CN106373186B - A kind of 3D model topology mapping method based on deformation driving - Google Patents
A kind of 3D model topology mapping method based on deformation driving Download PDFInfo
- Publication number
- CN106373186B CN106373186B CN201610782774.3A CN201610782774A CN106373186B CN 106373186 B CN106373186 B CN 106373186B CN 201610782774 A CN201610782774 A CN 201610782774A CN 106373186 B CN106373186 B CN 106373186B
- Authority
- CN
- China
- Prior art keywords
- mapping
- model
- node
- search
- file
- Prior art date
- Legal status (The legal status 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 status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Computer Graphics (AREA)
- Geometry (AREA)
- Software Systems (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention relates to a kind of 3D model topology mapping methods based on deformation driving, comprising the following steps: 1) extracts model file;2) corresponding 3D model is established according to the model file;3) mapping status is searched for by heuritic approach, described search process is driven by Mapping Energy;4) optimal mapping status is selected according to search result;5) visualization of corresponding source model and object module is realized according to the optimal mapping status;6) destination file is exported.Compared with prior art, the present invention has many advantages, such as that high-efficient, to return the result quality high.
Description
Technical field
Graphics analysis techniques field of the present invention is related to a kind of method of 3D model mapping, more particularly, to one kind based on change
The 3D model topology mapping method of shape driving is a kind of other model mapping for handling Geometrical change and change in topology of component-level
Method.
Background technique
Model mapping algorithm is a kind of underlying issue in graphics, it is widely used in the demand of multiple fields, than
Such as object identification, cluster, model deformation, static modelling etc..For a long time, model mapping algorithm has caused people's
Pay attention to, also emerges a large amount of correlative study.Before, most of the classical related algorithm of model mapping problems all focuses on tool
" the Robust global registration " of the model, such as GELFAND for thering is particular geometric to require et al., ZHANG et al.
" Deformation-driven shape correspondence ", " Non-rigid of Huang et al.
registration under isometric deformations".For these algorithms, when two model gaps increase
When big, mapping search will be more difficult to, and the computation complexity of algorithm also can be accordingly higher.
In recent years, with the proposition of the field new concept and various technological innovations, some new algorithms also occur therewith.Than
Such as the relevant KIM of machine learning, " the Learning Part-based Templates from Large of V.G et al.
Collections of 3D Shapes " and Van et al. " Unsupervised co-segmentation of a set of
Shapes via descriptor-space spectral clustering ", " the Fit and diverse:Set of Xu et al.
evolution for inspiring 3d shape galleries".But the result that generally speaking, these methods return is opposite
For more coarse and dispersion, particularly with there is mapping of the topological distorted pattern to that can not obtain high quality.
Summary of the invention
It is an object of the present invention to overcome the above-mentioned drawbacks of the prior art and provide a kind of high-efficient, return knots
The high 3D model topology mapping method based on deformation driving of fruit quality.
The purpose of the present invention can be achieved through the following technical solutions:
A kind of 3D model topology mapping method based on deformation driving, comprising the following steps:
1) model file is extracted;
2) corresponding 3D model is established according to the model file;
3) mapping status is searched for by heuritic approach, described search process is driven by Mapping Energy;
4) optimal mapping status is selected according to search result;
5) visualization of corresponding source model and object module is realized according to the optimal mapping status;
6) destination file is exported.
The model file includes node structure information, connection structure information and component information, the node structure information
Including the fine-grained divided parts information of model, the connection structure information includes the physical connection information between each component, institute
Stating component information includes model topology tectonic information.
In the step 2), when establishing 3D model, each divided parts structure in analytic modell analytical model file, according to model text
Segmented mode in part obtains fine-grained component.
Search process in the step 3) specifically:
301) heuritic approach constructs search tree, and a kind of search condition of each node on behalf on the search tree is described to search
Strand state includes Mapping Energy, mapping means and non-mapping means;
302) node is expanded by Mapping Energy driving method, when carrying out node expansion, it is minimum only expands mapping consumption
Preceding k child node, each child node increase the mapping of at least one component on the basis of father node;
303) when the branch line of all search reaches leaf node, search process terminates, and obtains fullpath.
In the heuritic approach, for each node, each searching position is commented in search condition space
Estimate, is scanned for from optimal location.
During described search, the factor that the estimation function that each node scans for considers includes Mapping Energy, source portion
Pantograph ratio and each component participation weight relative to estimation function of the part to target component.
The estimation formulas of the Mapping Energy are as follows:
E=Ed+wc*Ec+ws*Es
Wherein, E is Mapping Energy, and Ec is connectivity changing value, and Es is change in topology value, and wc and ws are weighted value.
The destination file includes work summary, journal file, mapped file and mapping graph.
Compared with prior art, the invention has the following advantages that
(1) present invention carries out heuristic search as reference frame using the mapping consumption of deformation driving, passes through control
Optimal solution quantity processed to carry out child node beta pruning in each search step, and finally obtains optimal case, realizes more precisely
Mapping Energy estimate mode.
(2) when the present invention establishes 3D model according to model file, Model Abstraction is the structure chart comprising curve and surface, drop
The complex properties of low true model are interfered to algorithm bring.
(3) it when the present invention utilizes heuristic, for each search node, provides in search condition space to each
A searching position is assessed, and obtains best position, then scan for can be omitted so big until target from this position
Meaningless searching route is measured, is improved efficiency.
(4) factor that the estimation function that the present invention searches for considers includes the scaling of Mapping Energy, source block to target component
Than the participation weight with each component relative to estimation function, different weights of the different components relative to estimation function, energy are given
It is enough to carry out model mapping more flexiblely.
(5) it when the present invention scans for, is carried out using prune approach, only retains current Mapping Energy most in each expand
Small preceding k child node, can effectively command deployment space, improve efficiency of algorithm.
Detailed description of the invention
Fig. 1 is the structural diagram of the present invention;
The time of k value is set when Fig. 2 is beta pruning;
Fig. 3 is the cost time under each estimation function.
Specific embodiment
The present invention is described in detail with specific embodiment below in conjunction with the accompanying drawings.The present embodiment is with technical solution of the present invention
Premised on implemented, the detailed implementation method and specific operation process are given, but protection scope of the present invention is not limited to
Following embodiments.
As shown in Figure 1, being specifically described such as the present embodiment provides a kind of 3D model topology mapping method based on deformation driving
Under.
Provide input: source model M, object module M '.
The first step extracts model file, extracts the specifying informations such as model geometry, topological structure, position attribution.Model
File is inputted with xml document, and model file contains partitioning site and the connection of figure, including node structure information, connection structure
Information and component information, node structure information include the fine-grained divided parts information of model, and connection structure information includes each portion
Physical connection information between part, component information include model topology tectonic information.In each structure, there is detailed geometry
Information is (such as the three-dimensional coordinate put, name of parts, node type etc..
Second step establishes corresponding 3D model according to model file.According to file content, each of analytic modell analytical model is drawn
Sub-unit structure obtains fine-grained component according to the segmented mode in model file.It is to include curve and table by Model Abstraction
The structure chart in face, the method for structural texture figure " Topology-varying 3D shape as the method for Alhashim et al.
Creation via structural blending ", wherein curve describes one-dimensional model data, and surface then describes two
Dimension module data.By abstracting model in this way, the complex properties that can reduce true model are interfered to algorithm bring.
In this example, modular construction refers to: being drawn by certain (usually common-sense) criterion to model itself
Point.Such as in seat model, the components such as backrest, seat cushion, handrail, chair leg can be greatly classified into.It divides in example by mould
Type file provides.
Third step searches for mapping status by heuritic approach, and search process is driven by Mapping Energy.Search process will lead to
Heuritic approach search mapping status is crossed, and finds out the solution for meeting demand condition.In search process, searched with heuristic construction
Suo Shu, constantly expansion node, until finding fullpath.The reference for expanding node is Mapping Energy, search process specifically:
301) heuritic approach constructs search tree, and a kind of search condition of each node on behalf on the search tree searches for shape
State includes Mapping Energy, mapping means and non-mapping means;
302) node is expanded by Mapping Energy driving method, when carrying out node expansion, it is minimum only expands mapping consumption
Preceding k child node is expanded using prune approach, each child node increases at least one portion on the basis of father node
Part mapping;
303) when the branch line of all search reaches leaf node, search process terminates, and obtains fullpath, fullpath
Refer to, the paths in search tree from source node to target.The end of fullpath must be the leaf node in search tree.
Wherein, leaf node refers to, not the search node of child node, and in leaf node, all components of source model are
The a certain component being mapped as in object module.
Wherein, beta pruning refers to, in each search step, the quantity of siding stopping node when expanding for next layer of search tree.
In this example, an integer k (k > 0) is set, then only retains the smallest preceding k son section of current Mapping Energy in each expand
Point.In this way, can effectively command deployment space, improve efficiency of algorithm.
In the heuristic applied in this example, for each search node, provide in search condition space to every
One searching position is assessed, and obtains best position, then scan for until target from this position.It can be omitted in this way
A large amount of meaningless searching routes, improve efficiency.
In heuristic search, the appraisal to position is highly important.Can have using different appraisals different
Effect.
Following evaluation function can be used:
F'(n)=g'(n)+h'(n)
Wherein, f'(n) it is evaluation function;It g'(n) is minimum Mapping Energy value of the starting point to node n;It h'(n is) n to mesh
The inspiration value of target minimum Mapping Energy.
In search process, the factor that estimation function that each node scans for considers includes that Mapping Energy, source block arrive
The participation weight of the pantograph ratio of target component and each component relative to estimation function.
Wherein, Mapping Energy refers to, it will be appreciated that for the cost of mapping.For instinctively, Mapping Energy is smaller, illustrates this
It is mapped in source model and the deformation of object module corresponding component bring is smaller, also just closer to our needs.In general, it calculates
Method needs to guarantee not change in conversion process the properties of original image: such as symmetry, relative position, connectivity.Specifically
It says, Mapping Energy is determined by several factors.In " Deformation-Driven Topology-Varying 3D Shape
Correspondence " in, estimate the function of Mapping Energy are as follows:
E=Ed+wc*Ec+ws*Es
Wherein, Ed refers to geometry deformation.Geometry deformation refers to the variations such as position, the shape of mapping correspondence point.In order to weigh
This parameter is measured, each component is sampled in function, and is come really by comparing the Geometrical change between corresponding sampled point
Determine deformation.In this example, Px and Py respectively represents two corresponding components.Ec refers to that connectivity changes, and connectivity becomes
Change and changes mainly for the relative position between the component that is connected.Es refers to change in topology.Change in topology refers in mapping process
The case where number of components changes, such as division or fusion.Wc is the weighted value of Ec, and ws is the weighted value of Es.
Pantograph ratio EsclRefer to source block to target component scaling multiple:
Escl=min { v (px)/v (py), 2-v (px)/v (py) }
Wherein Px and Py respectively represents two corresponding components, and v indicates component geometric volume.
In existing main stream approach, all components are coequally treated.However for general semanteme, these components
Usually possess different importance.Therefore, the ginseng that the present invention gives one weight of each component to determine it to estimation function
With degree.In this example, this weight is determined by the geometric volume of component.
4th step selects optimal mapping status according to search result, and optimal mapping status, which refers to, spends Mapping Energy minimum
Mapping scheme.It is specific as follows:
Solution=arg Min energy (c) | c ∈ leafNode }
Wherein, energy (c) indicates that the Mapping Energy of node c, leafNode indicate leaf segment point set.
5th step displays source model, object module, mapping situation visualization, to reach intuitivism apprehension.In this reality
In example, visualize mainly for model.
6th step writes the result into file and stores.It as a result include work summary, journal file, mapped file, mapping
Figure.Work summary (job) includes the position of remaining destination file, and job number, optimal mapping expends and optimal mapping scheme;Log
File (log) includes source model file, object module file, optimal mapping consuming;Mapped file (.match) has recorded optimal
Mapping scheme provides a pair of of component mapping in every a line of mapped file;Mapping graph (.png) provides mapping with visual means
The directviewing description of scheme, the identical component of color represent them and map relatively.
According to above-mentioned steps, model data is tested and is analyzed, the test set have chosen 20 pairs of models into
Row test, as shown in table 1.All tests realize on PC computer, the major parameter of the PC computer are as follows: 2 centres
Manage device Intel (R) Core (TM) i5-2450M CPU@2.50GHz, memory 8GB.
Table 1
The results show that the method for the present invention is a kind of efficient and high quality Model Matching algorithm, and pantograph ratio is introduced
There is improvement effect for some examples with component weight, and the time is spent to increase, as shown in Figure 2.To sum up, these experiment tables
Bright, estimation function appropriate is extremely important to result.
In the following experiment, different estimation functions is used:
Estimation function 1 selected parameter Ed=1.1, Es=0.4, Ec=0.7;
Estimation function 2 selects Ed=1.1, Es=0.4, Ec=0.7, Ed=0.2, Escl=0.3;
Ed=1.1, Es=0.4, Ec=0.6, Ed=0.3, E in estimation function 3scl=0.2;
The parameter selection of estimation function 4 introduces component weight with 3.
Mapping Energy under above-mentioned estimation function is as shown in table 2.
Table 2
Example number | Estimation function 1 | Estimation function 2 | Estimation function 3 | Estimation function 4 |
1 | 0.1658 | 0.1423 | 0.1322 | 0.1211 |
2 | 0.1375 | 0.1342 | 0.1221 | 0.1321 |
3 | 0.1653 | 0.0912 | 0.1322 | 0.1232 |
4 | 0.121468 | 0.1221 | 0.1203 | 0.1123 |
5 | 0.0915 | 0.0812 | 0.0678 | 0.0622 |
6 | 0.10322 | 0.0823 | 0.0921 | 0.0812 |
7 | 0.0478 | 0.0432 | 0.0232 | 0.0221 |
8 | 0.10853 | 0.1008 | 0.1123 | 0.1036 |
9 | 0.14445 | 0.1232 | 0.1321 | 0.1123 |
10 | 0.0825 | 0.1121 | 0.1012 | 0.0982 |
11 | 0.08217 | 0.0762 | 0.0621 | 0.0528 |
12 | 0.0405 | 0.0213 | 0.0423 | 0.0321 |
13 | 0.09404 | 0.0621 | 0.0452 | 0.0423 |
14 | 0.140699 | 0.0921 | 0.1219 | 0.1125 |
15 | 0.16525 | 0.1219 | 0.1148 | 0.1123 |
16 | 0.121468 | 0.1123 | 0.1021 | 0.1024 |
17 | 0.18288 | 0.1387 | 0.1432 | 0.1323 |
18 | 0.1234 | 0.1212 | 0.1172 | 0.1134 |
19 | 0.13306 | 0.1421 | 0.1321 | 0.1343 |
20 | 0.154673 | 0.0982 | 0.1321 | 0.1068 |
Different value of K is set in beta pruning, due to selection is in test optimal solution, so do not have substantially on solution quality
Too big variation, but can be seen that the influence to the time is arranged in k value, as shown in Figure 3.
Claims (6)
1. a kind of 3D model topology mapping method based on deformation driving, which comprises the following steps:
1) model file is extracted;
2) corresponding 3D model is established according to the model file;
3) mapping status is searched for by heuritic approach, search process is driven by Mapping Energy;
4) optimal mapping status is selected according to search result;
5) visualization of corresponding source model and object module is realized according to the optimal mapping status;
6) destination file is exported;
In the step 2), when establishing 3D model, each divided parts structure in analytic modell analytical model file, according in model file
Segmented mode obtain fine-grained component;
Search process in the step 3) specifically:
301) heuritic approach constructs search tree, a kind of search condition of each node on behalf on the search tree, described search shape
State includes Mapping Energy, mapping means and non-mapping means;
302) node is expanded by Mapping Energy driving method, when carrying out node expansion, only expands mapping and consume minimum preceding k
Child node, each child node increase the mapping of at least one component on the basis of father node;
303) when the branch line of all search reaches leaf node, search process terminates, and obtains fullpath.
2. the 3D model topology mapping method according to claim 1 based on deformation driving, which is characterized in that the model
File includes node structure information, connection structure information and component information, and the node structure information includes that model is fine-grained
Divided parts information, the connection structure information include the physical connection information between each component, and the component information includes mould
Type topological structure information.
3. the 3D model topology mapping method according to claim 1 based on deformation driving, which is characterized in that the inspiration
In formula algorithm, for each node, each searching position is assessed in search condition space, from optimal location into
Row search.
4. the 3D model topology mapping method according to claim 1 based on deformation driving, which is characterized in that described search
In the process, the factor that the estimation function that each node scans for considers includes the contracting of Mapping Energy, source block to target component
It puts than the participation weight with each component relative to estimation function.
5. the 3D model topology mapping method according to claim 4 based on deformation driving, which is characterized in that the mapping
The estimation formulas of energy are as follows:
E=Ed+wc*Ec+ws*Es
Wherein, E is Mapping Energy, and Ec is connectivity changing value, and Es is change in topology value, and wc and ws are weighted value.
6. the 3D model topology mapping method according to claim 1 based on deformation driving, which is characterized in that the result
File includes work summary, journal file, mapped file and mapping graph.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610782774.3A CN106373186B (en) | 2016-08-30 | 2016-08-30 | A kind of 3D model topology mapping method based on deformation driving |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610782774.3A CN106373186B (en) | 2016-08-30 | 2016-08-30 | A kind of 3D model topology mapping method based on deformation driving |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106373186A CN106373186A (en) | 2017-02-01 |
CN106373186B true CN106373186B (en) | 2019-04-02 |
Family
ID=57899787
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610782774.3A Active CN106373186B (en) | 2016-08-30 | 2016-08-30 | A kind of 3D model topology mapping method based on deformation driving |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106373186B (en) |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103871074A (en) * | 2012-12-18 | 2014-06-18 | 帕洛阿尔托研究中心公司 | Simultaneous mapping and registering thermal images |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US10134177B2 (en) * | 2015-01-15 | 2018-11-20 | Samsung Electronics Co., Ltd. | Method and apparatus for adjusting face pose |
-
2016
- 2016-08-30 CN CN201610782774.3A patent/CN106373186B/en active Active
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103871074A (en) * | 2012-12-18 | 2014-06-18 | 帕洛阿尔托研究中心公司 | Simultaneous mapping and registering thermal images |
Non-Patent Citations (1)
Title |
---|
Deformation-Driven Topology-Varying 3D Shape Correspondence;Ibraheem Alhashim等;《ACM Transactions on Graphics》;20151130;第34卷(第6期);全文 |
Also Published As
Publication number | Publication date |
---|---|
CN106373186A (en) | 2017-02-01 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US7995841B2 (en) | Hybrid graph model for unsupervised object segmentation | |
JP2019091434A (en) | Improved font recognition by dynamically weighting multiple deep learning neural networks | |
CN107704638B (en) | Elevator machinery room diagram generating device and method, modeling data generating device and method | |
CN101159064A (en) | Image generation system and method for generating image | |
CN109002834A (en) | Fine granularity image classification method based on multi-modal characterization | |
JP6612487B1 (en) | Learning device, classification device, learning method, classification method, learning program, and classification program | |
CN110163208B (en) | Scene character detection method and system based on deep learning | |
JP2014179090A (en) | Segmentation of image based on color and color difference | |
Jacobsen et al. | Metrosets: Visualizing sets as metro maps | |
US10410380B2 (en) | Image rendering utilizing procedural yarn model generated in multi-stage processing pipeline | |
CN105354593B (en) | A kind of threedimensional model sorting technique based on NMF | |
JP2019045894A (en) | Retrieval program, retrieval method and information processing apparatus operating retrieval program | |
CN115081087A (en) | Decoration cloud design method, device, equipment and storage medium based on Internet of things | |
US8564594B2 (en) | Similar shader search apparatus and method using image feature extraction | |
Jang et al. | Automatic wall detection and building topology and property of 2D floor plan (short paper) | |
CN106373186B (en) | A kind of 3D model topology mapping method based on deformation driving | |
CN108399288A (en) | A kind of device adding decorative element automatically in planar design | |
JP4883719B2 (en) | Similar image retrieval method and apparatus | |
CN107220481B (en) | Quantitative evaluation method for comprehensive quality of planar element map | |
Gralka et al. | 2016 IEEE scientific visualization contest winner: Visual and structural analysis of point-based simulation ensembles | |
Yu et al. | Grammar-based automatic 3D model reconstruction from terrestrial laser scanning data | |
JP6160522B2 (en) | Design management apparatus and program | |
US20160259864A1 (en) | Fluid transition route acquisition apparatus, fluid transition route acquisition method, and program | |
Wu | Research on embroidery image recognition based on deep learning | |
US20150302115A1 (en) | Method and apparatus for creating 3d model |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |