CN101694727A - Ancient Chinese construction process modeling method based on construction drawings - Google Patents
Ancient Chinese construction process modeling method based on construction drawings Download PDFInfo
- Publication number
- CN101694727A CN101694727A CN200910235591A CN200910235591A CN101694727A CN 101694727 A CN101694727 A CN 101694727A CN 200910235591 A CN200910235591 A CN 200910235591A CN 200910235591 A CN200910235591 A CN 200910235591A CN 101694727 A CN101694727 A CN 101694727A
- Authority
- CN
- China
- Prior art keywords
- model
- obtains
- tree
- construction
- rule
- 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.)
- Pending
Links
Images
Landscapes
- Processing Or Creating Images (AREA)
Abstract
An ancient Chinese construction process modeling method based on construction drawings comprises steps as follows: (1) separating the construction drawings on the basis of a repeat characteristic of construction components to obtain construction components, (2) separating the drawings from top to bottom to obtain a tree structure of buildings, and then obtaining semantics of construction modules by utilizing the belief propagation method to identify types of the construction types on the basis of the tree structure, and (3) concluding rules on the basis of the model semantics, and finally building construction modules by combining a model component library.
Description
Technical field
The invention belongs to computer virtual reality, process model building technical field, specifically belong to identification Chinese ancient architecture drawing, extracting rule generates three-dimensional building model field automatically.
Background technology
Building scholar Stiny[Stiny72] notion of shape grammar at first proposed.Shape grammar successfully is used for buildings design and analysis, structure.Original shape grammar directly acts on line and point, determines their distribution.Inference rule can be expressed as setting-out and point and name them.In the reality, this will cause a derivation problem, because in each step derivation, will have many rules to select, and many different conversion are arranged.
For making shape grammar be applicable to graphics, Wonka[Wonka03 more] and Muller[Muller06] etc. the people introduced a framework, regular automated reasoning mechanism be replaced and introduced to shape wherein can by one, two or more sub-shapes.Have only automatically rule to derive and to carry out large-scale process model building.The main energy of original grammer of people such as Wonka design concentrates on and adopts splitting operation to make that the size between the rule is independent.Splitting operation can be along cleave plane cutting basic configuration (for example cube).The size that is independent of rule allows the deviser to specify a cleave plane position how to change when basic configuration changes, and how many cleave plane should be used for a shape of determining size.
Muller[Muller06] on the basis of division rule, added some other functions, the CGA grammer has been proposed.At first, he has introduced the shape manipulation of rough model, and the rough model of BUILDINGS MODELS is constituted by some simple shapes.Rough model uses very direct, and consistent with the actual design process of building.Secondly, they have introduced the generation that the context dependent rule is come controlling models, make that a large amount of different BUILDINGS MODELS can rule-based automatic generation.
Lipp[Lipp08] introduced interaction syntax editor's method, replace the text editing rule, make and adopt graphical user interface design, editing rule to become possibility fully.Semantic selection and consistance that the document has mainly solved in the model editing are revised problem, and this method makes that process model building is easier is accepted by the user.
Recently, the method for computer vision and process model building is combined obtained certain success.Muller[Muller07] etc. by the thought of division rule, designed one at the top-down analytical approach of elevation of building single image, at first, find symmetrical repetitive structure important in the image, stipulations obtain Molecular Graphs (irreducible facade) again then.It is used for seeking divisural line to determine the row and column of facade with global optimization.Further disaggregated classification is like the division rule of [Muller06], the last parameter that extracts the shape grammar rule.This makes that rebuilding the building that generates variation becomes possibility.
Aliaga[Aliaga07] a kind of Style Grammar proposed, automatically from cut apart and the image of mark extracting rule, can generate the difference building of same style fast.Adopt some models that obtained, the model that can very fast design makes new advances, they have similar style, can adopt the method for the sense of reality or feeling of unreality to draw out.In addition, he also utilizes multiple image to solve occlusion issue and uses the colour equation algorithm to make it be applicable to the situation that light alters a great deal.
Also do not generate the achievement in research report of building rule in the present document, but some achievements in research are being arranged aspect the architectural drawing identification based on architectural drawing.Building modeling based on drawing has had further investigation, Yin etc. [Yin09] have summarized relevant work, at architectural plan,, mainly solve problems such as removing noise, literal extraction, vector quantization and Symbol recognition by identification special symbol and Straight Line Identification drawing.Lu etc. [Lu07] have designed an architectural plane drawing recognition system, and earlier the identification parallel segment is to as wall and remove, and remaining figure carries out pattern match as symbol and predefined symbol, reaches identifying purpose.But below all be at the architectural plane drawing, generate the building interior model.Chen etc. [Chen08] have studied the building modeling based on sketch, use is based on the method identification sketch of MAP (Maximum APosteriori), generate rough model, obtain the basic plane of buildings, retrieval component and texture add in each basic plane in the sample storehouse then, construct the three-dimensional model of complete buildings.But the problem that this method mainly solves is based on the three-dimensional reconstruction and the retrieval of sketch, rather than the Direct Recognition drawing.Above-mentioned these methods all are to mate by pattern or symbol with predefined, realize the automatic identification of architectural drawing.But, in process model building,, need the identification architectural elevation, and this can't predefined for automatic create-rule, therefore very big from the difficulty of the automatic create-rule of architectural drawing.
[Stiny72]G.Stiny?and?J.Gips.Shape?Grammars?and?the?Generative?Specification?ofPainting?and?Sculpture.Proc.IFIP?Congress?71，North-Holland，1972：1460-1465.
[Wonka03]Wonka?P.，Wimmer?M.，Sillion?F.，Ribarsky?W.Instant?architecture.ACMTransactions?on?Graphics?22，3(2003)，669-677.
[Müller06]Müller?P.，WONKA?P.，HAEGLER?S.，ULMER?A.，VAN?GOOL?L.Procedural?modelingof?buildings.ACM?Transactions?on?Graphics?25，3(2006)，614-623.
[Lipp08]Lipp?M.，Wonka?P.，Wimmer?M.Interactive?visual?editing?of?grammars?forprocedural?architecture.ACM?Transactions?on?Graphics?27，3(2008)，1-10.
[Müller07]Müller?P.，ZENG?G.，Wonka?P.，Gool?L.V.Image-based?procedural?modelingof?facades.ACM?Transactions?on?Graphics?24，3(2007)，85.
[Aliaga07]Aliaga?D.G.，Rosen?P.A.，Bekins?D.R.Style?grammars?for?interactivevisualization?of?architecture.IEEE?Transactions?on?Visualization?and?ComputerGraphics?13，4(2007)，786-797.
[Yin09]X.T.Yin，P.Wonka，A.Razdan.Generating?3D?Building?Models?fromArchitectural?Drawings：A?Survey.IEEE?Computer?Graphics?and?Appl?ications，29(1)，2009：20-30.
[lu07]T.Lu，etc.Automatic?Analysis?and?Integration?of?Architectural?Drawings.International?Journal?on?Document?Analysis?and?Recognition，9(1)，2007：31-47.
[Chen08]Chen，X.，Kang，S.B.，Xu，Y.etc.Sketching?reality：Realisticinterpretation?of?architectural?designs.ACM?Trans.Graph.27(2)：109，2008.
Summary of the invention
The technical problem to be solved in the present invention: process model building technology and computer vision are combined, and extracting rule from architectural drawing generates the three-dimensional building model automatically.
The solution that the present invention adopts, it is as follows to relate to major technique:
(1) repeats characteristics based on construction set, cut apart drawing, obtain construction set.
(2) the top-down drawing of cutting apart, the hierarchical tree structure that obtains building based on this hierarchical structure, adopts BeliefPropagation method identification construction set type, obtains the BUILDINGS MODELS semanteme.
(3) based on model semantics, induction rule, the combination model Component Gallery generates BUILDINGS MODELS.
Description of drawings
Fig. 1 is the drawing image of input;
Fig. 2 is for filling the regional group model that obtains, and the same area group is used with a kind of color showing;
Fig. 3 is the conversion of zone group 1;
Fig. 4 is the RPT of zone group 1;
Fig. 5 is the conversion of zone group 2;
Fig. 6 is the RPT of zone group 2;
Fig. 7 is the result after the pattern match;
The assembly segmentation result of Fig. 8 for obtaining behind the region growing;
The hierarchical model of Fig. 9 for obtaining by divided in horizontal direction;
Figure 10 is an abstract architectural elevation;
Figure 11 is the hierarchical tree of elevation of building, numeral node numbering wherein, and English alphabet is represented the leafy node type;
Figure 12 a and Figure 12 b three-dimensional building model for generating.
Embodiment
(1) repeats characteristics based on construction set, cut apart drawing, obtain construction set
The input the Chinese ancient architecture drawing image as shown in Figure 1,
Notice model component, as door, window, post, watt, sets of brackets on top of the columns etc., all have repeatability in the drawings, so utilize this repetition characteristics from data, to be partitioned into repetitive.Mainly solve the combinatorial problem of different repeat units.
At first,, obtain the fundamental region with indiscriminate (Flood Fill) the method blank map picture of inundation, and matched shape size same area in the horizontal direction, being grouped together obtains the zone group, fills out identical color, as shown in Figure 2.This is six doors.Zone group 1 is represented two different zone groups with zone group 2.
For the zone group that will have same repeat pattern is grouped together, obtain construction set, we have designed the repeat pattern tree, and (Repetitive Pattern Tree RPT) represents all possible repeat pattern of each zone group, and is used for mating different zone groups.A zone group might all be coded among its RPT by repeat pattern.The make of RPT is as follows: the map table that at first will construct each zone group, zone of each vertex representation, the conversion between adjacent two zones is represented on every limit, for zone group 1, its map table as shown in Figure 3, identical conversion is labeled identical letter, the total different conversion of ABC three classes in zone group 1.
After obtaining map table, construct RPT with it.At first delete the minimum limit of multiplicity in the map table, if remaining be identical together with component, then finish once division, obtain one deck of RPT.Continue the minimum limit of deletion multiplicity then, so circulation is gone down, till not having fillet.In last example, at first delete limit C, what discovery was remaining is ABA together with component, ABA, therefore ABA generates the ground floor of RPT, second step deletion limit B, remaining is A together with component, A, A, A, A, therefore A generates the second layer of PRT, and is last, delete limit A, obtain last one deck of RPT, the RPT of generation as shown in Figure 4.A kind of repeat pattern of each of RPT layer this zone group of expression.Zone group 2 map tables and RPT are shown in Figure 6 as Fig. 5 respectively.
The coupling of two RPT: if the n layer of two RPT coupling, so the their the 0th, 1 ..., the n layer all mates.At two RPT, we successively mate (the 0th layer is root node, and the 0th layer of any two RPT all mates) since the 0th layer, till can not mating, obtain mating number of stories m.If m＞0 is thought that then two trees mate, otherwise is not matched.
The 0th, 1,2 layers of zone group 1 and zone group 2 equal coupling in last example, thereby obtain zone group 1 by repeating in twos, zone group 2 repeats can match together by three or three, also can obtain the repeat pattern of door assembly.In drawing, we try to achieve the RPT of each zone group, and mate them, and the zone combination that will be complementary also obtains the result as shown in Figure 7.
Coupling based on RPT is strict matching process, it can be grouped together the zone group of strictness coupling, but owing to blocking or reason such as noise, some zone that should mate can't match together, mated after the All Ranges with RPT, obtain a plurality of match patterns, the priority definition of every kind of match pattern is total number of matching area in each pattern.Descending according to priority, for each pattern, it as the standard weight complex pattern, is used for it the growth of adjacent areas coupling, calculate adjacent area its redundancy rate and miss rate relatively, the approximate match adjacent area, growth obtains model component.Growth result as shown in Figure 8
(2) generating structure tree is adopted BP recognizer component type
In the previous step, we obtain construction set in bottom-up merging zone, below we are top-down cuts apart, obtain the structure tree of whole building, discern the type of each assembly then, obtain the semantic structure tree of model.
Because Chinese ancient architecture has tangible hierarchy feature, we cut apart drawing in the horizontal and vertical directions, till having only the single component type, obtain the structure tree of representation model by different level like this.The leafy node of structure tree is represented a component type, and non-leafy node is represented the compound of component type in its child node.The segmentation result of first level as shown in Figure 9.Because have the mutual restriction relation between the different assemblies,, should determine that the dependence between them is discerned together so different component type should not be discerned separately.We are according to the dependence between the definite assembly of model structure tree that obtains previously.For the positive facade structures of building as shown in figure 10, its structure tree as shown in figure 11, numeral node numbering wherein, English alphabet is represented the leafy node type.
We adopt with different levels recognition methods, discern a level at every turn.In each layer, be defined as follows energy function:
Wherein E (l) represents the 1st layer energy, E
1(c
i) defined assembly c
iInternal energy, E
2(c
i, c
I+1) defined assembly c
i, c
I+1Between the energy of mutual restriction.Assembly c
i, c
I+1Between have mutual restriction relation, and if only if c
i, c
I+1It is adjacent sibling.For non-leafy node, need from its child node, select of the representative of a component type as it.For certain non-leafy node of the 1st layer, the component type of its all child nodes is all regarded as in 1 layer to be occurred once, we in the 1st layer, selects the maximum assembly of frequency of occurrence and not Shi Bie the representative of this non-leafy node of assembly conduct.If the maximum assembly of a plurality of occurrence numbers is arranged, we are chosen in the 1st layer and cut apart a longest assembly of direction length and represent as it.After the dependence between having determined to set up, use Belief Propagation algorithm optimization energy function E (l).
For the example among Figure 11, the 0th layer is root node, does not need identification.In the 1st layer, node 2 is non-leafy nodes, selects assembly B to represent as it, and in the 2nd layer, the representative of node 5,7,9 is assembly C.
(3) induction rule and model generate
Based on the model semantics structure tree that obtains in (2) step, adopt the method for process model building, induction rule, generation model.The semantic structure tree representation the top-down hierarchical structure and the semantic information of model.We adopt the CGA grammer, obtain the model rule from structure tree, and in conjunction with the priori of Chinese ancient architecture, determine absolute and relative parameter.The depth data that wherein lacks has user interactions to specify.We have set up the three-dimensional model Component Gallery, and the three-dimensional model of retrieval component from Component Gallery according to assembly drawing and type then according to the rule of concluding previously, to together, generates the three-dimensional building model that has semantic information with combination of components.By revising parameter, can generate the different models of variation.Figure 12 adds the door and window number, the different models that obtain automatically for the three-dimensional building model of generation with by revising parameter.
Claims (4)
1. based on the Chinese ancient architecture process modeling approach of architectural drawing, it is characterized in that comprising the steps:
(1) repeat characteristics based on construction set, the bottom-up zone of convergency obtains construction set;
(2) the top-down drawing of cutting apart, the hierarchical tree structure that obtains building based on this hierarchical tree structure, adopts BeliefPropagation method identification construction set type, obtains the BUILDINGS MODELS semantic tree;
(3) based on above-mentioned BUILDINGS MODELS semantic tree, induction rule, the combination model Component Gallery generates BUILDINGS MODELS.
2. the Chinese ancient architecture process modeling approach based on architectural drawing according to claim 1, it is characterized in that: repeat characteristics based on construction set in the described step (1), cut apart drawing, the method that obtains construction set is: after the filling drawing obtains the zone, mate identical zone in the horizontal direction and obtain the zone group, the repeat pattern that generates each zone group is then set, and merges according to described repeat pattern tree coupling to have the zone group of identical repeat pattern, thereby obtains construction set.
3. the Chinese ancient architecture process modeling approach based on architectural drawing according to claim 1, it is characterized in that: the top-down drawing of cutting apart in the described step (2), the hierarchical tree structure that obtains building, based on this hierarchical tree structure, adopt Belief Propagation method identification construction set type, the method that obtains the BUILDINGS MODELS semantic tree is: cut apart drawing in level and vertical direction, till the model component that obtains, obtain the model structure tree in (1) step; In the model structure tree, make up the dependence graph between the different component type, adopt Belief Propagation model of cognition assembly, obtain the model semantics tree; The construction method of described dependence graph is: the sibling of described model semantics tree constitutes chain successively and relies on, and has between the different sibling groups of same type assembly to have relation of interdependence; For non-leafy node, it is the highest then to choose multiplicity, and assembly that length is the longest and the assembly of having discerned constitute the summit of dependence graph as the representative of this non-leafy node.
4. the Chinese ancient architecture process modeling approach based on architectural drawing according to claim 1, it is characterized in that: described step is set based on model semantics in (3), induction rule, the combination model Component Gallery, the method that generates BUILDINGS MODELS is: based on the model semantics tree that obtains in the priori of Chinese ancient architecture and the step (2), conclude the model create-rule, generate BUILDINGS MODELS, and generate the different models that change based on the model create-rule in conjunction with the three-dimensional model Component Gallery.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN200910235591A CN101694727A (en) | 2009-09-29 | 2009-09-29 | Ancient Chinese construction process modeling method based on construction drawings |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN200910235591A CN101694727A (en) | 2009-09-29 | 2009-09-29 | Ancient Chinese construction process modeling method based on construction drawings |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN101694727A true CN101694727A (en) | 2010-04-14 |
Family
ID=42093696
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN200910235591A Pending CN101694727A (en) | 2009-09-29 | 2009-09-29 | Ancient Chinese construction process modeling method based on construction drawings |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101694727A (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103279983A (en) * | 2013-05-31 | 2013-09-04 | 西安理工大学 | China Tang dynasty style historic building modeling method |
| WO2014005415A1 (en) * | 2012-07-04 | 2014-01-09 | Thomson Licensing | System and method for multi-level repetitive structure based 3d model compression |
| CN104063896A (en) * | 2014-06-24 | 2014-09-24 | 南京大学 | 3D building model structure discovery method based on transformation space |
| CN104637090A (en) * | 2015-02-06 | 2015-05-20 | 南京大学 | Indoor scene modeling method based on single picture |
| CN104765614A (en) * | 2015-04-24 | 2015-07-08 | 广东小天才科技有限公司 | Color filling processing method and device |
| CN108363867A (en) * | 2018-02-09 | 2018-08-03 | 广州大学 | Building method for automatic modeling based on architectural drawing |
| CN110598634A (en) * | 2019-09-12 | 2019-12-20 | 山东文多网络科技有限公司 | Machine room sketch identification method and device based on graph example library |
-
2009
- 2009-09-29 CN CN200910235591A patent/CN101694727A/en active Pending
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2014005415A1 (en) * | 2012-07-04 | 2014-01-09 | Thomson Licensing | System and method for multi-level repetitive structure based 3d model compression |
| CN103279983A (en) * | 2013-05-31 | 2013-09-04 | 西安理工大学 | China Tang dynasty style historic building modeling method |
| CN103279983B (en) * | 2013-05-31 | 2016-01-27 | 西安理工大学 | The modeling method of China Tang dynasty style ancient building |
| CN104063896A (en) * | 2014-06-24 | 2014-09-24 | 南京大学 | 3D building model structure discovery method based on transformation space |
| CN104063896B (en) * | 2014-06-24 | 2016-08-17 | 南京大学 | A kind of three-dimensional building model structure based on transformation space finds method |
| CN104637090A (en) * | 2015-02-06 | 2015-05-20 | 南京大学 | Indoor scene modeling method based on single picture |
| CN104765614A (en) * | 2015-04-24 | 2015-07-08 | 广东小天才科技有限公司 | Color filling processing method and device |
| CN104765614B (en) * | 2015-04-24 | 2018-04-10 | 广东小天才科技有限公司 | Color in processing method and processing device |
| CN108363867A (en) * | 2018-02-09 | 2018-08-03 | 广州大学 | Building method for automatic modeling based on architectural drawing |
| CN108363867B (en) * | 2018-02-09 | 2021-04-20 | 广州大学 | Building automatic modeling method based on building drawing |
| CN110598634A (en) * | 2019-09-12 | 2019-12-20 | 山东文多网络科技有限公司 | Machine room sketch identification method and device based on graph example library |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101694727A (en) | Ancient Chinese construction process modeling method based on construction drawings | |
| Martinovic et al. | Bayesian grammar learning for inverse procedural modeling | |
| CN105931294B (en) | A method of BIM physical model is converted into detail GIS standardized model | |
| Kelly et al. | Interactive architectural modeling with procedural extrusions | |
| Yin et al. | Generating 3d building models from architectural drawings: A survey | |
| CN106558100B (en) | A kind of stratigraphic model method for automatic modeling based on borehole data | |
| CN103098100B (en) | Based on the three-dimensional model shape analysis method of perception information | |
| CN101763652B (en) | Three-dimensional framework fast extraction method based on branch feathers | |
| CN110992490B (en) | Method for automatically extracting indoor map based on CAD building plan | |
| CN104036550B (en) | Laser radar point-cloud interpretation and reconstruction method for building elevations on basis of shape semantics | |
| Hou et al. | Drawing-based procedural modeling of Chinese architectures | |
| CN105741348A (en) | Structure adaptive three-dimension model editing method | |
| Domínguez et al. | Semiautomatic detection of floor topology from CAD architectural drawings | |
| Zhi et al. | A graph-based algorithm for extracting units and loops from architectural floor plans for a building evacuation model | |
| CN106127857A (en) | Synthetic data drives the on-board LiDAR data modeling method with model-driven | |
| CN104318618B (en) | Three-dimensional sectioning method of generalized tri-prism spatial data model | |
| CN113570275A (en) | Water resource real-time monitoring system based on BIM and digital elevation model | |
| Boulch et al. | Semantizing complex 3D scenes using constrained attribute grammars | |
| Yang et al. | Semantic decomposition and recognition of indoor spaces with structural constraints for 3D indoor modelling | |
| CN101320487B (en) | Scene pretreatment method for fire disaster simulation | |
| Ripperda et al. | Reconstruction of façade structures using a formal grammar and rjmcmc | |
| CN114757864A (en) | Multi-level fine-grained image generation method based on multi-scale feature decoupling | |
| CN102074027B (en) | Two-dimensional geological model agglomerating method | |
| CN108561127A (en) | A kind of Formation pressure prediction method based on stochastic simulation | |
| Benes et al. | Urban brush: Intuitive and controllable urban layout editing |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C12 | Rejection of a patent application after its publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20100414 |