versions
CN101719140A (en )
Inventor
王召福
罗笑南
许晓伟
Original Assignee
中山大学
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
Grant date

Links

Abstract

The invention discloses a figure retrieving method, comprising the steps of: (1) modeling a polygon; establishing a three-dimensional grid model base; (2) matching three-dimensional grid models with two-dimensional images or figures; (3) extracting frameworks of the three-dimensional grid models; (4) retrieving the three-dimensional models according to the frameworks; (5) extracting characteristic points of the three-dimensional grid models; (6) computing control points of the three-dimensional grid models; (7) computing frequency spectra of the control points of the three-dimensional grid models; and (8) computing the similarity of the frequency spectra and retrieving a corresponding figure according to the similarity. The technical scheme is more convenient for retrieval and supports multimode retrieval.

Description

一种图形检索方法 A graphical search method

技术领域 FIELD

[0001] 本发明涉及图形处理技术领域,具体涉及一种图形检索方法。 [0001] The present invention relates to image processing technology, and in particular relates to a pattern search method. 背景技术 Background technique

[0002] 网络与计算机图形学逐渐渗入日常生活中,人们已不再满足于只能在网络上看到二维图像。 [0002] Network and Computer Graphics gradually infiltrated everyday life, people are no longer satisfied with the two-dimensional image can be seen on the network. 另一方面,随着三维建模技术的日益成熟和计算机软硬件技术的飞速发展,三维模型的数量在最近的十年中有了飞跃性的增长。 On the other hand, with the rapid development of technology has become more sophisticated three-dimensional modeling and computer hardware and software technology, the number of three-dimensional model of a leap of growth in the last decade. 相对于二维图像,人们可以从任意角度浏览自己感兴趣部分,因此更受人喜爱,用途更广泛。 With respect to the two-dimensional image, one can view from any angle yourself interesting parts, and therefore more likeable, more versatile. 网络游戏和动漫技术、网络教育技术、基于Web的信息服务关键技术及产品和数据库与数据挖掘技术等热点领域的研究均不能缺少三维图形这一媒体。 Research in the field of online games and animation technology, learning technology, Web-based information services and database products and key technologies and data mining technology and other hot spots are not the lack of three-dimensional graphics of the media.

[0003] 充分利用已有的三维模型数据资源,可以大大减轻设计新模型的工作量,同时也可以促进三维数据的流通和在各领域的应用。 [0003] full use of the existing three-dimensional model data resources, can greatly reduce the workload of the design of the new model, but also can promote the circulation and application of three-dimensional data in various fields. 然而,如何在海量的三维模型库中快速的搜索到自己感兴趣的模型,为三维模型库建立搜索引擎是一个困难的问题。 However, how fast searching in massive library of three-dimensional model to model their own interest, to establish a three-dimensional model library search engines is a difficult problem. 现有技术的图形检索方法是根据几何内容对三维模型进行分类检索,用户不能方便地通过检索界面表达检索要求。 Graphic search method of the prior art are classified according to the geometric three-dimensional model retrieval content, the user can not easily retrieval request expressed by the search interface.

发明内容 SUMMARY

[0004] 本发明要解决的技术问题是提供一种图形检索方法,能够克服现有技术的不足, 实现图形的多模态检索,使得检索更为方便,完善目前的多媒体搜索技术和填补目前网络三维图形搜索引擎的空白,推动下一代智能多模态搜索引擎的实现。 [0004] The present invention is to solve the technical problem of providing a graphic search method capable of overcoming the disadvantages of the prior art, multi-modal pattern retrieved, so more convenient retrieval, multimedia search techniques to improve the current to fill the current network and three-dimensional graphics engine search blank, promote the realization of next-generation intelligent multi-modal search engines.

[0005] 本发明提供的技术方案如下: [0005] The present invention provides the following technical solution:

[0006] 本发明提供一种图形检索方法,包括: [0006] The present invention provides a graphic search method, comprising:

[0007] 1)建立三维网格模型库; [0007] 1) establishing a three-dimensional mesh model library;

[0008] 2)当用户输入的是二维图形或图像时,与三维网格模型库中的三维网格模型的轮廓进行匹配,根据匹配参数将三维网格模型投影到二维空间,得到投影的二维图像或图形, 然后计算投影得到的二维图像或图形与输入的图形或图像之间的相关度,根据相关度检索得到三维网格模型; [0008] 2) When the user inputs the two-dimensional pattern or image, to match the contour of the 3D mesh model 3D mesh model library, the matching parameter 3D mesh model is projected onto a two-dimensional space, obtained projection two-dimensional images or graphics, and then calculate the correlation between two-dimensional image or a graphic image or graphic input with a projection obtained, to obtain a three-dimensional mesh model according to the correlation retrieval;

[0009] 3)当用户输入的是三维网格模型时,对输入的三维网格模型进行骨架提取,根据提取的三维网格模型骨架,在三维网格模型库中初步检索得到三维网格模型; [0009] 3) When the user inputs a three-dimensional mesh model, the three-dimensional mesh skeleton extraction inputted, based on the extracted three-dimensional mesh skeleton, a three-dimensional mesh model library retrieved preliminary 3D mesh model ;

[0010] 4)将检索得到的三维网格模型和用户输入的三维网格模型进行特征点提取,代替原始三维网格模型,再进行三角剖分,对剖分后的分割线进行分段拟合,得到原始三维网格模型的控制点,然后根据拓扑结构对控制点进行频域变换; [0010] 4) The three-dimensional mesh model retrieved and the user input 3D mesh model feature point extraction, instead of the original three-dimensional mesh model, then triangulation, the division line of the split segmented Quasi together, to give the control point of the original 3D mesh model, then the control point according to a frequency domain conversion topology;

[0011] 5)计算得到的用户输入三维网格模型的控制点频域坐标值与三维网格模型库中的三维网格模型的控制点频域坐标值之间的相似度,根据相似度检索出对应的图形; [0011] 5) the calculated control point user input 3D mesh model in the frequency domain coordinate values ​​of control points of the three-dimensional mesh model 3D mesh model library degree of similarity between the frequency domain coordinate values, according to the similarity search the corresponding pattern;

[0012] 其中,步骤4)中根据三维网格模型的拓扑结构对得到的控制点进行频域变换,包括:[0013] (1)以控制点到网格中心的矢量的模对粗糙网格的控制点进行排序; [0012] wherein in step 4) frequency domain transform on the control points obtained in accordance with the topology of a three-dimensional mesh model, comprising: [0013] (1) to the control point vector to the center of the grid meshes of rough mold sorting control point;

[0014] (2)从网格拓扑关系获得Kirchhoff矩阵 [0014] (2) a matrix obtained from the Kirchhoff mesh topology

[0015] K = DA (6) [0015] K = DA (6)

[0016] D是对角矩阵,其对角线上的元素Dii与顶点vi的价相对应,A是网格的邻接矩阵; [0016] D is a diagonal matrix whose elements on the diagonal vertex vi Dii corresponding valence, A is the adjacency matrix of the grid;

[0017] 对KirchhofT矩阵进行特征值分解得到的η个特征向量Wi进行升序排列,组成的η*η映射矩阵W ; [0017] The eigenvalue decomposition KirchhofT matrix obtained eigenvectors Wi [eta] in ascending order, consisting of η * η mapping matrix W is;

[0018] (3)从先前排好序的η个控制点的空间坐标构造3个向量: [0018] (3) constructed from three space coordinates vector previously sorted η control points:

[0019] X = (X1, X2, —, Xn), Y = (y1? I2,…,yn),Z = (Z1, z2,…,zn) (8) [0019] X = (X1, X2, -, Xn), Y = (? Y1 I2, ..., yn), Z = (Z1, z2, ..., zn) (8)

[0020] 将这3个向量投影到特征向量基W上得到频域向量: .Xs = WX [0020] These three vectors projected to obtain a frequency domain vector based on the feature vector W: .Xs = WX

[0021] [0021]

Figure CN101719140BD00051

[0022] 每个顶点对应的频谱的幅值Si计算公式为: [0022] each vertex corresponding to a magnitude spectrum Si is calculated as:

[0023] [0023]

Figure CN101719140BD00052

[0024] 优选的,步骤1)中的三维网格模型库是对各种三维数据格式进行重新组织和多边形建模后得到的。 [0024] Preferably, the three-dimensional mesh model library in step 1) is performed after the reorganization polygonal modeling and three-dimensional data of various formats obtained.

[0025] 优选的,骨架提取过程为:首先为输入的网格模型建立渐进网格表示,然后对渐进网格不断的进行边塌缩变换,在塌缩的过程中如果一条边没有相邻三角形,则该条边标记为骨架边,并且一直保留到塌缩结束,最终获得的边构成模型的骨架。 [0025] Preferably, the skeleton extraction process is as follows: Firstly, progressive mesh model representation of the input grid, and then continue to carry out progressive mesh edge collapse transformation, if an edge is not adjacent triangles in the process of collapse , the article side edges is marked as a skeleton, and kept until the end of the collapse, the obtained final edge of a skeletal model.

[0026] 优选的,步骤4)中对于检索得到的三维网格模型和用户输入的三维网格模型,根据其空间形状进行特征点提取,用脐带点作为特征点代替原始三维网格模型。 [0026] Preferably, step 4) and three-dimensional mesh model of the 3D mesh model retrieved user input, extracting feature points in accordance with its spatial shape, instead of the original 3D mesh model with umbilical point as a feature point.

[0027] 优选的,步骤4)中根据特征点对三维网格模型进行三角剖分,对剖分后的分割线进行分段拟合,得到分割点,将这些分割点作为原始三维网格模型的控制点。 [0027] Preferably, the 4) step according to the three-dimensional mesh model characteristic point triangulation, the division line of the split segmented obtained by fitting the split point, the division point such as an original 3D mesh model the control points.

[0028] 本发明具有以下有益效果: [0028] The present invention has the following advantages:

[0029] (1)可以根据单个二维图像检索三维图形,用户可以输入bmp、jpeg、tiff等常见格式的图片,通过本发明方法可以检索出对应的图形。 [0029] (1) three-dimensional graphics can be retrieved, the user can enter the common image format bmp, jpeg, tiff, etc. from a single two-dimensional image, by the process of the present invention can retrieve the corresponding pattern.

[0030] (2)支持输入的三维模型的表示方法更加广泛,对于三角形网格数据、点云数据、 体数据或多边形网格模型均适用。 Three-dimensional model representation [0030] (2) supports a wider input for the triangular mesh data, point cloud data, the volume data or polygonal mesh models are applicable.

[0031] (3)采用了分级搜索,更快更准确。 [0031] (3) using a hierarchical search, faster and more accurate. 首先根据三维图形的骨架进行粗糙搜索,对搜索结果中的三维图形采用其它特征提取技术进行准确搜索,保证了实时性和准确性。 First, according to three-dimensional graphics skeleton rough search, the search results for the three-dimensional graphics that other feature extraction accurately searched, ensure real-time and accuracy.

附图说明 BRIEF DESCRIPTION

[0032] 为了更清楚地说明本发明实施例或现有技术中的技术方案,下面将对实施例或现有技术描述中所需要使用的附图作简单地介绍,显而易见地,下面描述中的附图仅仅是本发明的一些实施例,对于本领域普通技术人员来讲,在不付出创造性劳动的前提下,还可以根据这些附图获得其它的附图。 [0032] In order to more clearly illustrate the technical solutions in the embodiments or the prior art embodiment of the present invention, briefly introduced hereinafter, embodiments are described below in the accompanying drawings or described in the prior art needed to be used in describing the embodiments the drawings are only some embodiments of the present invention, those of ordinary skill in the art is concerned, without creative efforts, can derive from these drawings other drawings.

[0033] 图1是用户输入二维图形或图像的处理示意图;[0034] 图2是三维与二维的匹配流程图; [0033] FIG. 1 is a schematic two-dimensional pattern or image processing user input; [0034] FIG 2 is a flowchart illustrating the matching of 3D and 2D;

[0035] 图3是对于用户输入三维网格数据时的处理流程图。 [0035] FIG. 3 is a three-dimensional network for user input of data processing flowchart.

具体实施方式 detailed description

[0036] 下面将结合本发明实施例中的附图,对本发明实施例中的技术方案进行清楚、完整地描述,显然,所描述的实施例仅仅是本发明一部分实施例,而不是全部的实施例。 [0036] below in conjunction with the present invention in the accompanying drawings, technical solutions of embodiments of the present invention are clearly and completely described, obviously, the described embodiments are merely part of embodiments of the present invention, but not all embodiments example. 基于本发明中的实施例,本领域普通技术人员在没有作出创造性劳动前提下所获得的所有其它实施例,都属于本发明保护的范围。 Based on the embodiments of the present invention, those of ordinary skill in the art without making all of the other embodiments given herein without creative efforts fall within the scope of the present invention.

[0037] 本发明方法主要研究图形的多模态检索,其原理是根据三维图形的特征,通过二维图像、二维图形、三维模型等信息,计算用户输入的图形或图像与三维图形之间的相似度,从而实现三维图形多模态检索。 [0037] Multi-modal searching method according to the present invention, the main pattern, the principle is based on the characteristics of the three-dimensional graphics, the two-dimensional image, two-dimensional graphics, three-dimensional model information, and the calculation between images or three-dimensional graphic pattern input by the user similarity, thereby achieving a three-dimensional graphics multimodal retrieval. 这里多模态是指支持用户以常见格式的二维图像、二维图形和三维模型进行查询。 Here it refers to support multi-modal user queries two-dimensional image, two-dimensional graphics and three-dimensional model of a common format.

[0038] 本发明方面从三维数据中提取出很小的数据量,并将其作为对应图形或图像的主要特征,可以根据该主要特征进行检索。 [0038] aspect of the present invention, the three-dimensional data extracted from a small amount of data, as the main feature and the corresponding graphics or images, can be searched in accordance with the primary feature. 该特征基本不受噪声、相似变换、不同分辨率采样等因素的影响。 This feature substantially unaffected by noise, similarity transformation, the effects of different factors such as the sampling resolution.

[0039] 本发明方法主要包括以下8个环节: [0039] The method of the present invention mainly comprises the following eight aspects:

[0040] (1)首先对各种数据格式的三维模型数据进行重新组织,进行多边形建模,得到三维网格模型,并建立三维网格模型库。 [0040] (1) First, the three-dimensional model data of various data formats reorganized, polygon modeling, to obtain a three-dimensional mesh model, and dimensional mesh model library.

[0041] (2)当用户输入的是二维图形或图像时,与三维网格模型库中的三维网格模型的轮廓进行匹配,然后根据匹配参数将三维网格模型投影到二维空间,得到投影的二维图像或图形,然后计算投影得到的二维图像或图形与输入的图形或图像之间的相关度,根据相关度检索出三维网格模型。 [0041] (2) When the user inputs the two-dimensional pattern or image, to match the contour of the 3D mesh model 3D mesh model library, then according to the matching 3D mesh model parameter to the projected two-dimensional space, to obtain a two-dimensional projection images or graphics, and then calculating the degree of correlation between the two-dimensional image or a graphic pattern or image obtained by projecting the input, a three-dimensional mesh model according to the retrieved affinity.

[0042] (3)当用户输入的是三维网格模型时,对输入的三维网格模型进行骨架提取。 [0042] (3) When the user inputs a three-dimensional mesh model, the three-dimensional mesh skeleton extraction inputted.

[0043] (4)根据提取的三维网格模型骨架,在三维网格模型库中进行快速的初步检索,得到初步检索出的三维网格模型。 [0043] (4) based on the extracted three-dimensional mesh skeleton, quick retrieval of the initial three-dimensional network model library to obtain a preliminary model of the three-dimensional mesh retrieved.

[0044] (5)根据图形学理论将初步检索出的三维网格模型和用户输入的三维网格模型, 根据其空间形状进行特征点提取,代替原始三维网格模型,大大减少数据量。 [0044] (5) The graphics theory retrieved preliminary 3D mesh model and a three-dimensional mesh model of the user's input, according to which the spatial shape of the feature point extraction, instead of the original 3D mesh model, significantly reduce the amount of data.

[0045] (6)根据提取的特征点对三维网格模型进行三角剖分,对剖分后的分割线进行分段拟合,得到分割点,将这些分割点作为原始三维网格模型的控制点。 [0045] (6) according to the extracted feature points to triangulate three-dimensional mesh, the division line of the split segmented obtained by fitting the split point, the division point such as the control of the original three-dimensional mesh model point.

[0046] (7)根据三维网格模型的拓扑结构对控制点进行频域变换,得到控制点的频域坐标值。 [0046] (7) a three-dimensional mesh model of the topology of the control points according to the frequency domain transformation, the frequency domain to obtain the coordinate values ​​of control points.

[0047] (8)计算得到的用户输入三维网格模型的控制点频域坐标值与三维网格模型库中的三维网格模型的控制点频域坐标值之间的相似度,根据相似度检索出对应的图形。 [0047] The similarity between the frequency domain coordinate values ​​(8) calculated control point user input 3D mesh model in the frequency domain coordinate values ​​of control points of the three-dimensional mesh model 3D mesh model library, according to the similarity retrieve the corresponding graphic.

[0048] 本发明的技术特点主要体现如下: [0048] Technical features of the present invention are mainly as follows:

[0049] (1)系统可以根据单个二维图像检索三维图形。 [0049] (1) The three-dimensional graphics system may retrieve a single two-dimensional image. 用户可以输入bmp、jpeg、tiff等常见格式的图片,系统将从图片中分割对象并提取轮廓信息,与三维网格模型库中的三维图形进行匹配,向用户返回匹配的三维网格模型。 Users can enter a common picture formats bmp, jpeg, tiff, etc., the system will split the picture and extract the object contour information, matching 3D mesh models with three-dimensional graphics library, 3D mesh model is returned to the user match.

[0050] (2)系统支持输入的三维模型的表示方法更加广泛。 [0050] representation of the three-dimensional model input support (2) is more extensive system. 三角形网格数据、点云数据、 体数据或多边形网格模型均适用。 Triangular mesh data, point cloud data, the volume data or polygonal mesh models are applicable. [0051] (3)系统采用分级搜索。 [0051] (3) a hierarchical search system. 首先根据三维图形的骨架进行粗糙搜索,对搜索结果中的三维图形再采用特征提取技术进行准确搜索,保证了实时性和准确性。 First, according to three-dimensional graphics skeleton rough search, the search results for the three-dimensional graphics feature extraction and then using accurately searched, ensure real-time and accuracy.

[0052] 下面对本发明做进一步详细说明。 [0052] The following further detailed description of the present invention.

[0053] 本发明支持多模态的图形检索方法主要步骤包括: [0053] The method of the present invention supports retrieval pattern multimodal main steps comprising:

[0054] (1)进行多边形建模,预先建立三维网格模型库; [0054] (1) in polygonal modeling, three-dimensional mesh model library established in advance;

[0055] (2)三维网格模型与二维图像或图形的匹配; [0055] (2) matching 3D mesh model or pattern of two-dimensional image;

[0056] (3)三维网格模型骨架的提取; [0056] (3) extracting the 3D mesh model skeleton;

[0057] (4)根据骨架进行三维模型检索; [0057] (4) The three-dimensional model retrieval skeleton;

[0058] (5)三维网格模型特征点提取; [0058] (5) 3D mesh model feature point extraction;

[0059] (6)三维网格模型控制点的计算; [0059] calculation point (6) Control 3D mesh model;

[0060] (7)三维网格模型控制点频谱的计算; [0060] (7) a three-dimensional mesh model spectrum calculated control points;

[0061] (8)计算频谱相似度,根据相似度检索出对应的图形。 [0061] (8) spectrum is calculated similarity, similarity search in accordance with the corresponding pattern.

[0062] 以下分别对上述步骤进行详细介绍。 [0062] The following steps are described in detail above.

[0063] (1)进行多边形建模,预先建立三维网格模型库。 [0063] (1) in polygonal modeling, three-dimensional mesh model of a pre-established database.

[0064] 本发明根据三维网格模型数据,进行多边形建模,构造多边形三维网格模型。 [0064] According to the present invention, the three-dimensional mesh model data, the polygon modeling, three-dimensional mesh model of a polygonal configuration. 多边形建模是利用许多的多边形模拟曲面进行,多边形越多,则模型越逼近真实曲面。 Polygon modeling is the use of a number of polygons simulated curved surfaces, the more polygons, the model is more close to the real surface. 多边形建模是最广泛又易于实现的一种建模技术,并可以获得高精度的模型,通常采用三角形网格形式。 Polygon modeling is the most extensive and readily implemented in a modeling technique and model can be obtained with high precision, usually in the form of a triangular lattice.

[0065] 为了下述描述的方便和统一,利用数学符号给出三维网格模型的定义。 [0065] For convenience of the following description and unity, using a mathematical notation defines dimensional mesh model is given. 网格模型M = {V, C},由顶点集合V和连接关系集合C组成,其中集合V包含N个顶点Vi且每个顶点的坐标值由(Xi^yijZi)确定,即 Mesh M = {V, C}, the set of vertices V and connection relationship set C, where the set of vertices V contains N and each coordinate value of the vertex Vi is determined by (Xi ^ yijZi), i.e.

[0066] V = {vj , i = 0,1, Nl, Vi = (xi? yi; Zi) (1) [0066] V = {vj, i = 0,1, Nl, Vi = (xi yi;? Zi) (1)

[0067] 而连接关系集合C表示成 [0067] C is connected to a set of relationships expressed as

[0068] C= {{ik, jk}}k = 0, ...„,-!,0^ Nl,O^ Jk^NI (2) [0068] C = {{ik, jk}} k = 0, ... ", - !, 0 ^ Nl, O ^ Jk ^ NI (2)

[0069] 这里{ik,jk}表示由第ik个顶点和第jk个顶点确定的第k条边。 [0069] Here {ik, jk} denotes the k-th edge ik determined by the first vertex and the second vertex jk.

[0070] (2)三维网格模型与二维图像或图形的匹配 [0070] (2) a three-dimensional mesh model of two-dimensional image or pattern matching

[0071] 图1是用户输入二维图形或图像的处理示意图。 [0071] FIG. 1 is a schematic two-dimensional pattern or image processing user input.

[0072] 如图1所示,对于单个图像/图形,与模型库图形进行轮廓匹配,再进行2D投影, 然后进行图形匹配相关度计算。 [0072] As shown in FIG. 1, for a single image / graphics, with the model contour matching library pattern, then the 2D projection, then pattern matching correlation computation.

[0073] 在很多情况下,用户的检索输入为二维图像或图形,与三维网格模型的匹配具体过程如图2所示。 [0073] In many cases, the user's search input two-dimensional image or pattern, matching the three-dimensional mesh model of the specific procedure shown in FIG. 首先,本发明采用图像的轮廓(或图形轮廓)与三维网格模型库中的三维网格模型轮廓先进行初始匹配;轮廓匹配后,将得到的三维网格模型根据匹配参数投影到2D空间,得到投影图像,然后用图像相关匹配方法进行进一步精确匹配。 First, the present invention uses contour (outline or graphic) image and the 3D mesh model 3D mesh model library first initial contour matching; after contour matching, the resulting 3D mesh model is projected onto a 2D space as matching parameters, to obtain a projection image, and then further refined correlation image matching method of matching.

[0074] 三维网格模型库中的所有网格模型已完成X、Y、Z三个正方向的轮廓提取和投影计算。 [0074] All grid model 3D mesh model library has been completed X, Y, Z three directions of the positive contour extraction and projection calculation.

[0075] 对于三维网格模型,本发明方法是遍历网格中的每一个边来提取的轮廓,具体方法如下: [0075] For the 3D mesh model, the method of the present invention is that each side to traverse extracted contour grid, as follows:

[0076] 1.如果当前边仅与一个三角形相连接,那么它属于轮廓; [0076] 1. If the current edge is connected only to a triangle, it belongs to the contour;

[0077] 2.如果当前边与两个三角形F1和F2,则定义其法向量分别为;;,和&,当前镜头位置与当前边的一个顶点之间的向量为P。 [0077] 2. If the current edge F1 and F2 of two triangles, which are defined respectively ;; normal vector, and &, and the current lens position vector between a side of the current vertex is P. 如果(ϋ)·(ϋ)<0,即&和&相对于镜头的轴 If (ϋ) · (ϋ) <0, i.e., & and & axis relative to the lens

处于不同的方向,说明F1和F2 —个正对着镜头一个背对着镜头,因此当前边为轮廓边,否则,当前边不是显著的轮廓边。 In different directions, instructions F1 and F2 - a positive front of the camera back to the camera, so the current side is the side profile, otherwise, the current edge is not significant contour edges.

[0078] 对于图形,不需要轮廓提取过程。 [0078] For graphics, the contour extraction process not required.

[0079] 对于图像,可以采用Sobel、Prewitt等算子提取轮廓。 [0079] For the image, may be employed Sobel, Prewitt operator, etc. contour extraction.

[0080] 本发明轮廓匹配过程可采用基于对应形状匹配方法,如Hausdoff距离;或者多边形分解匹配方法,如进行线段化处理,并且以轮廓重心为中心进行三角剖分,确定边界顶点,连接起来便成一凸多边形,用这一凸多边形近似表示原图形,将多边形重心与顶点连线,组成一系列三角形从而进行图形轮廓匹配。 [0080] The process of the present invention may match the contour of a corresponding shape-based matching method, such as the distance Hausdoff; or a polygon match decomposition method, such as line segment processing, and contour to the center of gravity of triangulation, to determine the boundary vertex, they are connected into a convex polygon, with the original pattern represented by a convex polygon approximation, the center of gravity of the polygon vertex lines to form a series of triangles to perform pattern matching profile. 轮廓匹配过程简单高效,可快速实现初始检索。 Contour matching process is simple and efficient, quick implementation of the initial crawl.

[0081] 本发明图像相关匹配方法是计算三维网格模型投影得到的二维图像或图形与输入的图形或图像之间的相关度,如利用归一化相关测度计算图像区域中每一对像素的相似性。 [0081] The image matching method of the present invention is related to the degree of correlation between the two-dimensional image or a graphic image or graphic input with the calculated projection 3D mesh model obtained, as calculated using the normalized correlation measures the image area of ​​each pixel the similarity. 对于待匹配的两幅图像I1(Ly)和I2(x,y),待检测图位置(i,j)上归-义为: For two images I1 (Ly) it is matched and I2 (x, y), the return to be detected FIG position (i, j) - is defined as:

[0082] [0082]

Figure CN101719140BD00081

[0083] 当相关度大于一个设定阈值,即检索到匹配的三维网格模型。 [0083] When the correlation is greater than a set threshold, i.e. the retrieved matching 3D mesh model.

[0084] (3)三维网格模型骨架的提取 [0084] (3) extracting the 3D mesh model skeleton

[0085] 骨架是一种性质优良的图形几何特征,又称中轴(Medial Axis),是一种有效的图形描述手段。 [0085] The skeleton is excellent in a property of graphic geometric characteristics, also known as the axis (Medial Axis), a graphical depiction is an effective means. 顾名思义,骨架是一种线型的几何体,居于图形的对称中心,有着与原图形相同的拓扑结构,并保留着原图形的形状信息。 As the name suggests, it is a linear backbone geometry, the center of symmetry living pattern, the original pattern has the same topology, and retains the shape information of the original pattern.

[0086] 图3是对于用户输入三维网格数据时的处理流程图。 [0086] FIG. 3 is a three-dimensional network for user input of data processing flowchart.

[0087] 图3的过程包括以下的过程,即(¾三维网格模型骨架的提取;(4)根据骨架进行三维模型检索;(¾三维网格模型特征点提取;(6)三维网格模型控制点的计算;(7)三维网格模型控制点频谱的计算;(8)计算频谱相似度,根据相似度检索出对应的图形。具体内容参加下面的描述。 Process [0087] FIG. 3 includes a process, i.e., (¾ extract 3D mesh model skeleton; (4) The three-dimensional model retrieval skeleton; (¾ 3D mesh model feature point extraction; (6) 3D mesh model calculating a control point; (7) a three-dimensional mesh model spectrum calculated control points; (8) spectrum is calculated similarity, a similarity search in accordance with a pattern corresponding to the specific content to participate in the following description.

[0088] 本发明设计了一种快速的基于多分辨率网格的骨架提取算法:首先为三维网格模型建立渐进网格表示,然后对渐进网格不断的进行边塌缩变换,在塌缩的过程中如果一条边没有相邻三角形,那么这条边标记为骨架边,并且一直保留到塌缩结束,最终获得的边就构成了网格模型的骨架。 [0088] The present invention is designed a fast algorithm for extracting the skeleton of a multi-resolution grid: Firstly progressive mesh model representation of the three-dimensional mesh, and then continue to carry out progressive mesh edge collapse transformation, the collapse If the process is not adjacent one side of the triangle, then this skeleton side edges marked, and kept until the end of the collapse, while the finally obtained constitutes a skeletal grid model. 三维网格模型库中的所有网格模型已完成骨架提取计算。 All mesh 3D mesh model library skeleton extraction has finished calculation. 用户输入网格模型时需要提取骨架。 When the user need to extract skeleton model input mesh.

[0089] (4)根据骨架进行三维模型检索 [0089] (4) The three-dimensional model retrieval backbone

[0090] 提取骨架是将3D图形转换为3D线段的过程,3D线段的数据量相对于原始3D图形大大减少,因此可以加快检索速度。 [0090] Extraction skeleton is converted to 3D graphics 3D segment process, the amount of data of the 3D line segment relative to the original 3D graphics greatly reduced, it is possible to accelerate the retrieval speed.

[0091] 用户输入的三维网格模型与库中的网格模型进行骨架比较,比较骨架采用主成分PCA分析法,先将骨架定位,分段计算骨架的距离,然后直接采用欧氏距离进行排序,根据距离信息比较结果,欧氏距离差最小的即为初步检索的三维网格模型。 [0091] 3D mesh model with the mesh model library user inputted comparison skeleton, skeleton comparison using principal component analysis PCA, positioning first skeleton, the skeleton of the segment to calculate the distance, Euclidean distance is then used directly to sort the information from the comparison result, the minimum Euclidean distance difference is the initial 3D mesh model retrieval. [0092] (5)三维网格模型特征点提取 [0092] (5) 3D mesh model feature point extraction

[0093] 根据图形学理论将用户输入的三维网格模型,根据其空间形状进行特征点提取, 代替原始三维网格模型,大大减少数据量。 [0093] The theory of the graphics 3D mesh model input by the user, depending on its spatial shape feature point extraction, instead of the original 3D mesh model, significantly reduce the amount of data. 三维网格模型库中的所有网格模型已完成特征点提取、控制点频谱计算。 All mesh 3D mesh model library feature point extraction has been completed, the control point spectrum calculation.

[0094] 本发明使用脐带点作为任意三角形三维网格模型的特征点。 [0094] Using as an arbitrary point of a triangle cord 3D mesh model of the features of the present invention. 以脐带点作为三维网格模型表面的特征点,对于噪声、剪切、旋转、平移、缩放、不同分辨率采样等由于不同采集设备造成的影响具有很强的鲁棒性。 Umbilical point to a three-dimensional mesh model of the surface feature points, the noise, shearing, rotation, translation, scaling, effects due to different acquisition devices caused by sampling different resolutions, which are highly robust. 由于跨过网格模型的边的曲率较大,所以曲率张量可以定义为网格模型的边的每一个点。 Due to the larger side of the curvature across the grid model, the curvature tensor can be defined for each point edge of the grid model. 在任意网格区域B内,定点ν的曲率张量可以用下式估计: In an arbitrary grid areas B, point ν curvature tensor can be estimated by the following formula:

[0095] [0095]

Figure CN101719140BD00091

[0096] 其中,|B|是ν的邻域的面积,β (e)是边e的两个邻接三角形的法向量之间的夹角,|e η B是区域B中的边e的长度J是沿着边e的单位法向量。 [0096] where, | B | ν is the area of ​​the neighborhood, β (e) is the angle between the normal vectors of two adjacent triangles of edge e, | e η B is the length of a side of the region B e J along the edge of the unit normal vector e. ν的邻接区域B是由以ν为球心,以r为半径的球体与网格模型的相交圆定义的。 ν B is a region adjacent to ν center of the sphere, a sphere of radius r intersects with the circle defined by the grid model. 半径r是指定曲率估计的尺度参数。 R is the radius of curvature designated scale parameter estimation.

[0097] 得到每一个顶点的曲率张量以后,在每个三角形网格上进行线性插值以获得连续的曲率张量场。 After [0097] Each of the obtained vertex curvature tensor, linear interpolation is performed on each of the triangular mesh to obtain a continuous curvature tensor field. 顶点的法向量方向与曲率张量的幅值最小的特征值相对应,其它两个特征值分别对应顶点ν的最小曲率和最大曲率,当这两个特征值相等时,顶点ν被称为脐带点, 即本发明中使用的三维网格模型的特征点。 The magnitude of the smallest eigenvalue is the vector direction of curvature tensor method corresponding to the apex, the other two eigenvalues ​​correspond to the maximum curvature and the minimum curvature at the vertex ν when these two characteristic values ​​are equal, the umbilical cord is called the vertex ν point, i.e., feature points of the 3D mesh model used in the present invention.

[0098] 显然曲率估计的关键是尺度参数。 [0098] Obviously the key to the curvature of the scale parameter is estimated. 本发明采用不同的尺度进行曲率张量估计,以便于平滑张量场并且估计不同尺度下的脐带点对于噪声以及仿射变换等因素的鲁棒性。 The present invention is carried out using different scales curvature tensor estimation for smoothing tensor field and estimates the umbilical point at different scales robustness with respect to noise and other factors of affine transformation. 为了避免丢失网格模型表面的局部信息以及减小计算复杂度,尺度参数不能取得过大。 To avoid losing information of the local mesh model of the surface and to reduce the computational complexity, the scale parameter can not be made too large. 本发明在寻找鲁棒性最高的尺度参数时,选用自适应遗传算法作为优化搜索技术。 The present invention is in the search for the most robust scale parameter, adaptive genetic algorithm as the choice of optimization search technique. 在鲁棒性最高的几个尺度参数中,我们选用使区域B内的平均曲率最大的点,即曲率张量的迹最大的点ο In several highest robustness scale parameter, we use so that the point of maximum average curvature in the region B, that is the point of maximum curvature of the trace of the tensor ο

[0099] (6)三维网格模型控制点的计算 [0099] (6) Calculation of the 3D mesh model control point

[0100] 根据提取的特征点对三维网格模型进行三角剖分,对剖分后的分割线进行分段拟合,得到分割点,将这些分割点作为原始三维网格模型的控制点。 [0100] The feature points extracted for the three-dimensional mesh model triangulation, the split line of the split segmented obtained by fitting the split point, the division point such as control points of the original 3D mesh model.

[0101] 获得特征点以后,下一步任务是减小数据量,这对于特征注册以及三维网格模型的检索都非常重要。 [0101] After the feature points are obtained, the next task is to reduce the amount of data, which is important for registration and retrieval characteristics of the 3D mesh model.

[0102] 获得特征点以后,第一步是将网格进行三角分割。 [0102] After the feature points are obtained, the first step is the triangular mesh is divided. 在2D空间中,Delaimay三角分割能够产生形状较均勻的三角形,并且具有唯一性。 In the 2D space, Delaimay triangular division can generate a relatively uniform shape of triangles, and unique. 在3D表面中,Delaimay三角分割不是使用欧氏距离而是使用测地距离(测地距离是指3D空间中网格表面的两个顶点沿着表面的最短距离)。 In 3D surface, Delaimay triangular segmentation is not used but the use of Euclidean distance measuring distance (geodesic distance means the shortest distance between two vertices of the mesh surface in 3D space along the surface). 本发明采用波阵面方法对整个Voronoi图及其二重Delaimay三角分割进行估计。 The present invention is a method of using wavefront entire Voronoi diagram and triangular double Delaimay segmentation estimate. 采用该方法的好处是能够使三角分割不受采样率的大小的影响。 The advantage of using this method is the ability to split the triangle is not affected by the size of the sample rate. 波阵面是通过计算以种子点为球心,半径不断增长的球与三维网格表面的相交圆获得的。 Wavefront by the seed point calculated in center of the sphere, the radius of the circle intersects the growing surface of the ball and the three-dimensional network obtained. 采用基于不同的球半径的波阵面方法比采用基于网格拓扑关系的波阵面方法得到的结果要好。 Based on the results using the wavefront better topological relationship based on a grid obtained wavefront method different than using radius of the sphere.

[0103] (7)三维网格模型控制点频谱的计算 [0103] (7) a three-dimensional mesh model spectra calculated control points

[0104] 根据三维网格模型的拓扑结构对控制点进行频域变换,得到控制点的频域坐标值。 [0104] 3D mesh model of the topology of the control points according to the frequency domain transformation, the frequency domain to obtain the coordinate values ​​of control points.

[0105] 由于3D网格是图形而非图像,所以每个顶点坐标没有其固有的影像函数,因此对3D网格进行变换域处理首先要构造一个影像函数,以便于将经典的变换域处理算法应用于3D网格处理。 [0105] Since a graphics 3D grid instead of the image, the coordinates of each vertex function without its inherent video, 3D mesh is therefore transform domain to construct a first image processing function, to facilitate processing algorithms transform domain classic 3D mesh processing is applied.

[0106] 首先,对控制点进行排序。 [0106] First, control points are sorted. 本发明以控制点到网格中心的矢量的模对粗糙网格的控制点进行排序。 The present invention is a control point to control point vector to the center of the grid to the coarse mesh die sort.

[0107]定义网格中心的坐标为 Defined by the coordinates of the center of the grid [0107] of

[0108] [0108]

Figure CN101719140BD00101

[0109] 每个控制点到网格中心的矢量的模定义如下 [0109] each control point vector to the center of the grid mode defined below

[0110] [0110]

Figure CN101719140BD00102

[0111] 将控制点根据模大小进行升序排列以后,开始进行频域变换。 [0111] The control point after the die size in ascending order, the frequency domain transformation started. 首先从网格拓扑关系中获得组合拉普拉斯算子或者叫做Kirchhoff矩阵。 First, obtain a combination of Laplacian or called matrix from Kirchhoff mesh topology relationship. 该矩阵定义如下: This matrix is ​​defined as follows:

[0112] K = DA (6) [0112] K = DA (6)

[0113] 其中,D是对角矩阵,其对角线上的元素Dii与顶点Vi的价相对应(价即是从顶点放射出的边的个数),A是网格的邻接矩阵,其元素定义如下: [0113] where, D is a diagonal matrix whose vertex Vi Dii tetravalent element on a diagonal line corresponding to (the number of edges is a monovalent i.e. radiated from the vertex), A is the adjacency matrix grid which elements are defined as follows:

[0114] [0114]

Figure CN101719140BD00103

[0115] 对于有η个控制点的网格模型,矩阵A、D和K的尺寸均为η*η。 [0115] [eta] For mesh of control points, the matrix A, D and K are the sizes η * η. 对Kirchhoff矩阵进行特征值分解得到η个特征值λ i和η个特征向量Wi。 Kirchhoff of matrix eigenvalue decomposition η eigenvalues ​​λ i and η eigenvectors Wi. 将这η个特征向量进行升序排列,可以得到其对应的特征向量,这些排序后的特征向量是频率不断增大的函数基。 Η these ascending order eigenvectors can be obtained corresponding eigenvectors, eigenvectors are ordered after those increasing function of the frequency group. 该函数基仅取决于网格模型的拓扑结构,而与网格模型的几何特性无关。 This function depends only on the topology of the base mesh model, irrespective of the geometric characteristics of the grid model. 将这η个排序后的特征向量组成的η*η映射矩阵记做W。 [eta] [eta] After these feature vector consisting of * η sort denoted mapping matrix W.

[0116] 从先前排好序的η个控制点的空间坐标构造3个向量: [0116] 3 is configured from the space vector coordinate previously sorted η control points:

[0117] X = (X1, χ2, ···, xn), Y = (y1? y2, ..., yn), Z = (Z1, z2,…,zn) (8) [0117] X = (X1, χ2, ···, xn), Y = (y1? Y2, ..., yn), Z = (Z1, z2, ..., zn) (8)

[0118] 将这3个向量投影到特征向量基W上,即可得到空域坐标的频域分解向量: [0118] These three vectors projected onto the feature vector group W, to obtain the spatial frequency domain decomposition vector coordinates:

[0119] [0119]

Figure CN101719140BD00104

[0120] 空间坐标也可由频率坐标还原得到: [0120] spatial coordinates may also be reduced to give the frequency coordinates:

[0121] [0121]

Figure CN101719140BD00105

[0122] 每个顶点对应的频谱的幅值Si可以由下式计算: [0122] Each amplitude spectrum corresponding to the vertex Si may be calculated by the following formula:

[0123] [0123]

Figure CN101719140BD00106

[0124] (8)计算频谱相似度,根据相似度检索出对应的图形。 [0124] (8) spectrum is calculated similarity, similarity search in accordance with the corresponding pattern.

[0125] 计算用户输入的三维网格模型的控制点频域坐标值与库中的三维网格模型的控制点频域坐标值之间的相关度,根据相关度检索出对应的图形。 Correlation between the [0125] control point calculating 3D mesh model in the frequency domain the user input control point coordinate value and the three-dimensional mesh model library frequency domain coordinate values, according to the degree of correlation corresponding to the retrieved pattern.

[0126] 比较两个网格模型的相似度时,则通过比较频域系数的波形相似度来判断。 [0126] When the similarity comparison between two grid model, the similarity of the waveform judged by comparing the frequency domain coefficients. 比较相似度时的准则可采用归一化相关系数: Criteria may be employed when comparing the degree of similarity of the normalized correlation coefficient:

[0127] [0127]

Figure CN101719140BD00111

[0128] 其中歹为由用户输入的三维网格模型计算得到的频谱值,云为库中的三维网格频谱值。 [0128] wherein bad 3D mesh model by spectral values ​​calculated user input, three-dimensional network cloud spectral values ​​in the library. NC越大,则频谱系数越相关。 NC is increased, the spectral correlation coefficient.

[0129] 给定一个阈值T,如果NC >T,则认为用户输入的三维网格与库中的三维网格相匹配,检索流程结束。 [0129] Given a threshold T, if the NC> T, the three-dimensional network with a three-dimensional network that the user input database matches retrieval process ends.

[0130] 本发明具有以下有益效果: [0130] The present invention has the following advantages:

[0131] (1)可以根据单个二维图像检索三维图形,用户可以输入bmp、jpeg、tiff等常见格式的图片,通过本发明方法可以检索出对应的图形。 [0131] (1) three-dimensional graphics can be retrieved, the user can enter the common image format bmp, jpeg, tiff, etc. from a single two-dimensional image, by the process of the present invention can retrieve the corresponding pattern.

[0132] (2)支持输入的三维模型的表示方法更加广泛,对于三角形网格数据、点云数据、 体数据或多边形网格模型均适用。 Three-dimensional model representation [0132] (2) supports a wider input for the triangular mesh data, point cloud data, the volume data or polygonal mesh models are applicable.

[0133] (3)采用了分级搜索,更快更准确。 [0133] (3) using a hierarchical search, faster and more accurate. 首先根据三维图形的骨架进行粗糙搜索,对搜索结果中的三维图形采用其它特征提取技术进行准确搜索,保证了实时性和准确性。 First, according to three-dimensional graphics skeleton rough search, the search results for the three-dimensional graphics that other feature extraction accurately searched, ensure real-time and accuracy.

[0134] 本领域普通技术人员可以理解上述实施例的各种方法中的全部或部分步骤是可以通过程序来指令相关的硬件来完成,该程序可以存储于一计算机可读存储介质中,存储介质可以包括:只读存储器(ROM,Read Only Memory)、随机存取存储器(RAM,Random Access Memory)、磁盘或光盘等。 [0134] Those of ordinary skill in the art can appreciate that various embodiments of the method of the above-described embodiments all or part of the steps may be relevant hardware instructed by a program, the program may be stored in a computer-readable storage medium, the storage medium It may include: a read only memory (ROM, Read Only memory), a random access memory (RAM, random access memory), a magnetic disk or optical disk.

[0135] 以上对本发明实施例所提供的一种图形检索方法,进行了详细介绍,本文中应用了具体个例对本发明的原理及实施方式进行了阐述,以上实施例的说明只是用于帮助理解本发明的方法及其核心思想;同时,对于本领域的一般技术人员,依据本发明的思想,在具体实施方式及应用范围上均会有改变之处,综上所述,本说明书内容不应理解为对本发明的限制。 Illustrates a graphic search method embodiment provided in [0135] the above embodiment of the present invention, described in detail herein through specific examples of the principles and embodiments of the invention are set forth in the above embodiments are only used to help understand the method and the core idea of ​​the present invention; while those of ordinary skill in the art, according to the ideas of the present invention, there are modifications to the specific embodiments and application scope, summary, the specification shall not construed as limiting the present invention.

Claims (5)

1. 一种图形检索方法,其特征在于,包括:1)建立三维网格模型库;2)当用户输入的是二维图形或图像时,与三维网格模型库中的三维网格模型的轮廓进行匹配,根据匹配参数将三维网格模型投影到二维空间,得到投影的二维图像或图形,然后计算投影得到的二维图像或图形与输入的图形或图像之间的相关度,根据相关度检索得到三维网格模型;3)当用户输入的是三维网格模型时,对输入的三维网格模型进行骨架提取,根据提取的三维网格模型骨架,在三维网格模型库中初步检索得到三维网格模型;4)将检索得到的三维网格模型和用户输入的三维网格模型进行特征点提取,代替原始三维网格模型,再进行三角剖分,对剖分后的分割线进行分段拟合,得到原始三维网格模型的控制点,然后根据拓扑结构对控制点进行频域变换;5)计算得到的用户输入三维 A graphic search method, characterized by comprising: 1) establishing a three-dimensional mesh model library; 2) when the user input is a two-dimensional pattern or image, the 3D mesh model with 3D mesh model in the library contour matching, the matching parameter 3D mesh model is projected onto a two-dimensional space, to obtain a two-dimensional projection images or graphics, and then calculating the degree of correlation between the two-dimensional image or a graphic pattern or image obtained by projecting the input, in accordance with correlation retrieved 3D mesh model; 3) when the user inputs a three-dimensional mesh model, the three-dimensional mesh skeleton extraction inputted, based on the extracted three-dimensional mesh skeleton, the initial three-dimensional mesh model library 3D mesh model retrieved; 4) the three-dimensional mesh model retrieved and the user input 3D mesh model feature point extraction, instead of the original three-dimensional mesh model, then triangulation, the dividing line of the split segmented obtained by fitting the original control points of the 3D mesh model, then the frequency domain transformation of the control point based topology; 5) the calculated three-dimensional user input 网格模型的控制点频域坐标值与三维网格模型库中的三维网格模型的控制点频域坐标值之间的相似度,根据相似度检索出对应的图形;其中,步骤4)中根据三维网格模型的拓扑结构对得到的控制点进行频域变换,包括:(1)以控制点到网格中心的矢量的模对粗糙网格的控制点进行排序;(2)从网格拓扑关系获得Kirchhoff矩阵K = DA (6)D是对角矩阵,其对角线上的元素Dii与顶点vi的价相对应,A是网格的邻接矩阵;对Kirchhoff矩阵进行特征值分解得到的η个特征向量Wi进行升序排列,组成的η*η 映射矩阵W ;(3)从先前排好序的η个控制点的空间坐标构造3个向量:X = (χ1? χ2, ...,χη),Y = (y” y2,···» yn), Z = (z1? z2, zn) (8)将这3个向量投影到特征向量基W上得到频域向量:'Xs 二WX< Ys=WY (9)Λ署每个顶点对应的频谱的幅值Si计算公式为:《WW2+W2+IM (H)0 The control points of the grid model in the frequency domain coordinate values ​​of control points of the three-dimensional mesh model 3D mesh model library degree of similarity between the frequency domain coordinate value corresponding to the retrieved based on the similarity of the pattern; wherein, in step 4) according to the topology of a three-dimensional mesh model of the control points obtained frequency domain transform, comprising: (1) control point to the control point vector mode of the coarse mesh grid sorting center; (2) from the grid obtaining topology Kirchhoff matrix K = DA (6) D is a diagonal matrix whose elements on the diagonal vertex vi Dii corresponding valence, a is the adjacency matrix grid; Kirchhoff of matrix eigenvalue decomposition obtained [eta] Wi eigenvectors ascending order, η * η consisting mapping matrix W; (3) constructed from three space coordinates vector previously sorted control point [eta]: X = (χ1 χ2, ...,? χη), Y = (y "y2, ···» yn), Z = (z1 z2, zn) (8) these three vectors projected to obtain a frequency domain vector based on the feature vector W:? 'Xs two WX <Ys = WY (9) Λ Department each vertex corresponding to a magnitude spectrum Si calculated as: "WW2 + W2 + IM (H) 0
2.根据权利要求1所述的图形检索方法,其特征在于:步骤1)中的三维网格模型库是对各种三维数据格式进行重新组织和多边形建模后得到的。 2. The pattern retrieval method according to claim 1, wherein: the three-dimensional mesh model library in step 1) is performed after the reorganization polygonal modeling and three-dimensional data of various formats obtained.
3.根据权利要求1或2所述的图形检索方法,其特征在于:骨架提取过程为:首先为输入的网格模型建立渐进网格表示,然后对渐进网格不断的进行边塌缩变换,在塌缩的过程中如果一条边没有相邻三角形,则该条边标记为骨架边,并且一直保留到塌缩结束,最终获得的边构成模型的骨架。 The graphic search method of claim 1 or claim 2, wherein: the skeleton extraction process: First, to establish a progressive mesh mesh model representation of the input, and then continue to carry out progressive mesh edge collapse transformation, in the collapsing process if no adjacent one side of the triangle, the edge strip marked as the backbone side, and kept until the end of the collapse, the obtained final edge of a skeletal model.
4.根据权利要求1或2所述的图形检索方法,其特征在于:步骤4)中对于检索得到的三维网格模型和用户输入的三维网格模型,根据其空间形状进行特征点提取,用脐带点作为特征点代替原始三维网格模型。 The graphic search method of claim 1 or claim 2, wherein: in step 4) and three-dimensional mesh model of the 3D mesh model retrieved user input, extracting feature points in accordance with its spatial shape, with umbilical point as a feature point instead of the original 3D mesh model.
5.根据权利要求1或2所述的图形检索方法,其特征在于:步骤4)中根据特征点对三维网格模型进行三角剖分,对剖分后的分割线进行分段拟合,得到分割点,将这些分割点作为原始三维网格模型的控制点。 The graphic search method of claim 1 or claim 2, wherein: in step 4) the three-dimensional triangulated mesh model according to the feature points of the dividing lines for the split segment fitted, to give dividing points of these division points as control points of the original 3D mesh model.
CN 200910214068 2009-12-23 2009-12-23 Figure retrieving method CN101719140B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN 200910214068 CN101719140B (en) 2009-12-23 2009-12-23 Figure retrieving method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN 200910214068 CN101719140B (en) 2009-12-23 2009-12-23 Figure retrieving method

Publications (2)

Publication Number Publication Date
CN101719140A true CN101719140A (en) 2010-06-02
CN101719140B true CN101719140B (en) 2012-04-18

Family

ID=42433714

Family Applications (1)

Application Number Title Priority Date Filing Date
CN 200910214068 CN101719140B (en) 2009-12-23 2009-12-23 Figure retrieving method

Country Status (1)

Country Link
CN (1) CN101719140B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9817845B2 (en) 2013-12-12 2017-11-14 Xyzprinting, Inc. Three-dimensional image file searching method and three-dimensional image file searching system

Families Citing this family (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102375831B (en) * 2010-08-13 2014-09-10 富士通株式会社 Three-dimensional model search device and method thereof and model base generation device and method thereof
US8437506B2 (en) * 2010-09-07 2013-05-07 Microsoft Corporation System for fast, probabilistic skeletal tracking
CN101957992B (en) * 2010-09-30 2012-05-02 清华大学 Data characteristic extracting and matching method for two-dimensional injective mapping curved surface
CN102004922B (en) * 2010-12-01 2012-12-26 南京大学 High-resolution remote sensing image plane extraction method based on skeleton characteristic
CN102063475B (en) * 2010-12-22 2012-10-10 张丛喆 Webpage user terminal presenting method of three-dimensional model
CN103309938A (en) * 2013-04-24 2013-09-18 中国科学院遥感与数字地球研究所 Method for realizing high-performance of polygonal maximum inner circle
CN103412947A (en) * 2013-08-26 2013-11-27 浙江大学 Polygon search method for big space data
CN103544734A (en) * 2013-10-11 2014-01-29 深圳先进技术研究院 Street vie based three-dimensional map modeling method
CN104503275B (en) * 2014-11-21 2017-03-08 深圳市超节点网络科技有限公司 Non-contact gesture-based control method and apparatus
CN105354866A (en) * 2015-10-21 2016-02-24 郑州航空工业管理学院 Polygon contour similarity detection method
CN106909539A (en) * 2015-12-22 2017-06-30 成都理想境界科技有限公司 Image retrieval system, server and database and relevant method

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9817845B2 (en) 2013-12-12 2017-11-14 Xyzprinting, Inc. Three-dimensional image file searching method and three-dimensional image file searching system

Also Published As

Publication number Publication date Type
CN101719140A (en) 2010-06-02 application

Similar Documents

Publication Publication Date Title
Huang et al. Shape registration in implicit spaces using information theory and free form deformations
Tung et al. The augmented multiresolution Reeb graph approach for content-based retrieval of 3D shapes
Krishnamurthy et al. Fitting smooth surfaces to dense polygon meshes
Campbell et al. A survey of free-form object representation and recognition techniques
de La Gorce et al. Model-based 3d hand pose estimation from monocular video
Dong et al. Harmonic functions for quadrilateral remeshing of arbitrary manifolds
Sundar et al. Skeleton based shape matching and retrieval
Pauly et al. Example-based 3D scan completion
Rusu et al. Close-range scene segmentation and reconstruction of 3D point cloud maps for mobile manipulation in domestic environments
Wang et al. Feature based 3D garment design through 2D sketches
US20040249809A1 (en) Methods, systems, and data structures for performing searches on three dimensional objects
Surazhsky et al. Explicit surface remeshing
Huang et al. Combinatorial manifold mesh reconstruction and optimization from unorganized points with arbitrary topology
Lai et al. Fast mesh segmentation using random walks
Shamir A survey on mesh segmentation techniques
Attene et al. Hierarchical mesh segmentation based on fitting primitives
Chang et al. Automatic registration for articulated shapes
Lee et al. Mesh scissoring with minima rule and part salience
Nishino et al. Robust simultaneous registration of multiple range images
Ju Fixing geometric errors on polygonal models: a survey
Ikeuchi et al. The great buddha project: Digitally archiving, restoring, and analyzing cultural heritage objects
Ahmed et al. Dense correspondence finding for parametrization-free animation reconstruction from video
Wand et al. Reconstruction of deforming geometry from time-varying point clouds
Ke et al. Feature-based reverse modeling strategies
Mehra et al. Abstraction of man-made shapes

Legal Events

Date Code Title Description
C06 Publication
C10 Entry into substantive examination
C14 Grant of patent or utility model
EXPY Termination of patent right or utility model