CN106960032A - 3D shape expression and device - Google Patents
3D shape expression and device Download PDFInfo
- Publication number
- CN106960032A CN106960032A CN201710171841.2A CN201710171841A CN106960032A CN 106960032 A CN106960032 A CN 106960032A CN 201710171841 A CN201710171841 A CN 201710171841A CN 106960032 A CN106960032 A CN 106960032A
- Authority
- CN
- China
- Prior art keywords
- shape
- minor structure
- segmentation
- expression
- mixed type
- 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.)
- Granted
Links
- 230000011218 segmentation Effects 0.000 claims abstract description 49
- 239000000284 extract Substances 0.000 claims abstract description 8
- 238000000605 extraction Methods 0.000 claims description 10
- 210000000988 bone and bone Anatomy 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 3
- 238000000034 method Methods 0.000 abstract description 14
- 238000010586 diagram Methods 0.000 description 7
- 230000008569 process Effects 0.000 description 6
- 230000006870 function Effects 0.000 description 4
- 238000000513 principal component analysis Methods 0.000 description 4
- 230000000007 visual effect Effects 0.000 description 4
- 244000062793 Sorghum vulgare Species 0.000 description 3
- 235000019713 millet Nutrition 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000006399 behavior Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000004590 computer program Methods 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 238000013479 data entry Methods 0.000 description 1
- 238000005315 distribution function Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 239000000686 essence Substances 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000035800 maturation Effects 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
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
-
- 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
- G06F16/5854—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 shape and object relationship
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Library & Information Science (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Apparatus For Radiation Diagnosis (AREA)
- Image Analysis (AREA)
Abstract
The embodiments of the invention provide a kind of 3D shape expression and device.Wherein method includes step：Extract the mixed type skeleton of 3D shape；By the segmentation to the mixed type skeleton, the segmentation of the 3D shape is obtained；According to the 3D shape of segmentation, the minor structure of the 3D shape is obtained；And according to the minor structure of the 3D shape, the expression of the 3D shape is set up using bag of words.The embodiment of the present invention, can realize the succinct of 3D shape and efficient expression.
Description
Technical field
The present invention relates to graphics techniques, more particularly to a kind of 3D shape expression and device.
Background technology
The geological information of 3D shape obtains the gradual perfection of equipment and the maturation of three-dimensional modeling mode, allows 3D shape
Quantity obtained great lifting, this also proposes higher requirement to the retrieval and contrast of 3D shape.Based on such
Observation, it is the key solved the problems, such as that expression how is efficiently carried out to 3D shape.
In recent years, to 3D shape expression and grinding using different 3D shape expression progress Shape-memory behavior correlations
Study carefully achievement to be on the increase.And in the prior art, the expression to 3D shape is general by the way of global or local feature, use
Different Feature Descriptors go describe 3D shape, for example by the volume of 3D shape, area, Fourier Transform Coefficients and its
His statistics describes 3D shape as global characteristics, or based on the distance between random table millet cake in 3D shape,
Different 3D shapes, etc. are expressed in the distribution of angle, area and volume.
In a word, in existing scheme, the expression way to 3D shape can be divided mainly into three major types：(1) table of feature based
Reach；(2) expression based on figure；(3) expression based on visual angle.And the expression of feature based simply merely considers shape face
Geometric attribute, not in view of the overall structure of 3D shape, and people to the direct feel of shape often from structure
On, rather than in details.And the mode based on figure then expresses a 3D shape only by connection figure, to be done more using it
Further contrast or retrieval application, often also needs to extra definition and calculating.And for the expression way based on visual angle, its
Core concept is that each visual angle of 3D shape is captured with multiple two dimensional images, but often in order to be able to comprehensively capturing as far as possible
The information of 3D shape is, it is necessary to substantial amounts of two-dimension picture.If two-dimension picture quantity, which is crossed, can cause at least loss much in shape
Details, so as to influence shape expression of results.Moreover, while these two dimensional images are obtained, in addition it is also necessary to these two-dimension pictures
Calculate specific description, it is impossible to simply these pictures are directly applied in the types of applications of 3D shape.
Therefore, the problem of existing expression way to 3D shape has certain, there is worth improved space.
The content of the invention
The embodiments of the invention provide a kind of 3D shape expression way and device, it can be come in simple and efficient mode
Express 3D shape.
The embodiments of the invention provide a kind of 3D shape expression, including step：Extract the mixed type of 3D shape
Skeleton；By the segmentation to the mixed type skeleton, the segmentation of the 3D shape is obtained；According to the 3D shape of segmentation, obtain
To the minor structure of the 3D shape；And according to the minor structure of the 3D shape, described three are set up using bag of words
Tie up the expression of shape.
Wherein, the mixed type skeleton for extracting 3D shape, including：The surface of the 3D shape is sampled,
Obtain sampled point；And the sampled point is expressed again, to obtain the mixed type comprising one-dimensional curve and two-dimensional slice
Skeleton.
Wherein, the segmentation by the mixed type skeleton, obtains the segmentation of the 3D shape, including：Segmentation
The mixed type skeleton；According to the corresponding relation between the mixed type skeleton and the sampled point, by the mixed type skeleton
Segmentation obtain the segmentation of the 3D shape.
Wherein, the 3D shape according to segmentation, obtains the minor structure of the 3D shape, including：By the segmentation
3D shape, obtain multiple parts of the 3D shape；Create the connection figure of the multiple part of connection；Extract the company
Subgraph in map interlinking, is used as the minor structure of the 3D shape.
Wherein, the minor structure according to the 3D shape, the table of the 3D shape is set up using bag of words
Reach, including：Each candidate's minor structure that the minor structure of the 3D shape and candidate's minor structure are concentrated is matched, to determine
The frequency that each described candidate's minor structure occurs in the 3D shape；According to each described candidate's minor structure in the three-dimensional
The frequency occurred in shape, creates the term vector of the 3D shape；The term vector is normalized, to obtain described three
Tie up the bag of words expression of shape.
Wherein, it is described to be set up using bag of words before the expression of the 3D shape, also include：Create the candidate
Minor structure collection；Wherein, establishment candidate's minor structure collection includes：Obtain the son for all 3D shapes that input data is concentrated
Structure；Determine the similitude between the minor structure of the acquisition；According to the similitude between the minor structure of the acquisition, from described
Candidate's minor structure is selected in the minor structure of acquisition, to constitute candidate's minor structure collection.
Wherein, the similitude between the minor structure for determining the acquisition, including：Define the acquisition minor structure it
Between the kernel of graph；According to the kernel of graph, to determine the similitude between the minor structure obtained.
Wherein, the kernel of graph between the minor structure for defining the acquisition, including：Definition node core and side core；Wherein, institute
Stating node core is：The side core is：
The knode(ni,nj) represent node core, kedge(ei,ej) represent side core, ni,njRepresent node, hiAnd hjIt is by section respectively
Point niAnd njThe geometric properties histogram of part be formed by connecting, D is hiAnd hjNormalization it is related,It is any two couples of hi
And hjApart from D (hi,hj) maximum, uiAnd ujAll-pair is connected in two parts on angle and distance vertically upward
The two-dimensional histogram of composition.
The embodiments of the invention provide a kind of 3D shape expression device, including：Skeleton extraction module, for extracting three-dimensional
The mixed type skeleton of shape；Split module, for by the segmentation to the mixed type skeleton, obtaining point of the 3D shape
Cut；Minor structure extraction module, for the 3D shape according to segmentation, obtains the minor structure of the 3D shape；And expression mould
Block, for the minor structure according to the 3D shape, the expression of the 3D shape is set up using bag of words.
Wherein, the skeleton extraction module, specifically for：The surface of the 3D shape is sampled, sampled
Point；And the sampled point is expressed again, to obtain the mixed type skeleton comprising one-dimensional curve and two-dimensional slice.
Wherein, the segmentation module, specifically for splitting the mixed type skeleton；And according to the mixed type skeleton with
Corresponding relation between the sampled point, the segmentation of the 3D shape is obtained by the segmentation of the mixed type skeleton.
The beneficial effect of the embodiment of the present invention is：
The embodiment of the present invention, for 3D shape, establishes the bag of words expression based on minor structure, and such a expression way has
The features such as having simple and efficient.
Brief description of the drawings
Fig. 1 is the schematic flow sheet of the embodiment of the 3D shape expression of the embodiment of the present invention；
Fig. 2 is the schematic flow sheet of the embodiment of step 101 in Fig. 1；
Fig. 3 is the schematic flow sheet of the embodiment of step 102 in Fig. 1；
Fig. 4 is the schematic flow sheet of the embodiment of step 103 in Fig. 1；
Fig. 5 is the schematic flow sheet of the embodiment of step 104 in Fig. 1；
Fig. 6 is the schematic diagram of the mixed type skeleton of 3D shape；
Fig. 7 (a)~(c), which is respectively three, is used for the local feature schematic diagram that mixed type skeleton is split；
Fig. 8 is the process schematic for creating 3D shape graph structure；
Fig. 9 is the bag of words expression schematic diagram of 3D shape；
Figure 10 is the schematic diagram for carrying out 3D shape retrieval；
Figure 11 is the structural representation of the embodiment of the 3D shape expression device of the embodiment of the present invention.
Embodiment
In order that technical problem solved by the invention, technical scheme and beneficial effect are more clearly understood, below in conjunction with
Drawings and Examples, the present invention will be described in further detail.It should be appreciated that specific embodiment described herein is only used
To explain the present invention, it is not intended to limit the present invention.
As shown in figure 1, being the schematic flow sheet of the embodiment of the 3D shape expression of the embodiment of the present invention, it includes
Following steps：
Step 101：Extract the mixed type skeleton of 3D shape.
Step 102：By the segmentation to mixed type skeleton, the segmentation of 3D shape is obtained.
Step 103：According to the 3D shape of segmentation, the minor structure of 3D shape is obtained.And
Step 104：According to the minor structure of 3D shape, the expression of 3D shape is set up using bag of words.
Wherein, the mixed type skeleton of step 101 includes：One-dimensional curve and two-dimensional slice.Wherein, as shown in Fig. 2 being step
The schematic flow sheet of 101 embodiment.Specifically, in order to extract mixed type skeleton, first, the surface of 3D shape is carried out
Sampling, obtains multiple sampled points (step 201)；Then, all sampled points are expressed again, to obtain including one-dimensional curve
With the mixed type skeleton (step 202) of two-dimensional slice.Wherein, in step 201, Poisson can be used to the surface of 3D shape
Disk sample mode is sampled, and wherein sampled point is set as about 20000.Wherein, the sampled point obtained by step 201 is
Surface point, and in step 202., these sampled points can be expanded to depth point, and by each surface point and 3D shape
The corresponding skeletal point in portion is combined, and by the arrangement of the point on optimised shape surface and skeleton, makes direction and the table of depth point line
The normal vector of millet cake is consistent, and the convergence of final optimization pass function will obtain a mixed type being made up of one-dimensional curve and two-dimensional slice
Skeleton.For example, as shown in fig. 6, being the schematic diagram of the mixed type skeleton of the 3D shape of chair.
Wherein, as shown in figure 3, being the schematic flow sheet of the embodiment of step 102.Specifically, first, mixed type bone is split
Frame (step 301)：Then, can be with by the segmentation of mixed type skeleton according to the corresponding relation between mixed type skeleton and sampled point
Obtain the segmentation (step 302) of 3D shape.Wherein, in step 301, to each mixed type skeleton in mixed type internal frame
Point, calculates three local features for being based on PCA (Principal Component Analysis, principal component analysis).And calculate
These three features each put are, it is necessary to which a geodetic neighborhood union for choosing the point calculates its eigenvalue λ1≥λ2≥λ3>=0, and it is fixed
Justice：
Wherein, L, P, S are three above-mentioned local features, respectively describe the linear of the vertex neighborhood shape, flatness,
The degree of spherical property；For example, it is the schematic diagram of above three local feature, i.e. Fig. 7 respectively as shown in Fig. 7 (a)~(c)
(a)~(c) shows the linear of region, flatness and spherical property in visual mode respectively.Wherein, above-mentioned local feature
The cluster process of segmentation mixed type skeleton will be used for.Also, in embodiments of the present invention, have more semanteme to obtain one
3D shape segmentation, semi-supervised Spectral Clustering can be used, and combine to the man-machine interactively of mixed type skeleton, be mixed
The segmentation of mould assembly skeleton.Wherein, allow to set " must connect " to the difference on mixed type skeleton in the process of man-machine interactively
Two kinds of " can not connect " constraint guides the result of segmentation.
As shown in figure 4, being the schematic flow sheet of the embodiment of step 103.Specifically, in Fig. 4, first, by splitting
3D shape, obtains multiple parts (step 401) of 3D shape；Then, the connection figure (step of the plurality of part of connection is created
402)；Finally, the subgraph in connection figure is extracted, the minor structure (step 403) of 3D shape is used as.Wherein, obtain in step 402
To connection figure node be 3D shape in part, and create connection figure when, if any in any two part
Point to distance, less than the 2% of whole 3D shape bounding box catercorner length, then two parts are connected by a line.
Wherein, in step 403, it is n=1 ... connected node number in the connection figure, 5 subgraph is extracted, so as to obtain the three-dimensional
The corresponding minor structure of shape, minor structure herein can represent that wherein n value is not also limited to using a series of geometric properties
In above-mentioned example.Wherein, as shown in figure 8, it illustrates the process for creating 3D shape graph structure, left side in wherein Fig. 8
It is shaped as combining being shaped as by mixed type bone in the middle of in the segmentation schematic diagram that mixed type skeleton is obtained after man-machine interactively, Fig. 8
Each portion for being shaped as connecting 3D shape on right side in the segmentation for contacting acquisition 3D shape of frame and shape table millet cake, Fig. 8
Part sets up connection figure.
As shown in figure 5, being the schematic flow sheet of the embodiment of step 104.Specifically, first, the son of 3D shape is tied
Structure is matched with each candidate's minor structure that candidate's minor structure is concentrated, to determine that each candidate's minor structure goes out in 3D shape
Existing frequency (step 501)；Then, the frequency occurred according to each candidate's minor structure in 3D shape, creates 3D shape
Term vector (step 502)；Finally, term vector is normalized, to obtain the expression (step 503) of 3D shape.
Wherein, candidate's minor structure collection in step 501 can be pre-created, all three-dimensionals for example concentrated by input data
The information of shape is created.Specifically, the minor structure of all 3D shapes of input data concentration can be obtained, in this herein
Acquisition modes can for example include：The mixed type skeleton of each 3D shape is extracted respectively and is split, and then obtains three
The segmentation of shape is tieed up, so as to obtain the minor structure of 3D shape；It is then determined that the similitude between the minor structure obtained；Finally,
According to the similitude between the minor structure of acquisition, candidate's minor structure is selected from the minor structure of acquisition, to constitute candidate's
Structure collection.In addition, when selecting candidate's minor structure, the quantity of candidate's minor structure can also be determined simultaneously.
Wherein, in order to select representative minor structure, two interstructural distances of son can be used.And calculate two portions
Distance between number of packages identical minor structure, can first pass through the kernel of graph (graph kernel) calculating and obtain two with same node point number
Subgraph similitude so that the similitude between obtaining minor structure.The process wherein calculated can use node core (node
Kernel) and side core (edge kernel), in the present embodiment, node core is defined as：
Wherein, hiAnd hjIt is on node n respectivelyiAnd njAll geometric properties histograms of part be formed by connecting, these
Geometric properties include shape diameter function (Shape Diameter Function) and three offices based on PCA above used
Portion's feature, and the dimension of each feature histogram is 16.D is hiAnd hjNormalization correlation (normalized
correlation).SymbolIt is any two couples of hiAnd hjApart from D (hi,hj) maximum.
Wherein, side core is used for the similitude for capturing two pairs of connected components, is defined as：
Wherein, uiAnd ujIt is connected to all-pair in two parts straight on the two dimension that angle and distance vertically upward is constituted
Fang Tu.The two are characterized in vertical with 3D shape to the distance of the line segment constituted, and the line segment by calculating these
The angle upwardly-directed constituted is obtained.The similitude of two minor structures is that figure step (graph walks) is less than or equal to p in the kernel of graph
Similitude be added summation gained, wherein p be minor structure node quantity.The similitude of i.e. two minor structures is by a consideration
Node core is calculated with the figure step core (graph walk kernel) of side core and obtained.
, just can be from all initial minor structure (sons of i.e. above-mentioned all 3D shapes after the similitude of minor structure has been defined
Structure) in extract candidate minor structure collection C, to set up the dictionary in bag of words.For initial minor structure collection, it is necessary to solve
A problem be that wherein there are a large amount of similar minor structures.The main cause for producing this problem is 3D shape minor structure
Distinctiveness it is not high.Therefore, in order to avoid a large amount of similar or unrelated minor structures of processing, the present invention is by minor structure phase
Density analysis is carried out like property space, selection obtains candidate's minor structure collection.The density of all minor structures is calculated, is then only protected
Those density are stayed to be in the minor structure of peak value.The density peaks and the cluster centre of the initial minor structure collection with same node point number
It is associated.Surrounded so only picking out in similar spatial by similar minor structure, and density is in the minor structure of peak value, is just avoided that
The unnecessary similar minor structure of processing.Simultaneously as these minor structures are density peaks, they can frequently occur on minor structure collection it
In.
, can be using the method clustered in order to calculate density in the way of relative robust.Before cluster, it is assumed that cluster centre
It can be surrounded by the minor structure of neighbouring less dense value, and relatively remote distance is kept with the minor structures of other more high intensity values.
Specifically, two minor structures are calculated by the kernel of graph (graph kernel) definition of similitude between above-mentioned minor structure first
Distance, and be designated as dij.Then, by some minor structure eiLocal density ρiIt is defined as：
ρi=∑jχ(dij-dc),
Wherein, x is worked as<When 1, otherwise χ (x)=1 is 0；dcTo block distance (cutoff distance), d is set herec
The 2%th value after being sorted from small to large for distance value between all minor structures.Next, again by a minor structure to it is other more
High density minor structure apart from δiIt is defined as：
And density highest minor structure, as a special case, it will be defined as δ to the distance of other high density minor structuresi
=maxjdij.Finally, to all minor structures according to δiSequence, and pick out before maximum of which K minor structure to obtain most
Big distance value δi.Wherein, minor structure number K can rule of thumb be set, such as according to variance γi=ρiδiTo determine.Pass through
With upper type, candidate's minor structure collection C is finally just obtained.It should be noted that the above method is for including number of components
Different minor structures are handled respectively.
After candidate's minor structure collection C is obtained, according to the concept of bag of words, candidate's minor structure collection C is represented with vectorial t
The frequency occurred in each 3D shape, is referred to as the term vector (term vector) of this 3D shape.T is a m
Dimensional vector, wherein m are the numbers of minor structure in candidate's minor structure collection C.In t it is every it is one-dimensional be counted certain candidate's minor structure appearance
Number of times in the 3D shape, and be then normalized with the quantity of all minor structures in 3D shape.If certain
One related sub-structures are not appeared in 3D shape, then corresponding dimension values are then zero.
In order to create term vector to given 3D shape S, using with obtain initial minor structure phase as flow from three
Dimension extracts minor structure in shape.Then, only retain the minor structure related to candidate's minor structure collection C, and count detected son
Structure sets the value in term vector corresponding to related sub-structures.In order to find in 3D shape S with phase in candidate's minor structure collection C
As minor structure, it is necessary in view of the similitude between minor structure.For the minor structure in figure, a son in 3D shape is given
A structure s ∈ S and candidate minor structure c ∈ C, if the core distance (kernel between minor structure s and minor structure c
Distance) it is less than threshold taus, then it is assumed that the two minor structures are similar.In order to obtain on each candidate's minor structure c's
Threshold value is, it is necessary to first calculate the core distance between minor structure c and other minor structures, and foundation obtains a histogram.Then, will
During the histogram-fitting is distributed to Beta, and set τsValue is that inverse Cumulative Distribution Function value is 0.05 part, it means that other
Minor structure has 95% to be more than τ into the distance of the minor structures。
As shown in figure 9, illustrate a three-dimensional chair shape and a three-dimensional stool shape, a and b represent respectively this two
The bag of words expression of individual 3D shape, i.e. term vector, and each the term vector t of 3D shape is represented with the m histograms tieed up.Directly
Same position in square figure correspond to common candidate's minor structure.In fig .9, the two 3D shapes can be observed all to have
There is common " cushion and supporting leg " minor structure.And for stool shape, it belongs to different shape classes from chair shape
Type, and without " handrail " minor structure in chair, therefore, tieed up accordingly in histogram and do not have corresponding data.
Existing mode is high to data entry requirement when expressing 3D shape；And the embodiment of the present invention, to input
3D shape strong adaptability.Meanwhile, 3D shape expression of the invention is the expression based on minor structure, therefore can more be captured
The global framework of 3D shape, so that adapting to the three-dimensional shape data with noise and missing problem.In addition, in existing
3D shape expression can use two-dimension picture, but generally for the form for intactly capturing 3D shape, it is necessary to multiple angles
Degree, causes to use substantial amounts of two-dimension picture；And the bag of words expression that the present invention is utilized is finally a simple histogram, and
The structure of 3D shape can be reflected well, its application extension is more convenient.In addition, to the final table of 3D shape in the present invention
Up to result contrasted or retrieved etc. apply when, can directly enter the calculating of row distance, it is not necessary to make further on the basis of this again
Processing and conversion.
The above-mentioned principle to the embodiment of the present invention is illustrated, the application note illustrated below to the embodiment of the present invention.
The 3D shape expression way of the embodiment of the present invention, can apply to retrieval, comparison, classification and the knowledge of 3D shape
Deng not field.Illustrated below by taking retrieval as an example come the application to the embodiment of the present invention.Newly inputted for given one
3D shape, splits and extracts " initial minor structure " in the 3D shape, then matched according to the similitude between minor structure
Candidate's minor structure, obtains the frequency of occurrences of all candidate's minor structures, so as to complete to express the 3D shape using bag of words
Process.Then, the distance between each 3D shape expression in the expression and data set of the 3D shape that calculating is newly inputted
Value, and according to the distance value of calculating, obtain retrieval result；For example, apart from smaller, represent that two 3D shapes are more similar.
For example, being the result figure of 3D shape retrieval application as shown in Figure 10, wherein being the 3D shape of retrieving in the square frame of left side
Input, and be then by retrieving the result returned in the square frame of right side.In Fig. 10, preceding ten corresponding three of minimum range have been taken
Dimension shape is used as the retrieval result of return, and finally, their contour structures also with the 3D shape of retrieval also most
To be similar.
The embodiment of the method to the embodiment of the present invention is illustrated above, and the device of the embodiment of the present invention is implemented below
Example is illustrated.
As shown in figure 11, be the present invention 3D shape expression device embodiment structural representation.It includes：Skeleton
Extraction module 111, the mixed type skeleton for extracting 3D shape；Split module 112, for by the mixed type skeleton
Segmentation, obtain the segmentation of the 3D shape；Minor structure extraction module 113, for the 3D shape according to segmentation, obtains institute
State the minor structure of 3D shape；And expression module 114, for the minor structure according to the 3D shape, utilize bag of words
To set up the expression of the 3D shape.Wherein, skeleton extraction module 111, specifically for：The surface of the 3D shape is entered
Row sampling, obtains sampled point；And the sampled point is expressed again, to obtain comprising one-dimensional curve and two-dimensional slice
Mixed type skeleton.Wherein, module 112 is split, specifically for splitting the mixed type skeleton；And according to the mixed type skeleton
With the corresponding relation between the sampled point, the segmentation of the 3D shape is obtained by the segmentation of the mixed type skeleton.
It should be noted that the step that the function of each module and effect are corresponded respectively in above method embodiment in the device
Rapid 101~step 104, and because above-mentioned steps have a detailed description foregoing, therefore description is not repeated for succinct.
One of ordinary skill in the art will appreciate that realize all or part of flow in above-described embodiment method, being can be with
The hardware of correlation is instructed to complete by computer program, described program can be stored in a computer read/write memory medium
In, the program is upon execution, it may include such as the flow of the embodiment of above-mentioned each method.Wherein, described storage medium can be magnetic
Dish, CD, read-only memory (Read-Only Memory, ROM) or random access memory (Random Access
Memory, RAM) etc..
Presently preferred embodiments of the present invention is the foregoing is only, is not intended to limit the invention, all essences in the present invention
Any modification, equivalent and improvement made within refreshing and principle etc., should be included within the scope of the present invention.
Claims (10)
1. a kind of 3D shape expression, it is characterised in that including step：
Extract the mixed type skeleton of 3D shape；
By the segmentation to the mixed type skeleton, the segmentation of the 3D shape is obtained；
According to the 3D shape of segmentation, the minor structure of the 3D shape is obtained；And
According to the minor structure of the 3D shape, the expression of the 3D shape is set up using bag of words.
2. 3D shape expression as claimed in claim 1, it is characterised in that the mixed type bone of the extraction 3D shape
Frame, including：
The surface of the 3D shape is sampled, sampled point is obtained；And
The sampled point is expressed again, to obtain the mixed type skeleton comprising one-dimensional curve and two-dimensional slice.
3. 3D shape expression as claimed in claim 2, it is characterised in that described by the mixed type skeleton
Segmentation, obtains the segmentation of the 3D shape, including：
Split the mixed type skeleton；
According to the corresponding relation between the mixed type skeleton and the sampled point, institute is obtained by the segmentation of the mixed type skeleton
State the segmentation of 3D shape.
4. 3D shape expression as claimed in claim 1, it is characterised in that the 3D shape according to segmentation, is obtained
To the minor structure of the 3D shape, including：
By the 3D shape of the segmentation, multiple parts of the 3D shape are obtained；
Create the connection figure of the multiple part of connection；
The subgraph in the connection figure is extracted, the minor structure of the 3D shape is used as.
5. 3D shape expression as claimed in claim 1, it is characterised in that the son according to the 3D shape is tied
Structure, the expression of the 3D shape is set up using bag of words, including：
Each candidate's minor structure that the minor structure of the 3D shape and candidate's minor structure are concentrated is matched, described to determine
The frequency that each candidate's minor structure occurs in the 3D shape；
The frequency occurred according to each described candidate's minor structure in the 3D shape, create the word of the 3D shape to
Amount；
The term vector is normalized, expressed with the bag of words for obtaining the 3D shape.
6. 3D shape expression as claimed in claim 5, it is characterised in that described to set up described using bag of words
Before the expression of 3D shape, also include：
Create candidate's minor structure collection；
Wherein, establishment candidate's minor structure collection includes：
Obtain the minor structure for all 3D shapes that input data is concentrated；
Determine the similitude between the minor structure of the acquisition；
According to the similitude between the minor structure of the acquisition, candidate's minor structure is selected from the minor structure of the acquisition,
So as to constitute candidate's minor structure collection.
7. the expression of 3D shape as claimed in claim 6, it is characterised in that the minor structure of the determination acquisition
Between similitude, including：
Define the kernel of graph between the minor structure of the acquisition；
According to the kernel of graph, to determine the similitude between the minor structure obtained.
8. the expression of 3D shape as claimed in claim 7, it is characterised in that the minor structure of the definition acquisition
Between the kernel of graph, including：
Definition node core and side core；
Wherein, the node core is：
The side core is：
The knode(ni,nj) represent node core, kedge(ei,ej) represent side core, ni,njRepresent node, hiAnd hjIt is by closing respectively
In node niAnd njThe geometric properties histogram of part be formed by connecting, D is hiAnd hjNormalization it is related,It is any two
To hiAnd hjApart from D (hi,hj) maximum, uiAnd ujBe connected in two parts all-pair on angle vertically upward and
The two-dimensional histogram that distance is constituted.
9. a kind of 3D shape expression device, it is characterised in that including：
Skeleton extraction module, the mixed type skeleton for extracting 3D shape；
Split module, for by the segmentation to the mixed type skeleton, obtaining the segmentation of the 3D shape；
Minor structure extraction module, for the 3D shape according to segmentation, obtains the minor structure of the 3D shape；And
Module is expressed, for the minor structure according to the 3D shape, the table of the 3D shape is set up using bag of words
Reach.
10. 3D shape expression device as claimed in claim 9, it is characterised in that
The skeleton extraction module, samples for the surface to the 3D shape, obtains sampled point, and to the sampling
Point is expressed again, to obtain the mixed type skeleton comprising one-dimensional curve and two-dimensional slice.
The segmentation module, for splitting the mixed type skeleton, and by between the mixed type skeleton and the sampled point
Corresponding relation, obtains the segmentation of the 3D shape.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710171841.2A CN106960032B (en) | 2017-03-21 | 2017-03-21 | Three-dimensional shape expression method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710171841.2A CN106960032B (en) | 2017-03-21 | 2017-03-21 | Three-dimensional shape expression method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106960032A true CN106960032A (en) | 2017-07-18 |
| CN106960032B CN106960032B (en) | 2021-02-19 |
Family
ID=59471025
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710171841.2A Active CN106960032B (en) | 2017-03-21 | 2017-03-21 | Three-dimensional shape expression method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106960032B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101308500A (en) * | 2008-05-23 | 2008-11-19 | 浙江大学 | Visual and efficient three-dimensional human body movement data retrieval method based on demonstrated performance |
| CN101488142A (en) * | 2008-12-09 | 2009-07-22 | 南京大学 | Three-dimensional solid model retrieval method based on face topological interconnection constraint |
| CN101763652A (en) * | 2009-06-03 | 2010-06-30 | 中国科学院自动化研究所 | Three-dimensional framework fast extraction method based on branch feathers |
| CN106021330A (en) * | 2016-05-06 | 2016-10-12 | 浙江工业大学 | A three-dimensional model retrieval method used for mixed contour line views |
-
2017
- 2017-03-21 CN CN201710171841.2A patent/CN106960032B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101308500A (en) * | 2008-05-23 | 2008-11-19 | 浙江大学 | Visual and efficient three-dimensional human body movement data retrieval method based on demonstrated performance |
| CN101488142A (en) * | 2008-12-09 | 2009-07-22 | 南京大学 | Three-dimensional solid model retrieval method based on face topological interconnection constraint |
| CN101763652A (en) * | 2009-06-03 | 2010-06-30 | 中国科学院自动化研究所 | Three-dimensional framework fast extraction method based on branch feathers |
| CN106021330A (en) * | 2016-05-06 | 2016-10-12 | 浙江工业大学 | A three-dimensional model retrieval method used for mixed contour line views |
Non-Patent Citations (1)
| Title |
|---|
| 王飞等: "拓扑和形状特征相结合的三维模型检索", 《计算机辅助设计与图形学学报》 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106960032B (en) | 2021-02-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Bali et al. | A review on the strategies and techniques of image segmentation | |
| Ko et al. | Tree genera classification with geometric features from high-density airborne LiDAR | |
| Varnousfaderani et al. | Weighted color and texture sample selection for image matting | |
| JP2012088796A (en) | Image area division device, image area division method, and image area division program | |
| CN110019914A (en) | A kind of three-dimensional modeling data storehouse search method for supporting three-dimensional scenic interaction | |
| CN109086777A (en) | A kind of notable figure fining method based on global pixel characteristic | |
| Valero et al. | New hyperspectral data representation using binary partition tree | |
| CN111046920A (en) | Method for training food image classification model and image classification method | |
| JP5464739B2 (en) | Image area dividing apparatus, image area dividing method, and image area dividing program | |
| Zhao et al. | Key‐Frame Extraction Based on HSV Histogram and Adaptive Clustering | |
| CN111833432B (en) | Three-dimensional reconstruction method based on core two-dimensional gray scale image | |
| CN101853398B (en) | Chinese paper cutting identification method based on space constraint characteristic selection and combination thereof | |
| CN104268535A (en) | Method for extracting features of two-dimensional image | |
| Zheng et al. | Superpixel-based image recognition for food images | |
| Abinaya et al. | Jasmine flower segmentation: A superpixel based approach | |
| Tu et al. | Texture pattern separation for hyperspectral image classification | |
| CN106960032A (en) | 3D shape expression and device | |
| CN110162654A (en) | It is a kind of that image retrieval algorithm is surveyed based on fusion feature and showing for search result optimization | |
| Fan et al. | A 3D point cloud segmentation method based on local convexity and dimension features | |
| WO2018170731A1 (en) | Method and device for expressing three-dimensional shape | |
| JP5391876B2 (en) | Representative feature extraction system, method and program | |
| Toony et al. | 3D-NCuts: Adapting normalized cuts to 3D triangulated surface segmentation | |
| CN106570123B (en) | Remote sensing image retrieval method and system based on adjacent object association rule | |
| Grana et al. | Similarity-based retrieval with MPEG-7 3D descriptors: performance evaluation on the Princeton Shape Benchmark | |
| Gong et al. | A Video Classification Method Based on Spatiotemporal Detail Attention and Feature Fusion |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| EE01 | Entry into force of recordation of patent licensing contract |
Application publication date: 20170718 Assignee: Shenzhen achievement transformation Center Co.,Ltd. Assignor: SHENZHEN INSTITUTES OF ADVANCED TECHNOLOGY CHINESE ACADEMY OF SCIENCES Contract record no.: X2023980044064 Denomination of invention: Method and device for expressing three-dimensional shapes Granted publication date: 20210219 License type: Common License Record date: 20231023 |
|
| EE01 | Entry into force of recordation of patent licensing contract |