CN103902657B - Three-dimensional model retrieval method based on sketch - Google Patents
Three-dimensional model retrieval method based on sketch Download PDFInfo
- Publication number
- CN103902657B CN103902657B CN201410074654.9A CN201410074654A CN103902657B CN 103902657 B CN103902657 B CN 103902657B CN 201410074654 A CN201410074654 A CN 201410074654A CN 103902657 B CN103902657 B CN 103902657B
- Authority
- CN
- China
- Prior art keywords
- krd
- profile diagram
- feature
- pixel
- visual angle
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/583—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
- G06F16/5838—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content using colour
Landscapes
- Engineering & Computer Science (AREA)
- Library & Information Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Image Analysis (AREA)
- Processing Or Creating Images (AREA)
Abstract
The invention discloses a three-dimensional model retrieval method based on a sketch. According to the method, a new sketch structural characteristic is provided and applied to three-dimensional model retrieval, and therefore the effect of three-dimensional model retrieval based on the sketch is improved. The method includes the steps that firstly, view angle profile diagrams are extracted from three-dimensional models in a database, and then KRD characteristics are extracted from the view angle profile diagrams and stored in the database. After the KRD characteristics are extracted from a sketch input by a user, the view angle profile diagram similar to the sketch is found out from the database through the EMD characteristic matching method, and accordingly a three-dimensional model corresponding to the view angle profile diagram is determined. According to the method, the characteristics of the sketch in structure are exploited fully, and the method has good robustness and accuracy, meanwhile is not large in calculation amount and can completely meet the requirements of practical application.
Description
Technical field
The invention belongs to computer information retrieval field, more particularly to a kind of three-dimensional model search side drawn based on grass
Method.This method propose a kind of new grass and paint image characteristic extracting method, this feature is the excavation to careless drawing structure feature, is
Grass is drawn as the desk study of structured features, with certain novelty.
Background technology
In the past few years, with the development of computer hardware and graphics, 3D technology more and more occurs in crowd
It is multi-field, and wherein threedimensional model is even more used widely in fields such as film, game, animations.Nowadays, people can be in electricity
The three-dimensional scenic of magnificence true to nature, such as famous 3D film great works are seen in shadow《A Fanda》The three-dimensional scenic and scene of the inside
In role and scenery it is marvellous.In field of play,《Winning eleven》、《World of Warcraft》Etc. all used substantial amounts of three-dimensional
Model.
At present, there is the technology of many three-dimensional model search.These technologies can be broadly divided into based on the retrieval of keyword
And content-based retrieval.Key search is the label of threedimensional model in the keyword by matching user input and data base
To return target three-dimensional.The defect of this technology is to need to manually enter semanteme to each threedimensional model in data base
Label, and semantic label much can not meet demand of the designer when target three-dimensional is searched.
It is current technology wide variety of in three-dimensional scenic is built based on the three-dimensional model search technology of content.Wherein again
The retrieval based on existing model and the retrieval based on image can be divided into.User is needed to provide one based on the retrieval of existing model
Existing threedimensional model, is then matched by extracting the feature of threedimensional model with the threedimensional model in data base, is returned most
Close threedimensional model.Although can reach good retrieval effectiveness in this technical know-how, user must be in retrieval
The front condition for providing threedimensional model limits the practicality of this technology.I.e. user needs quick lookup when three-dimensional scenic is built
To target three-dimensional, and this technical scheme requires that user provides an existing threedimensional model.
The two-value grass for being primarily referred to as user input based on the image of the three-dimensional model search of image is drawn, and user only needs to defeated
Enter the two dimensional image of a description threedimensional model.Here two dimensional image is typically the grass drawing of user input, existing to be based on
Although the three-dimensional model search technology that grass is drawn fully has excavated the shape and textural characteristics of careless drawing picture, grass is but have ignored
The architectural feature of drawing picture, and architectural feature can give full expression to careless drawing picture.
The content of the invention
The purpose of the present invention is for the deficiencies in the prior art, there is provided a kind of three-dimensional model search side drawn based on grass
Method, the method has fully excavated the key structure distribution characteristicss of careless drawing, and works well.
Solve the technical scheme that its technical problem adopted include it is as follows：The present invention is including offline pretreatment stage and online
Processing stage.
Described offline pretreatment stage is comprised the following steps that：
Step 1. is rendered using Direct3D to each threedimensional model in data base, to rendering after threedimensional model
Multiple view directions are selected to be projected；Threedimensional model after rendering produces a profile through the projection of each view directions
Figure, so as to produce several profile diagrams, every profile diagram is named as visual angle profile diagram；
Step 2. is that every visual angle profile diagram extracts key structure distribution characteristicss KRD：
First, vectorized process is done to visual angle profile diagram using the method for least square, visual angle profile diagram is converted to arrow
Amount profile diagram；Described vector outline figure is the figure being made up of some line segments；Then according in vector outline figure adjacent two
The corner dimension of bar line segment finds out the key structure in vector outline figure, and records each key structure in vector outline figure
Position；Last feature KRD of the spatial distribution rectangular histogram using key structure under polar coordinate system as various visual angles profile diagram.
The definition of the key structure of the vector outline figure described in above-mentioned steps 2 is：Key structure is by vector outline figure
Adjacent two lines section composition, the wherein angle of this two lines section is more than 30 degree.
The process for carrying out vectorized process to image in above-mentioned steps 2 is as follows：
（1）If set S={ v1, v2 ..., vn } expression picture black pixel point sets, set T=L0, L1, L2 ...,
Ln } represent the straight line set that has been fitted, v represent the black pixel point in image, L represent by set S in some pixels
The straight line of composition.
（2）If set S is not sky, then select a pixel v' at random in set S, and will be pixel v' from collection
Close and remove in S；Simultaneously the point on the basis of pixel v', selects one and belongs to set S's in its 8 UNICOM's neighbor pixel points
Pixel v0, using pixel v' and pixel v0,2 points are connected the straight line L' to be formed as initial straight, if the equation of L' is ax+
By+c=0；Algorithm terminates if set S is for sky.
（3）Each pixel in traversal set S, finds the minimum pixel v of the distance of straight line L'kIf, pixel
Point vkDistance to straight line L' is less than threshold value, then by vkIn adding the pixel point set on straight line L', and utilize method of least square
Recalculate the equation of L', repeat the（3）Step；If pixel vkDistance to L' is more than threshold value, and L' is added in set T and jumped
To（2）Step.
Visual angle profile diagram feature KRD that step 3. extracts step 2 is stored in data base, forms feature database.
The step of described online treatment stage, is as follows：
Step 4. user is input into width grass on system drawing board and draws；
The grass of step 5. extraction step 4 is drawn as KRD features f, and this feature extracting method is identical with step 2 method；
Step 6. utilizes EMD algorithms to calculate grass and draws as each in the feature database of generation in KRD features f and step 3
The distance between visual angle profile diagram feature KRD, and n grass in small distance is drawn as KRD features f before being returned using heapsort.
EMD algorithms calculate grass and draw as each the visual angle profile diagram in the feature database of generation in KRD features f and step 3
The distance between feature KRD formula, it is as follows：
P, Q represent two characteristic vectors in above-mentioned formula, represent two visual angle profile diagram features KRD, and i represents P i-th dimensions
Feature p', j represents feature q' that Q jth is tieed up, fijRepresent the difference of p' and q', dijRepresent the distance of p' and q'.Calculated using EMD algorithms
The result for going out is less, illustrates that the distance between two features are less.EMD algorithms are the histogram feature sides of computer realm
Method, detail refer to pertinent literature.
Step 7. is drawn as KRD features f determine the three-dimensional for generating the profile diagram finally by the n grass that step 6 is obtained
Model.
The present invention has the beneficial effect that：
The present invention then extract careless drawing key structure on this basis as carrying out vector quantization by drawing to grass first, and
New feature is generated by fundamental of key structure, the method has fully excavated careless picture of drawing the characteristics of configuration aspects,
With good robustness and accuracy, while its amount of calculation less, can meet completely the demand of practical application.
Description of the drawings
Fig. 1 is the flow chart of the inventive method.
Specific embodiment
Below in conjunction with the accompanying drawings the invention will be further described.
As shown in figure 1, a kind of method of the three-dimensional model search drawn based on grass, including offline pretreatment stage and online
Retrieval phase.It is wherein pre- offline to comprise the following steps that：
Step 1. is rendered using Direct3D to each threedimensional model in data base, to rendering after threedimensional model
Multiple view directions are selected to be projected；Threedimensional model after rendering produces a profile through the projection of each view directions
Figure, so as to produce several profile diagrams, every profile diagram is named as visual angle profile diagram；
Step 2. is that every visual angle profile diagram extracts key structure distribution characteristicss KRD：
First, vectorized process is done to visual angle profile diagram using the method for least square, visual angle profile diagram is converted to arrow
Amount profile diagram；Described vector outline figure is the figure being made up of some line segments；Then according in vector outline figure adjacent two
The corner dimension of bar line segment finds out the key structure in vector outline figure, and records each key structure in vector outline figure
Position；Last feature KRD of the spatial distribution rectangular histogram using key structure under polar coordinate system as various visual angles profile diagram.
The definition of the key structure of the vector outline figure described in above-mentioned steps 2 is：Key structure is by vector outline figure
Adjacent two lines section composition, the wherein angle of this two lines section is more than 30 degree.
The process for carrying out vectorized process to image in above-mentioned steps 2 is as follows：
（1）If set S={ v1, v2 ..., vn } expression picture black pixel point sets, set T=L0, L1, L2 ...,
Ln } represent the straight line set that has been fitted, v represent the black pixel point in image, L represent by set S in some pixels
The straight line of composition.
（2）If set S is not sky, then select a pixel v' at random in set S, and will be pixel v' from collection
Close and remove in S；Simultaneously the point on the basis of pixel v', selects one and belongs to set S's in its 8 UNICOM's neighbor pixel points
Pixel v0, using pixel v' and pixel v0,2 points are connected the straight line L' to be formed as initial straight, if the equation of L' is ax+
By+c=0；Algorithm terminates if set S is for sky.
（3）Each pixel in traversal set S, finds the minimum pixel v of the distance of straight line L'kIf, pixel
Point vkDistance to straight line L' is less than threshold value, then by vkIn adding the pixel point set on straight line L', and utilize method of least square
Recalculate the equation of L', repeat the（3）Step；If pixel vkDistance to L' is more than threshold value, and L' is added in set T and jumped
To（2）Step.
Visual angle profile diagram feature KRD that step 3. extracts step 2 is stored in data base, forms feature database.
The step of online treatment stage of the present invention, is as follows：
Step 4. user is input into width grass on system drawing board and draws；
The grass of step 5. extraction step 4 is drawn as KRD features f, and this feature extracting method is identical with step 2 method；
Step 6. utilizes EMD algorithms to calculate grass and draws as each in the feature database of generation in KRD features f and step 3
The distance between visual angle profile diagram feature KRD, and n grass in small distance is drawn as KRD features f before being returned using heapsort.
EMD algorithms calculate grass and draw as each the visual angle profile diagram in the feature database of generation in KRD features f and step 3
The distance between feature KRD formula, it is as follows：
P, Q represent two characteristic vectors in above-mentioned formula, represent two visual angle profile diagram features KRD, and i represents P i-th dimensions
Feature p', j represents feature q' that Q jth is tieed up, fijRepresent the difference of p' and q', dijRepresent the distance of p' and q'.Calculated using EMD algorithms
The result for going out is less, illustrates that the distance between two features are less.EMD algorithms are the histogram feature sides of computer realm
Method, detail refer to pertinent literature.
Step 7. is drawn as KRD features f determine the three-dimensional for generating the profile diagram finally by the n grass that step 6 is obtained
Model.
Claims (1)
1. it is a kind of based on grass draw method for searching three-dimension model, it is characterised in that the method include offline pretreatment stage and
Processing stage line；
Described offline pretreatment stage is comprised the following steps that：
Step 1, each threedimensional model in data base is rendered using Direct3D, to rendering after threedimensional model select
Multiple view directions are projected；Threedimensional model after rendering produces a profile diagram through the projection of each view directions, from
And several profile diagrams are produced, every profile diagram is named as visual angle profile diagram；
Step 2, it is that every visual angle profile diagram extracts key structure distribution characteristicss KRD：
First, vectorized process is done to visual angle profile diagram using the method for least square, visual angle profile diagram is converted to vector wheel
Exterior feature figure；Described vector outline figure is the figure being made up of some line segments；Then according to adjacent two lines in vector outline figure
The corner dimension of section finds out the key structure in vector outline figure, and records each key structure in the position of vector outline figure
Put；Last feature KRD of the spatial distribution rectangular histogram using key structure under polar coordinate system as visual angle profile diagram；
The definition of the key structure of the vector outline figure described in above-mentioned steps 2 is：Key structure is by adjacent in vector outline figure
Two lines section is constituted, and wherein the angle of this two lines section is more than 30 degree；
The process for carrying out vectorized process to image in above-mentioned steps 2 is as follows：
(1) set set S={ v1, v2 ..., vn } and represent picture black pixel point set, set T={ L0, L1, L2 ..., Ln } table
Show the straight line set being fitted, v1, v2 ..., vn represents the 1st, 2 in image ..., n black pixel point, L0, L1,
L2 ..., Ln represent by set S in some pixels constitute the 1st, 2 ..., n+1 bar straight lines；
(2) if set S is not sky, then a random selected pixel v' in set S, and will be pixel v' from set S
In remove；Simultaneously the point on the basis of pixel v', in its 8 UNICOM's neighbor pixel points a point for belonging to set S is selected
V0, using pixel v' and pixel v0,2 points are connected the straight line L' to be formed as initial straight, if the equation of L' is ax+by+c=
0；Algorithm terminates if set S is for sky；
(3) each pixel in set S is traveled through, the minimum pixel v of straight line L' distances is foundkIf, pixel vkTo straight
The distance of line L' is less than threshold value, then by vkIn adding the pixel point set on straight line L', and counted again using method of least square
The equation of L' is calculated, repeats (3rd) step；If pixel vkDistance to L' is more than threshold value, and L' is added in set T and the is skipped to
(2) step；
Step 3, visual angle profile diagram feature KRD that step 2 is extracted are stored in data base, form feature database；
The step of described online treatment stage, is as follows：
Step 4, user are input into width grass on system drawing board and draw；
Step 5, the grass of extraction step 4 are drawn as KRD features f, and this feature extracting method is identical with step 2 method；
Step 6, using EMD algorithms calculate grass draw as in KRD features f and step 3 generation feature database in each visual angle
The distance between profile diagram feature KRD, and n visual angle profile diagram feature KRD in small distance before being returned using heapsort；
EMD algorithms calculate grass and draw as each the visual angle profile diagram feature in the feature database of generation in KRD features f and step 3
The distance between KRD formula, it is as follows：
P, Q represent two characteristic vectors in above-mentioned formula, represent two visual angle profile diagram features KRD, and i represents the feature of P i-th dimensions
P', j represent feature q' that Q jth is tieed up, fijRepresent the difference of p' and q', dijRepresent the distance of p' and q'；Calculated using EMD algorithms
As a result it is less, illustrate that the distance between two features are less；
Step 7, n visual angle profile diagram feature KRD obtained finally by step 6 determine the three-dimensional mould for generating the profile diagram
Type.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410074654.9A CN103902657B (en) | 2014-03-03 | 2014-03-03 | Three-dimensional model retrieval method based on sketch |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410074654.9A CN103902657B (en) | 2014-03-03 | 2014-03-03 | Three-dimensional model retrieval method based on sketch |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103902657A CN103902657A (en) | 2014-07-02 |
| CN103902657B true CN103902657B (en) | 2017-04-19 |
Family
ID=50993979
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410074654.9A Active CN103902657B (en) | 2014-03-03 | 2014-03-03 | Three-dimensional model retrieval method based on sketch |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103902657B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106484692B (en) * | 2015-08-25 | 2019-11-29 | 北京师范大学 | A kind of method for searching three-dimension model |
| CN106874350B (en) * | 2016-12-27 | 2020-09-25 | 合肥阿巴赛信息科技有限公司 | Diamond ring retrieval method and system based on sketch and distance field |
| CN112541092B (en) * | 2020-12-25 | 2022-04-22 | 华南理工大学 | Three-dimensional image contour retrieval method and system based on tangential domain and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101004748A (en) * | 2006-10-27 | 2007-07-25 | 北京航空航天大学 | Method for searching 3D model based on 2D sketch |
| CN101084498A (en) * | 2004-12-31 | 2007-12-05 | 富士通株式会社 | Search method of 3d model and device thereof |
| CN101281545A (en) * | 2008-05-30 | 2008-10-08 | 清华大学 | Three-dimensional model search method based on multiple characteristic related feedback |
| EP2284770A1 (en) * | 2009-07-06 | 2011-02-16 | Stavros Perantonis | Method and system for content-based search and retrieval of 3D models using panoramic views |
| CN102110159A (en) * | 2011-02-21 | 2011-06-29 | 清华大学 | CAD three-dimensional model retrieval method and system |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2004288170A (en) * | 2003-03-05 | 2004-10-14 | Olympus Corp | Three-dimensional model retrieval method and system |
| US20090157649A1 (en) * | 2007-12-17 | 2009-06-18 | Panagiotis Papadakis | Hybrid Method and System for Content-based 3D Model Search |
-
2014
- 2014-03-03 CN CN201410074654.9A patent/CN103902657B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101084498A (en) * | 2004-12-31 | 2007-12-05 | 富士通株式会社 | Search method of 3d model and device thereof |
| CN101004748A (en) * | 2006-10-27 | 2007-07-25 | 北京航空航天大学 | Method for searching 3D model based on 2D sketch |
| CN101281545A (en) * | 2008-05-30 | 2008-10-08 | 清华大学 | Three-dimensional model search method based on multiple characteristic related feedback |
| EP2284770A1 (en) * | 2009-07-06 | 2011-02-16 | Stavros Perantonis | Method and system for content-based search and retrieval of 3D models using panoramic views |
| CN102110159A (en) * | 2011-02-21 | 2011-06-29 | 清华大学 | CAD three-dimensional model retrieval method and system |
Non-Patent Citations (2)
| Title |
|---|
| 一种基于二维草图的三维CAD模型检索方法;黄明聪等;《锻压装备与制造技术》;20130630;第107-110页 * |
| 基于视图特征点分布的三维模型检索算法;潘翔等;《浙江工业大学学报》;20131215;第41卷(第6期);第641-645页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103902657A (en) | 2014-07-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Mehra et al. | Abstraction of man-made shapes | |
| CN104392486A (en) | Point-cloud scene rebuilding method | |
| CN104751511A (en) | 3D scene construction method and device | |
| CN103902657B (en) | Three-dimensional model retrieval method based on sketch | |
| CN107194984A (en) | Mobile terminal real-time high-precision three-dimensional modeling method | |
| CN102567465B (en) | Method for searching geometrical similarity of three-dimensional models on basis of compound eye vision | |
| Sun et al. | Quadratic terms based point-to-surface 3D representation for deep learning of point cloud | |
| CN108961385A (en) | A kind of SLAM patterning process and device | |
| Shu et al. | CFSA-Net: Efficient Large-Scale Point Cloud Semantic Segmentation Based on Cross-Fusion Self-Attention. | |
| Demir et al. | Proceduralization of buildings at city scale | |
| CN103366395B (en) | A kind of volume data non-photorealistic rendering method accelerated based on GPU | |
| Li et al. | Art3d: 3d gaussian splatting for text-guided artistic scenes generation | |
| Wu et al. | Building reconstruction from high-resolution multiview aerial imagery | |
| Liu et al. | Sherpa3d: Boosting high-fidelity text-to-3d generation via coarse 3d prior | |
| Shen et al. | Make-it-4d: Synthesizing a consistent long-term dynamic scene video from a single image | |
| An et al. | RS-AUG: Improve 3D object detection on LiDAR with realistic simulator based data augmentation | |
| CN102708583A (en) | Automatic match method of two-dimensional animation characters | |
| Li et al. | 3d shape reconstruction of furniture object from a single real indoor image | |
| Cheng et al. | An augmented reality image registration method based on improved ORB | |
| Zhu et al. | A 3D reconstruction method based on RGB-D camera | |
| Zhou et al. | Multi-attribute joint point cloud super-resolution with adversarial feature graph networks | |
| Ye | Evolution and Application of Artificial Intelligence Art Design Based on Machine Learning Algorithm | |
| Cheng | Retracted: Application of VR computer image technology in 3D virtualization of sports human science | |
| Kong et al. | DreamDrone | |
| Divya Udayan et al. | Constrained procedural modeling of real buildings from single facade layout |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |