CN102110159B - CAD three-dimensional model retrieval method and system - Google Patents
CAD three-dimensional model retrieval method and system Download PDFInfo
- Publication number
- CN102110159B CN102110159B CN 201110044079 CN201110044079A CN102110159B CN 102110159 B CN102110159 B CN 102110159B CN 201110044079 CN201110044079 CN 201110044079 CN 201110044079 A CN201110044079 A CN 201110044079A CN 102110159 B CN102110159 B CN 102110159B
- Authority
- CN
- China
- Prior art keywords
- view
- dimensional model
- descriptor
- dimension
- distance
- 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.)
- Expired - Fee Related
Links
- 238000000034 method Methods 0.000 title claims abstract description 44
- 239000011159 matrix material Substances 0.000 claims description 22
- 238000005070 sampling Methods 0.000 claims description 22
- 239000000284 extract Substances 0.000 claims description 11
- 239000012141 concentrate Substances 0.000 claims description 8
- 238000000605 extraction Methods 0.000 claims description 7
- 238000010606 normalization Methods 0.000 claims description 7
- 230000015572 biosynthetic process Effects 0.000 claims description 6
- 238000011960 computer-aided design Methods 0.000 abstract description 25
- 238000010586 diagram Methods 0.000 description 8
- 238000012545 processing Methods 0.000 description 7
- 238000004364 calculation method Methods 0.000 description 4
- 230000000295 complement effect Effects 0.000 description 2
- 230000008878 coupling Effects 0.000 description 2
- 238000010168 coupling process Methods 0.000 description 2
- 238000005859 coupling reaction Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 239000012467 final product Substances 0.000 description 2
- 238000004088 simulation Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Images
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Processing Or Creating Images (AREA)
Abstract
The invention provides a computer aided design (CAD) three-dimensional model retrieval method and a CAD three-dimensional model retrieval system. The method solves the problem that the retrieval performance is limited because the conventional three-dimensional model retrieval method depends on the accuracy of three views of a three-dimensional model. The method comprises the following steps of: inputting a two-dimensional draft; performing rough retrieval on the two-dimensional draft by using a first descriptor, and selecting n three-dimensional models sequenced in the front as rough retrieval results; and performing fine retrieval on the rough retrieval results by using a second descriptor, and outputting m three-dimensional models sequenced in the front as final retrieval results, wherein m is smaller than n. The input of a user is not limited any more in the method, the two-dimensional draft can be input by the user, and retrieval can be performed according to the draft input by the user. The input of the user is simplified, the retrieval request of the user can be processed in real time, and validity and accuracy of the retrieval results are ensured.
Description
Technical field
The present invention relates to retrieval technique, particularly relate to a kind of CAD (Computer Aided Design, computer-aided design (CAD)) method for searching three-dimension model and system.
Background technology
Retrieval is a more popular topic always, and along with the development of 3-D technology, increasing 3 d model library appears on network.In the CAD field, because the repeatability of a lot of cad models is very high, if can utilize fully existing cad model, can greatly save the design time of new cad model, this just makes extremely important that a kind of CAD 3D model index of exploitation system becomes.In the past few decades, have both at home and abroad much about the research based on the method for searching three-dimension model of example.But, based on the method for searching three-dimension model of example and be not suitable for all situations.For example, the user only can express with two-dimentional sketch at the general profile of only knowing at first the model that oneself needs, can not provide three-dimensional model to retrieve by system.In this case, a kind of method for searching three-dimension model based on two-dimentional sketch is more applicable.
At present, carry out the method for three-dimensional model search according to the two-dimentional sketch of user's submission, all require the user to submit simultaneously three width sketches to, correspond respectively to front elevation, left view, the vertical view of target three-dimensional model.Therefore when retrieval, all need Calculation of Three Dimensional model database accurately three-dimensional model towards, thereby determine front elevation, left view, the vertical view of three-dimensional model, then carry out Corresponding matching with three width sketches of user's input respectively.But well-known, accurately the Calculation of Three Dimensional model towards being a very hard problem, and whole retrieval performance depends on the accuracy of three views of three-dimensional model, therefore the performance of existing three-dimensional model searching system is subject to larger restriction.
Summary of the invention
The invention provides a kind of CAD 3D model retrieval method and system, depend on the accuracy of three views of three-dimensional model and the restricted problem of retrieval performance to solve existing method for searching three-dimension model.
In order to address the above problem, the invention discloses a kind of CAD 3D model retrieval method, comprising:
Input a two-dimentional sketch;
Each three-dimensional model in the cad model database is carried out the exemplary view collection extract operation, obtain the exemplary view collection of corresponding each three-dimensional model;
In the cad model database, the exemplary view collection of all three-dimensional models consists of an exemplary view database;
Utilize the first descriptor to carry out coarse search to described two-dimentional sketch, and choose n the forward three-dimensional model that sorts as the coarse search result;
Utilize the second descriptor to carry out meticulous retrieval from described coarse search result, export m the final result for retrieval of three-dimensional model conduct that sequence is forward; Wherein, m is less than n;
Wherein, described the first descriptor is shape profile descriptor, and described the second descriptor is shape context descriptor; Described coarse search and meticulous retrieval are all to retrieve from described exemplary view database;
Further, described exemplary view collection extracts to operate and comprises:
Step 1, the coordinate of normalization three-dimensional model, and three-dimensional model is placed in unit ball, the diverse location that virtual camera is placed in unit ball carries out projection to this three-dimensional model, each position projection obtains a two dimension view, and a plurality of two dimension views consist of one to two dimension view collection that should three-dimensional model;
Step 2, the Two-dimensional view is concentrated the proper vector of each two dimension view;
Step 21, each two dimension view that two dimension view is concentrated carries out exterior contour and detects, and exterior contour is carried out uniform sampling, obtains to set the sampled point of number, and utilizes described sample point coordinate computing center distance vector
The exterior contour P of a two dimension view is expressed as follows by the uniform sampling point of setting number:
P={p(u)=(x(i),y(i)),i∈0,1,…,N-1}
Wherein N represents the sum of uniform sampling point;
(x wherein
C, y
C) be the mean value of the coordinate of upper all the uniform sampling points of P;
Step 22 is with the centre distance vector
Be transformed into frequency domain through Fourier transform, and get the size of Fourier coefficient, obtain new vectorial F:
Being divided into M section and defining histogram h the span equalization of F:
h(k)={#F(u)∈bin(k),k∈1,2,…,M}
Wherein # represents projection operation, and bin represents section, and k is the subscript of section, and h (k) expression falls into F (u) number of k section;
Step 23, each two dimension view is calculated the proper vector H of shape profile descriptor:
Step 3 is concentrated the distance between view in twos according to the proper vector Two-dimensional view of two dimension view, obtains a distance matrix, concentrates from two dimension view according to described distance matrix and extracts exemplary view formation exemplary view collection;
Step 31 is concentrated the distance between view in twos according to the proper vector Two-dimensional view of two dimension view, obtains a distance matrix;
Suppose that the exterior contour p of two views and the shape profile descriptor proper vector of q are respectively H
pAnd H
q, the distance between these two views is:
Wherein
Step 32, according to described distance matrix, all two dimension views that utilize clustering method adaptively two dimension view to be concentrated are divided into the class of optimal number, and the corresponding center two dimension view of each class, and the view collection that center two dimension view corresponding to all classes forms is required exemplary view collection;
Further, described first descriptor that utilizes carries out coarse search to described two-dimentional sketch, comprising:
Calculate respectively the first descriptor proper vector of all views in described two-dimentional sketch and exemplary view database;
Utilize described the first descriptor proper vector to calculate the distance between all views in described two-dimentional sketch and exemplary view database, and sort from low to high according to distance, wherein the corresponding three-dimensional model of each view;
Further, described second descriptor that utilizes carries out meticulous retrieval from described coarse search result, comprising:
Calculate respectively the second descriptor proper vector of n the corresponding view of three-dimensional model in described two-dimentional sketch and coarse search result;
Utilize described the second descriptor proper vector to calculate in described two-dimentional sketch and coarse search result distance between n the corresponding view of three-dimensional model, and according to apart from sorting from low to high, wherein three-dimensional model of each view correspondence.
The present invention also provides a kind of CAD 3D model index system, comprising:
Load module is used for two-dimentional sketch of input;
The cad model database is used for the storage three-dimensional model; Off-line module, be used for each three-dimensional model of cad model database is carried out exemplary view collection extraction operation, obtain the exemplary view collection of corresponding each three-dimensional model, in the cad model database, the exemplary view collection of all three-dimensional models consists of an exemplary view database;
The coarse search module is used for utilizing the first descriptor to carry out coarse search to described two-dimentional sketch, and chooses n the forward three-dimensional model that sorts as the coarse search result;
Meticulous retrieval module utilizes the second descriptor to carry out meticulous retrieval from described coarse search result, exports m the final result for retrieval of three-dimensional model conduct that sequence is forward; Wherein, m is less than n;
The exemplary view database is for the exemplary view collection of all three-dimensional models of storage;
Wherein, described the first descriptor is shape profile descriptor, and described the second descriptor is shape context descriptor;
Described coarse search module and meticulous retrieval module are all to retrieve from described exemplary view database;
Further, described off-line module comprises:
The view collection obtains submodule, the coordinate that is used for the normalization three-dimensional model, and three-dimensional model is placed in unit ball, the diverse location that virtual camera is placed in unit ball carries out projection to this three-dimensional model, each position projection obtains a two dimension view, and a plurality of two dimension views consist of one to two dimension view collection that should three-dimensional model;
The proper vector calculating sub module is used for the proper vector that the Two-dimensional view is concentrated each two dimension view;
Exemplary view is extracted submodule, concentrates the distance between view in twos according to the proper vector Two-dimensional view of two dimension view, obtains a distance matrix, concentrates from two dimension view according to described distance matrix and extracts exemplary view formation exemplary view collection; Further, described proper vector calculating sub module comprises:
Subelement 1, each two dimension view that two dimension view is concentrated carry out exterior contour and detect, and exterior contour is carried out uniform sampling, obtain to set the sampled point of number, and utilize described sample point coordinate computing center distance vector
The exterior contour P of a two dimension view is expressed as follows by the uniform sampling point of setting number:
P={p(u)=(x(i),y(i)),i∈0,1,…,N-1}
Wherein N represents the sum of uniform sampling point;
(x wherein
C, y
C) be the mean value of the coordinate of upper all the uniform sampling points of P;
Subelement 2, centre distance is vectorial
Be transformed into frequency domain through Fourier transform, and get the size of Fourier coefficient, obtain new vectorial F:
Being divided into M section and defining histogram h the span equalization of F:
h(k)={#F(u)∈bin(k),k∈1,2,…,M}
Wherein # represents projection operation, and bin represents section, and k is the subscript of section, and h (k) expression falls into F (u) number of k section;
Subelement 3, each two dimension view is calculated the proper vector H of shape profie descriptor:
Further, described exemplary view extraction submodule comprises:
Subelement 4, in twos distance view between concentrated according to the proper vector Two-dimensional view of two dimension view obtain a distance matrix;
Suppose that the exterior contour p of two views and the shape profile descriptor proper vector of q are respectively H
pAnd H
q, the distance between these two views is:
Wherein
Subelement 5, according to described distance matrix, all two dimension views that utilize clustering method adaptively two dimension view to be concentrated are divided into the class of optimal number, and the corresponding center two dimension view of each class, the view collection that center two dimension view corresponding to all classes forms is required exemplary view collection;
Further, described coarse search module comprises:
The proper vector calculating sub module is used for calculating respectively the first descriptor proper vector of described two-dimentional sketch and all views of exemplary view database;
Apart from calculating sub module, be used for utilizing the distance between described the first descriptor proper vector described two-dimentional sketch of calculating and all views of exemplary view database, and sort from low to high according to distance, choose n the forward three-dimensional model that sorts as the coarse search result; Corresponding three-dimensional model of each view wherein;
Further, described meticulous retrieval module comprises:
The proper vector calculating sub module is used for calculating respectively described two-dimentional sketch and coarse search the second descriptor proper vector of n the corresponding view of three-dimensional model as a result;
Apart from calculating sub module, be used for utilizing described the second descriptor proper vector to calculate described two-dimentional sketch and the coarse search distance between n corresponding view of three-dimensional model as a result, and sort from low to high according to distance, export m the final result for retrieval of three-dimensional model conduct that sequence is forward; Corresponding three-dimensional model of each view wherein.
Compared with prior art, the present invention includes following advantage:
The present invention no longer includes restriction to user's input, and the user can input a width two dimension sketch, just can retrieve according to the sketch of user's input.At first utilize efficient shape profile descriptor to carry out coarse search, filter out most of undesirable three-dimensional model, and then utilize flexibly, reliably, the shape context descriptor of quantity of information abundance carries out meticulous retrieval in remaining three-dimensional model, thereby output in real time meets the CAD 3D model of customer requirements.The present invention has not only simplified user's input, can also process in real time user's retrieval request, and guarantees validity and the accuracy of result for retrieval.
Description of drawings
Fig. 1 is main thought schematic diagram of the present invention;
Fig. 2 is the process flow diagram of the described processed offline of the embodiment of the present invention;
Fig. 3 is the process flow diagram of the described online processing of the embodiment of the present invention;
Fig. 4 is the described result for retrieval schematic diagram of the embodiment of the present invention;
Fig. 5 is the structural drawing of the described a kind of CAD 3D model index of embodiment of the present invention system.
Embodiment
For above-mentioned purpose of the present invention, feature and advantage can be become apparent more, the present invention is further detailed explanation below in conjunction with the drawings and specific embodiments.
The present invention proposes a kind of new CAD 3D model retrieval method and system, and user's input is no longer included restriction, and the user can input a width two dimension sketch, and system just can retrieve according to the sketch that the user inputs.
With reference to Fig. 1, thinking of the present invention is as follows:
Be divided into off-line and the online two large divisions of processing, in processed offline, in the cad model database, each three-dimensional model extracts operation through the exemplary view collection, obtains an exemplary view collection and represents this three-dimensional model, and the exemplary view collection of all three-dimensional models consists of an exemplary view database; In online the processing, the user submits a two-dimentional sketch to system, system is retrieved according to the two-dimentional sketch that the user submits to, at first system utilizes shape profile descriptor to carry out coarse search, and then utilize shape context descriptor to carry out meticulous retrieval, result for retrieval is sorted from low to high by matching distance, the CAD 3D model of the setting quantity that output sequence at last is forward.
Be elaborated below by embodiment.
With reference to Fig. 2, it is the process flow diagram of the described processed offline of the embodiment of the present invention.
The purpose of processed offline is that its main thought is: from all directions, three-dimensional model is carried out projection, thereby obtain a two dimension view collection for each three-dimensional model for online the processing provides an exemplary view database; Then the two dimension view collection is carried out cluster, the center two dimension view of all classes consists of the exemplary view collection corresponding to three-dimensional model, and the exemplary view collection of all three-dimensional models consists of an exemplary view database.
Processed offline to a three-dimensional model specifically comprises the following steps:
Step 1, the coordinate of normalization three-dimensional model, and three-dimensional model is placed in unit ball, the diverse location that virtual camera is placed in unit ball carries out projection to this three-dimensional model, each position projection obtains a two dimension view, and a plurality of two dimension views consist of one to two dimension view collection that should three-dimensional model;
Step 1 specifically comprises:
Step 11, the coordinate of normalization three-dimensional model also is placed in unit ball, and the centre coordinate of described three-dimensional model overlaps with the centre coordinate of described unit ball;
Step 12, it is all 30 each intersection point places that write music gauze that virtual camera is placed in longitude and latitude interval on unit sphere, uses the OpenGL simulation to take pictures and obtains a two dimension view collection;
Certainly, the embodiment of the present invention does not limit and is necessary for 30 gauzes of writing music, and can be the interval of other angles; Must use the OpenGL simulation nor limit, can use other projecting methods, the present embodiment is only with this explanation for example.
Step 2, the Two-dimensional view is concentrated the proper vector of each two dimension view;
Step 2 specifically comprises:
Step 21, each two dimension view that two dimension view is concentrated carries out exterior contour and detects, and exterior contour is carried out uniform sampling, obtains to set the sampled point of number, and utilizes described sample point coordinate computing center distance vector
The exterior contour P of a two dimension view is expressed as follows by the uniform sampling point of setting number:
P={p(u)=(x(i),y(i)),i∈0,1,…,N-1}
Wherein N represents the sum of uniform sampling point;
(x wherein
C, y
C) be the mean value of the coordinate of upper all the uniform sampling points of P;
Step 22 is with the centre distance vector
Be transformed into frequency domain through Fourier transform, and get the size of Fourier coefficient, obtain new vectorial F:
Wherein u gets the size of Fourier coefficient;
Being divided into M (M is natural number, supposes M=20) individual section and defining histogram h the span equalization of F:
h(k)={#F(u)∈bin(k),k∈1,2,…,20}
Wherein # represents projection operation, and bin represents section, and k is the subscript of section, and h (k) expression falls into F (u) number of k section;
Step 23, each two dimension view is calculated the proper vector H of shape profile descriptor:
Step 3 is concentrated the distance between view in twos according to the proper vector Two-dimensional view of two dimension view, obtains a distance matrix, concentrates from two dimension view according to described distance matrix and extracts exemplary view formation exemplary view collection.
Step 3 specifically comprises:
Step 31 is concentrated the distance between view in twos according to the proper vector Two-dimensional view of two dimension view, obtains a distance matrix;
Suppose that the exterior contour p of two views and the shape profile descriptor proper vector of q are respectively H
pAnd H
q, the distance between these two views is:
Wherein
M gets 20;
Step 32, according to described distance matrix, all two dimension views that utilize clustering method adaptively two dimension view to be concentrated are divided into the class of optimal number, and the corresponding center two dimension view of each class, and the view collection that center two dimension view corresponding to all classes forms is required exemplary view collection.
The present embodiment can adopt affinity propagation clustering method, can certainly adopt other clustering methods.
Through above-mentioned processed offline, the exemplary view collection of all three-dimensional models in the cad model database will consist of an exemplary view database.Based on described exemplary view database, the user just can online retrieving CAD 3D model.
With reference to Fig. 3, it is the process flow diagram of the described online processing of the embodiment of the present invention.
Step 301 is inputted a two-dimentional sketch;
Be that the user submits a two-dimentional sketch of oneself drawing to system;
Step 302 utilizes the first descriptor to carry out coarse search to described two-dimentional sketch, and chooses n the forward three-dimensional model that sorts as the coarse search result;
Wherein, described the first descriptor can be shape profile (shape profile) descriptor; Described coarse search is to retrieve from the exemplary view database that above-mentioned processed offline obtains.
Step 302 specifically comprises:
If calculate shape profile descriptor proper vector, can calculate according to the method that step 2 in above-mentioned processed offline provides, be not described in detail in this.
Equally, also can carry out the calculating of view spacing according to the method that step 31 in above-mentioned processed offline provides, be not described in detail in this.
Step 303 utilizes the second descriptor to carry out meticulous retrieval from described n coarse search result, exports m the final result for retrieval of three-dimensional model conduct that sequence is forward;
Wherein, m and n are all more than or equal to 0, and m is less than n.
Described the second descriptor can be shape context (Shape context) descriptor, and described meticulous retrieval is also to retrieve from the exemplary view database that above-mentioned processed offline obtains.The value of n can be n=10m, and the embodiment of the present invention is not limited to this certainly.
Step 303 specifically comprises:
If the second descriptor is shape context descriptor, the computing method of shape context descriptor proper vector are different from the calculating of shape profile descriptor proper vector: shape profile descriptor is that corresponding view computation draws a proper vector, and shape context descriptor is that each sampled point on view outline is calculated a proper vector, thereby then obtains two distances between view by the matching distance that minimizes two view outline up-sampling points.To each sampled point calculated characteristics vector, the method for therefore utilizing proper vector to calculate the view spacing also is different from shape profile descriptor due to shape context descriptor, and also therefore the calculating of shape context descriptor is more accurately meticulous.The calculating of shape context descriptor can be adopted existing computing method, is not described in detail in this.
Through above-mentioned online processing, the user can input a width two dimension sketch, after the sketch that system inputs according to the user in the exemplary view database carries out coarse search and meticulous retrieval, just can find m optimum view of coupling, therefore the corresponding three-dimensional model of each view has just found the three-dimensional model that is complementary.Due to each three-dimensional model in described exemplary view database corresponding from the two dimension view of multiple directions projection, therefore the user only need input a width two dimension sketch, system just can retrieve the two dimension view with this two dimension sketch coupling, and then can find the three-dimensional model that is complementary.As from the foregoing, the present invention not only can export the CAD 3D model that meets customer requirements in real time, has also simplified user's input, and guarantees validity and the accuracy of result for retrieval.
With reference to Fig. 4, it is the described result for retrieval schematic diagram of the embodiment of the present invention.
At the Far Left of schematic diagram, the user submits a two-dimentional sketch of oneself drawing to, and system is retrieved according to the two-dimentional sketch that the user submits to automatically, and the CAD 3D model of the Optimum Matching of number (in figure being 12) is set in output.
Based on said method embodiment, the present invention also provides corresponding system embodiment.
With reference to Fig. 5, it is the structural drawing of the described a kind of CAD 3D model index of embodiment of the present invention system.
Described CAD 3D model index system mainly comprises:
Load module 51 is used for two-dimentional sketch of input;
Coarse search module 52 is used for utilizing the first descriptor to carry out coarse search to described two-dimentional sketch, and chooses n the forward three-dimensional model that sorts as the coarse search result;
Meticulous retrieval module 53 utilizes the second descriptor to carry out meticulous retrieval from described coarse search result, exports m the final result for retrieval of three-dimensional model conduct that sequence is forward; Wherein, m is less than n.
Wherein, described the first descriptor can be shape profile descriptor, and described the second descriptor can be shape context descriptor.
Above-mentioned load module 51, coarse search module 52 and meticulous retrieval module 53 belong to online processing section, and described CAD 3D model index system also comprises the processed offline part, therefore can also comprise:
Cad model database 54 is used for the storage three-dimensional model;
Off-line module 55, be used for each three-dimensional model of cad model database is carried out exemplary view collection extraction operation, obtain the exemplary view collection of corresponding each three-dimensional model, in the cad model database, the exemplary view collection of all three-dimensional models consists of an exemplary view database;
Exemplary view database 56 is for the exemplary view collection of all three-dimensional models of storage;
Described coarse search module and meticulous retrieval module are all to retrieve from described exemplary view database.
Further, described off-line module 55 can comprise:
The view collection obtains submodule, the coordinate that is used for the normalization three-dimensional model, and three-dimensional model is placed in unit ball, the diverse location that virtual camera is placed in unit ball carries out projection to this three-dimensional model, each position projection obtains a two dimension view, and a plurality of two dimension views consist of one to two dimension view collection that should three-dimensional model; Concrete, the view collection obtains the method that submodule can adopt step 1 in above-mentioned Fig. 2;
The proper vector calculating sub module is used for the proper vector that the Two-dimensional view is concentrated each two dimension view; Concrete, the proper vector calculating sub module can adopt the method for step 2 in above-mentioned Fig. 2;
The exemplary view extraction module is concentrated the distance between view in twos according to the proper vector Two-dimensional view of two dimension view, obtains a distance matrix, concentrates from two dimension view according to described distance matrix and extracts exemplary view formation exemplary view collection.Concrete, the exemplary view extraction module can adopt the method for step 3 in above-mentioned Fig. 2.
Described coarse search module 52 further also can comprise:
The proper vector calculating sub module is used for calculating respectively the first descriptor proper vector of described two-dimentional sketch and all views of exemplary view database; Concrete, the proper vector calculating sub module can adopt the method for step 2 in above-mentioned Fig. 2;
Apart from calculating sub module, be used for utilizing the distance between described the first descriptor proper vector described two-dimentional sketch of calculating and all views of exemplary view database, and sort from low to high according to distance, choose n the forward three-dimensional model that sorts as the coarse search result; Corresponding three-dimensional model of each view wherein.Concrete, also can adopt the method for step 31 in above-mentioned Fig. 2 apart from calculating sub module.
Described meticulous retrieval module 53 further also can comprise:
The proper vector calculating sub module is used for calculating respectively described two-dimentional sketch and coarse search the second descriptor proper vector of n the corresponding view of three-dimensional model as a result;
Apart from calculating sub module, be used for utilizing described the second descriptor proper vector to calculate described two-dimentional sketch and the coarse search distance between n corresponding view of three-dimensional model as a result, and sort from low to high according to distance, export m the final result for retrieval of three-dimensional model conduct that sequence is forward; Corresponding three-dimensional model of each view wherein.
But, if described the first descriptor is shape profile descriptor, described the second descriptor is shape context descriptor, shape profile descriptor is that corresponding view computation draws a proper vector, and shape context descriptor is that each sampled point on view outline is calculated a proper vector, thereby then obtains two distances between view by the matching distance that minimizes two view outline up-sampling points.
In sum, above-mentioned CAD 3D model index system no longer includes restriction to user's input, and the user can input a width two dimension sketch, and system just can retrieve according to the sketch that the user inputs, not only simplify user's input, can also guarantee validity and the accuracy of result for retrieval.
For said system embodiment, because it is substantially similar to embodiment of the method, so description is fairly simple, relevant part gets final product referring to the part explanation of Fig. 2 and embodiment of the method shown in Figure 3.
Each embodiment in this instructions all adopts the mode of going forward one by one to describe, and what each embodiment stressed is and the difference of other embodiment that between each embodiment, identical similar part is mutually referring to getting final product.
Above to a kind of CAD 3D model retrieval method provided by the present invention and system, be described in detail, used specific case herein principle of the present invention and embodiment are set forth, the explanation of above embodiment just is used for helping to understand method of the present invention and core concept thereof; Simultaneously, for one of ordinary skill in the art, according to thought of the present invention, all will change in specific embodiments and applications, in sum, this description should not be construed as limitation of the present invention.
Claims (2)
1. a CAD 3D model retrieval method, is characterized in that, comprising:
Input a two-dimentional sketch;
Each three-dimensional model in the cad model database is carried out the exemplary view collection extract operation, obtain the exemplary view collection of corresponding each three-dimensional model;
In the cad model database, the exemplary view collection of all three-dimensional models consists of an exemplary view database;
Utilize the first descriptor to carry out coarse search to described two-dimentional sketch, and choose n the forward three-dimensional model that sorts as the coarse search result;
Utilize the second descriptor to carry out meticulous retrieval from described coarse search result, export m the final result for retrieval of three-dimensional model conduct that sequence is forward; Wherein, m is less than n;
Wherein, described the first descriptor is shape profile descriptor, and described the second descriptor is shape context descriptor; Described coarse search and meticulous retrieval are all to retrieve from described exemplary view database;
Further, described exemplary view collection extracts to operate and comprises:
Step 1, the coordinate of normalization three-dimensional model, and three-dimensional model is placed in unit ball, the diverse location that virtual camera is placed in unit ball carries out projection to this three-dimensional model, each position projection obtains a two dimension view, and a plurality of two dimension views consist of one to two dimension view collection that should three-dimensional model;
Step 2, the Two-dimensional view is concentrated the proper vector of each two dimension view;
Step 21, each two dimension view that two dimension view is concentrated carries out exterior contour and detects, and exterior contour is carried out uniform sampling, obtains to set the sampled point of number, and utilizes described sample point coordinate computing center distance vector
The exterior contour P of a two dimension view is expressed as follows by the uniform sampling point of setting number:
Wherein N represents the sum of uniform sampling point;
(x wherein
c, y
c) be the mean value of the coordinate of upper all the uniform sampling points of P;
Step 22 is with the centre distance vector
Be transformed into frequency domain through Fourier transform, and get the size of Fourier coefficient, obtain new vectorial F:
Being divided into M section and defining histogram h the span equalization of F:
h(k)={#F(u)∈bin(k),k∈1,2,…,M}
Wherein # represents projection operation, and bin represents section, and k is the subscript of section, and h (k) expression falls into F (u) number of k section;
Step 23, each two dimension view is calculated the proper vector H of shape profile descriptor:
Step 3 is concentrated the distance between view in twos according to the proper vector Two-dimensional view of two dimension view, obtains a distance matrix, concentrates from two dimension view according to described distance matrix and extracts exemplary view formation exemplary view collection;
Step 31 is concentrated the distance between view in twos according to the proper vector Two-dimensional view of two dimension view, obtains a distance matrix;
Suppose that the exterior contour p of two views and the shape profile descriptor proper vector of q are respectively H
pAnd H
q, the distance between these two views is:
Step 32, according to described distance matrix, all two dimension views that utilize clustering method adaptively two dimension view to be concentrated are divided into the class of optimal number, and the corresponding center two dimension view of each class, and the view collection that center two dimension view corresponding to all classes forms is required exemplary view collection;
Further, described first descriptor that utilizes carries out coarse search to described two-dimentional sketch, comprising:
Calculate respectively the first descriptor proper vector of all views in described two-dimentional sketch and exemplary view database;
Utilize described the first descriptor proper vector to calculate the distance between all views in described two-dimentional sketch and exemplary view database, and sort from low to high according to distance, wherein the corresponding three-dimensional model of each view;
Further, described second descriptor that utilizes carries out meticulous retrieval from described coarse search result, comprising:
Calculate respectively the second descriptor proper vector of n the corresponding view of three-dimensional model in described two-dimentional sketch and coarse search result;
Utilize described the second descriptor proper vector to calculate in described two-dimentional sketch and coarse search result distance between n the corresponding view of three-dimensional model, and according to apart from sorting from low to high, wherein three-dimensional model of each view correspondence.
2. a CAD 3D model index system, is characterized in that, comprising:
Load module is used for two-dimentional sketch of input;
The cad model database is used for the storage three-dimensional model; Off-line module, be used for each three-dimensional model of cad model database is carried out exemplary view collection extraction operation, obtain the exemplary view collection of corresponding each three-dimensional model, in the cad model database, the exemplary view collection of all three-dimensional models consists of an exemplary view database;
The coarse search module is used for utilizing the first descriptor to carry out coarse search to described two-dimentional sketch, and chooses n the forward three-dimensional model that sorts as the coarse search result;
Meticulous retrieval module utilizes the second descriptor to carry out meticulous retrieval from described coarse search result, exports m the final result for retrieval of three-dimensional model conduct that sequence is forward; Wherein, m is less than n;
The exemplary view database is for the exemplary view collection of all three-dimensional models of storage;
Wherein, described the first descriptor is shape profile descriptor, and described the second descriptor is shape context descriptor;
Described coarse search module and meticulous retrieval module are all to retrieve from described exemplary view database;
Further, described off-line module comprises:
The view collection obtains submodule, the coordinate that is used for the normalization three-dimensional model, and three-dimensional model is placed in unit ball, the diverse location that virtual camera is placed in unit ball carries out projection to this three-dimensional model, each position projection obtains a two dimension view, and a plurality of two dimension views consist of one to two dimension view collection that should three-dimensional model;
The proper vector calculating sub module is used for the proper vector that the Two-dimensional view is concentrated each two dimension view;
Exemplary view is extracted submodule, concentrates the distance between view in twos according to the proper vector Two-dimensional view of two dimension view, obtains a distance matrix, concentrates from two dimension view according to described distance matrix and extracts exemplary view formation exemplary view collection; Further, described proper vector calculating sub module comprises:
Subelement 1, each two dimension view that two dimension view is concentrated carry out exterior contour and detect, and exterior contour is carried out uniform sampling, obtain to set the sampled point of number, and utilize described sample point coordinate computing center distance vector
The exterior contour P of a two dimension view is expressed as follows by the uniform sampling point of setting number:
P={p(u)=(x(i),y(i)),i∈0,1,…,N-1}
Wherein N represents the sum of uniform sampling point;
(x wherein
c, y
c) be the mean value of the coordinate of upper all the uniform sampling points of P;
Subelement 2, centre distance is vectorial
Be transformed into frequency domain through Fourier transform, and get the size of Fourier coefficient, obtain new vectorial F:
Being divided into M section and defining histogram h the span equalization of F:
h(k)={#F(u)∈bin(k),k∈1,2,…,M}
Wherein # represents projection operation, and bin represents section, and k is the subscript of section, and h (k) expression falls into F (u) number of k section;
Subelement 3, each two dimension view is calculated the proper vector H of shape profile descriptor:
Further, described exemplary view extraction submodule comprises:
Subelement 4, in twos distance view between concentrated according to the proper vector Two-dimensional view of two dimension view obtain a distance matrix;
Suppose that the exterior contour p of two views and the shape profile descriptor proper vector of q are respectively H
pAnd H
q, the distance between these two views is:
Subelement 5, according to described distance matrix, all two dimension views that utilize clustering method adaptively two dimension view to be concentrated are divided into the class of optimal number, and the corresponding center two dimension view of each class, the view collection that center two dimension view corresponding to all classes forms is required exemplary view collection;
Further, described coarse search module comprises:
The proper vector calculating sub module is used for calculating respectively the first descriptor proper vector of described two-dimentional sketch and all views of exemplary view database;
Apart from calculating sub module, be used for utilizing the distance between described the first descriptor proper vector described two-dimentional sketch of calculating and all views of exemplary view database, and sort from low to high according to distance, choose n the forward three-dimensional model that sorts as the coarse search result; Corresponding three-dimensional model of each view wherein;
Further, described meticulous retrieval module comprises:
The proper vector calculating sub module is used for calculating respectively described two-dimentional sketch and coarse search the second descriptor proper vector of n the corresponding view of three-dimensional model as a result;
Apart from calculating sub module, be used for utilizing described the second descriptor proper vector to calculate described two-dimentional sketch and the coarse search distance between n corresponding view of three-dimensional model as a result, and sort from low to high according to distance, export m the final result for retrieval of three-dimensional model conduct that sequence is forward; Corresponding three-dimensional model of each view wherein.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN 201110044079 CN102110159B (en) | 2011-02-21 | 2011-02-21 | CAD three-dimensional model retrieval method and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN 201110044079 CN102110159B (en) | 2011-02-21 | 2011-02-21 | CAD three-dimensional model retrieval method and system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN102110159A CN102110159A (en) | 2011-06-29 |
CN102110159B true CN102110159B (en) | 2013-05-22 |
Family
ID=44174320
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN 201110044079 Expired - Fee Related CN102110159B (en) | 2011-02-21 | 2011-02-21 | CAD three-dimensional model retrieval method and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN102110159B (en) |
Families Citing this family (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103065353B (en) * | 2012-12-22 | 2015-09-09 | 中国科学院深圳先进技术研究院 | Method for extracting characteristics of three-dimensional model and system, method for searching three-dimension model and system |
CN103927311B (en) * | 2013-01-11 | 2017-02-22 | 深圳市伊爱高新技术开发有限公司 | Historical track rapid retrieval method based on bisection method |
CN103902657B (en) * | 2014-03-03 | 2017-04-19 | 浙江大学 | Three-dimensional model retrieval method based on sketch |
CN109871761A (en) * | 2019-01-16 | 2019-06-11 | 中德(珠海)人工智能研究院有限公司 | A kind of identifying system of threedimensional model |
Citations (2)
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 |
CN101281545A (en) * | 2008-05-30 | 2008-10-08 | 清华大学 | Three-dimensional model search method based on multiple characteristic related feedback |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR100753536B1 (en) * | 2006-05-04 | 2007-08-30 | 주식회사 아이너스기술 | Method for detecting 2 dimension sketch data of source model data for 3 dimension reverse modeling |
-
2011
- 2011-02-21 CN CN 201110044079 patent/CN102110159B/en not_active Expired - Fee Related
Patent Citations (2)
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 |
CN101281545A (en) * | 2008-05-30 | 2008-10-08 | 清华大学 | Three-dimensional model search method based on multiple characteristic related feedback |
Non-Patent Citations (1)
Title |
---|
JP特开2007-299373A 2007.11.15 |
Also Published As
Publication number | Publication date |
---|---|
CN102110159A (en) | 2011-06-29 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN101350016B (en) | Device and method for searching three-dimensional model | |
CN101477529B (en) | Three-dimensional object retrieval method and apparatus | |
CN102024036B (en) | Three-dimensional object retrieval method and device based on hypergraphs | |
CN101201845B (en) | Method for searching three-dimensional model based on axis point-set hierarchical helix information | |
CN103559689B (en) | A kind of point cloud noise spot removal method | |
CN101446952A (en) | 3d model retrieval method and system | |
CN111027140B (en) | Airplane standard part model rapid reconstruction method based on multi-view point cloud data | |
CN101882150B (en) | Three-dimensional model comparison and search method based on nuclear density estimation | |
CN105205135A (en) | 3D (three-dimensional) model retrieving method based on topic model and retrieving device thereof | |
CN106844620B (en) | View-based feature matching three-dimensional model retrieval method | |
CN102110159B (en) | CAD three-dimensional model retrieval method and system | |
CN105868706A (en) | Method for identifying 3D model based on sparse coding | |
CN105320764A (en) | 3D model retrieval method and 3D model retrieval apparatus based on slow increment features | |
CN101359342A (en) | Search method for three-dimensional model of mixing characteristic based on feedback | |
CN113628329B (en) | Zero-sample sketch three-dimensional point cloud retrieval method | |
CN111597367B (en) | Three-dimensional model retrieval method based on view and hash algorithm | |
CN101894267B (en) | Three-dimensional object characteristic view selection method | |
Bespalov and et al. | Scale-space representation and classification of 3d models | |
CN106547825B (en) | Method for searching three-dimension model based on global characteristics and block feature | |
CN104111947B (en) | A kind of search method of remote sensing images | |
CN109658489B (en) | Three-dimensional grid data processing method and system based on neural network | |
Liu et al. | An approach to 3D building model retrieval based on topology structure and view feature | |
CN109272013A (en) | Method for measuring similarity based on study | |
CN115222988B (en) | Fine classification method for urban ground object PointEFF based on laser radar point cloud data | |
Chen et al. | 3D model retrieval based on fuzzy weighted shape distributions |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20130522 |
|
CF01 | Termination of patent right due to non-payment of annual fee |