CN107154071A - The method that Case-based Reasoning generates individual face body Model according to anthropological measuring size data - Google Patents
The method that Case-based Reasoning generates individual face body Model according to anthropological measuring size data Download PDFInfo
- Publication number
- CN107154071A CN107154071A CN201610119075.0A CN201610119075A CN107154071A CN 107154071 A CN107154071 A CN 107154071A CN 201610119075 A CN201610119075 A CN 201610119075A CN 107154071 A CN107154071 A CN 107154071A
- Authority
- CN
- China
- Prior art keywords
- mrow
- size
- msub
- length
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T19/00—Manipulating 3D models or images for computer graphics
- G06T19/20—Editing of 3D images, e.g. changing shapes or colours, aligning objects or positioning parts
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2200/00—Indexing scheme for image data processing or generation, in general
- G06T2200/08—Indexing scheme for image data processing or generation, in general involving all processing steps from image acquisition to 3D model generation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2210/00—Indexing scheme for image generation or computer graphics
- G06T2210/44—Morphing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2219/00—Indexing scheme for manipulating 3D models or images for computer graphics
- G06T2219/20—Indexing scheme for editing of 3D models
- G06T2219/2008—Assembling, disassembling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2219/00—Indexing scheme for manipulating 3D models or images for computer graphics
- G06T2219/20—Indexing scheme for editing of 3D models
- G06T2219/2021—Shape modification
Abstract
The invention provides a kind of Human Modeling technology for meeting personalized human outside of Case-based Reasoning, applied to three-dimensional fitting system and CAD garment industry.It is main to include three steps:1st, the relation between the size set of 26 anthropological measuring size compositions, statistical analysis size is extracted from instance model, instance model is divided into 16 rigid elements;2nd, the statistical information of block analysis body shape:With the mapping relations of linear regression method statistical analysis body shape to sized data.The new method of customizing model can be accurately generated using a kind of, i.e., learns linear mapping relation between its local deformation parameter and semantic parameter to each human body;3rd, human body is modeled according to input size, and after new manikin is obtained, further accurate adjustment is carried out to the initial model, be allowed to more laminating input dimensional parameters.This method accurately can generate to few distortion the individual face body Model for virtual fitting.
Description
First, technical field
The present invention relates to a kind of personalized method for setting up virtual human model in three-dimensional fitting software, applied to three-dimensional
Dressing system and CAD garment industry.
2nd, background technology
With continuing to develop for society and economy in recent years, people start to pursue personalized clothes, and customized clothing is set
Meter also gradually becomes more popular.User wishes to go out to meet the creation of oneself stature according to the Dam Configuration Design of oneself,
And the trouble for avoiding scene from fitting, therefore the user that designers need to use computer to be different building shape sets up corresponding size
Manikin simulate the effect of real clothes dress.With the gradually rise in the field, meet user's stature feature
Human Modeling also plays more and more important effect wherein, and it is the basis of three-dimensional virtual fitting.In recent years, learn both at home and abroad
Person proposes the method for a variety of different individual face body Model generations and develops different modeling softwares, however, how fast
Speed and be conveniently generated high-quality manikin, be still urgent problem to be solved.Existing method includes:Using three-dimensional
Body-scanner directly gathers human body surface information modeling and sets up manikin by human dimension parameter.It is straight using scanner
The human body degree of accuracy for meeting scanning customer is high, but needs the equipment of specialty, and cost is higher;Dimensional parameters using user are as defeated
The manikin and actual human body difference for entering to generate the generation of personalized model method are larger, and the final of influence three-dimensional fitting uses effect
Really.
3rd, the content of the invention
【Goal of the invention】
In order to make up the deficiency of existing modeling method, it can accurately reflect that body configuration is special comprehensively the invention provides one kind
Levy, set up personalized manikin method.Case-based Reasoning of the present invention generates manikin according to anthropological measuring size, to improve
The degree of accuracy of manikin generation;Blocking processing is carried out to manikin, each human body can be entered by the form of combination
Row splicing, considerably increases the diversity of training dataset manikin;The method of partition that the present invention is used also can be finer
Change ground generation manikin;Meanwhile, as one kind towards industrialized manikin modeling method, The present invention reduces input number
According to requirement, by the correlation analysis of anthropological measuring size data, obtain the association between each dimensional values, user only needs
Input 7 anthropometric datas, you can generation meets real individual face body Model.
【Technical scheme】
For achieving the above object, technical scheme used in the present invention is as follows:
The first step, anthropological measuring size is obtained and human body piecemeal:According to《GB10000-88 Chinese adult human dimensions》
The definition of measurement standard, obtains the key point of anthropological measuring size from each instance model, and then from instance model, that is, scans
Extracted in the three-dimensional grid model model of acquisition by the size set of 26 anthropological measuring size compositions.Meanwhile, according to these passes
Instance model is divided into 16 rigid regions by key point with critical size;Numerical lineardependence analysis and input size class by size
Type is chosen, and foundation can express 7 critical sizes to the BP neural network of remaining 19 minor dimension numerical relation with linearly reflecting
Penetrate model;
Second step, the statistical information of block analysis body shape:After each manikin is divided into 16 rigid regions,
It is determined that the control dimensional parameters type in each region, and obtain the deformation parameter of all rigid regions;Modeled segments are using a kind of
The new method of customizing model can accurately be generated:Learn its local deformation parameter and semantic parameter to each human body
Between linear mapping relation, i.e., pass through each rigid the position form parameter and size number of dimensionality reduction with linear regression method statistical analysis
According to the mapping relations of parameter;
3rd step, personalized model generation:7 crucial dimensional values of user are inputted, by training obtained BP nerves
Network generation all 26 anthropometric data set;And the Linear Mapping parameter calculating user obtained according to step 2 is corresponding
Human body Stiff Block deformation parameter, and each Stiff Block is spliced, obtain initial model;To the model of generation in key degree of enclosing chi
The position on very little portion faces summit is finely adjusted, and the final mask of generation is dimensionally more conformed to the data of input, is reduced
Error.
As a kind of preferred scheme, 26 sizes described in step 1 are respectively:Height, neck circumference, shoulder enclose, bust, waist
Enclose, hip encloses, arm length, shoulder breadth, shoulder height, chest breadth, breastheight, waist is wide, waist is high, hip is wide, hip is high, upperarm length, forearm length, elbow enclose, hand
Wrist circumference, leg length, thigh circumference, knee enclose, ankle encloses, thigh length, lower-leg length, head circumference;16 rigid regions of manikin
Respectively:Head, shoulder-chest, waist-belly, lower abdomen, left upper arm, right upper arm, left forearm, right forearm, left hand, the right hand, Zuo great
Leg, right thigh, left leg, right leg, left foot, right crus of diaphragm.
As a kind of preferred scheme, statistical analysis, choosing are carried out in step one to the linear relationship between human dimension
The size of strong linear correlation is taken to as inputoutput pair, input size and Output Size are obtained using the method for linear regression
Linear relationship, while input size is chosen, it should be noted that choose the size easily measured, and ensures to input between size
Linear relationship is relatively weak.After acquisition has the size of strong linear relationship with input size, 7 are set up by BP neural network
The mapping relations of key input size and remaining non-strong linear relationship size.Therefore 7 crucial sizes are respectively:Height, neck
Enclose, shoulder encloses, bust, waistline, hip enclose, arm is long;19 minor dimensions are respectively:Shoulder breadth, shoulder height, chest breadth, breastheight, waist are wide, waist
It is high, hip is wide, hip is high, upperarm length, forearm length, elbow enclose, wrist is enclosed, leg length, thigh circumference, knee enclose, ankle encloses, thigh length, small
Leg length, head circumference.
As a kind of preferred scheme, in step 2, it is determined that the control dimensional parameters at each position are as follows:Control the chi on head
Very little parameter includes head circumference, neck circumference;The dimensional parameters of control shoulder-chest enclose including neck circumference, shoulder, shoulder breadth, shoulder height, bust, chest breadth, chest
It is high;Controlling the dimensional parameters of waist-belly includes that bust, chest breadth, breastheight, waistline, waist be wide, waist is high;Underbelly size is controlled to join
Number includes that waistline, waist are wide, waist is high, hip encloses, hip is wide, hip is high;Control left upper arm and right upper arm dimensional parameters include arm length, on
Brachium, elbow enclose;The dimensional parameters of control left forearm and right forearm include arm length, forearm length, wrist and enclose length;Control left hand and the right side
The dimensional parameters of hand include arm length, wrist and enclose length;Control left thigh and right thigh dimensional parameters include leg length, thigh circumference,
Knee encloses, thigh length;The dimensional parameters of control left leg and right leg include leg length, knee encloses, ankle encloses, lower-leg length;
The dimensional parameters of control left foot and right crus of diaphragm include leg length, ankle and enclosed.
As a kind of preferred scheme, in step 2, using a template model as reference model, each rigidity of model is completed
The parametrization of block is represented:By the triangle gridding k on each Stiff Block of template to the corresponding triangle gridding deformation of object module, with one
Individual 3*3 affine transformation matrix MkWeigh.Next, by matrix MkIt is rewritten as a 9*1 vectorial Vk, all triangles in the position
Shape VkMerge into the deformation parameter of the Stiff Block.
As a kind of preferred scheme, the method for deformation parameter dimensionality reduction is PCA (Principal in step 2
Component Analysis, PCA).Linear Mapping between each block models spot size and the deformation parameter of its dimensionality reduction is closed
Tie up in learning process and to be obtained by least square method, its Linear Mapping matrix is expressed as M(i), wherein i represent the matrix be i-th
The control dimensional parameters and the Linear Mapping matrix of deformation parameter of individual Stiff Block.
As a kind of preferred scheme, comprise the following steps in step 3:
Input the crucial dimensional values set S of user 7input, the sized data based on input has very strong linear relationship
Size to that can be calculated by following linear formula:fnew=finputβ+ε, the formula is calculated using least square method, is being obtained
After these sizes, all sizes can be added S by usknownIn set, then equally adopt with the aforedescribed process, calculate with
SknownThere is the size of strong linear relationship in set, and constantly update SknownSet, until all and SknownThere is by force size in set
The size of linear relationship is finished by calculating;
According to 7 critical sizes, by training obtained BP neural network and linear mapping relation generation to remove SknownCollection
Anthropometric data outside conjunction, obtains 26 complete anthropometric data set;
According to the parameter sets of 26 anthropological measuring sizes, according to the local deformation of each human body obtained in step 2
Linear mapping relation obtains the deformation parameter of each human body Stiff Block between parameter and semantic parameter;
The average value that position triangle deformation parameter is abutted by calculating adjacent block completes the splicing of adjacent block, so that
The smooth integration of whole manikin is completed, initial manikin is obtained;
The method minimized using error function is carried out to generation model in crucial girth size portion faces vertex position
Fine setting, make all errors and is minimized.Error function includes length dimension error term Edist, girth size error term Egirth、
Deformation error term Edeform。
As a kind of preferred scheme, the linear relationship in step one between statistical analysis human dimension uses skin
The inferior coefficient correlation (Pearson Correlation Coefficient) of that, formula is:
Wherein,WithIt is the average value of kth number and jth size, n is the quantity of instance model.The model of coefficient correlation
Enclose between -1 to 1, when coefficient is close to 1 or -1, the linear relationship between size k and size j is stronger, conversely, working as coefficient
During close to 0, there is no linear relationship between size k and size j or there was only very weak linear relationship.It is strong linear defined in the present invention
The incidence coefficient threshold value of relation is 0.9 and -0.9, chooses coefficient correlation and is more than 0.9 or the size pair less than -0.9, and therefrom
The size relatively easily measured is chosen as input information.
As a kind of preferred scheme, in step 3, three error functions are respectively length dimension error term Edist, degree of enclosing chi
Very little error term Egirth, deformation error term EdeformIt is defined respectively as:Length dimension error:
Wherein fdThe set of human body length dimension, dist () be calculate two key points or key central point Europe it is several in
Obtain the function of distance;Girth size error:
Wherein fgIt is the set of each human body girth size, cir () is to calculate i-th of crucial cross section perpendicular to skeleton
The function of convex closure length;Define deformation error as follows, explain the change of body shape spatial shape state:
In this, Δ is minitUpper summit and mnewThe difference on upper summit, the purpose of this error term is to minimize deformation
And minimize the change of the relative position between key point.The error term can prevent the appearance of local deformity and distortion situation.
Final majorized function can be expressed as:E=α Edist+βEgirth+γEdeform.W in preceding two formulaiIt is i-th of size
Weight.The weight of 7 critical sizes of initial input is larger, because the accuracy of these sizes can be guaranteed.
【Beneficial effect】
Compared with prior art, the present invention has following conspicuousness progress and beneficial effect:
Example manikin of the invention based on one group of scanning, sets up three-dimensional (3 D) manikin each just by linear regression method
Property block portion position deformation parameter and corresponding anthropological measuring size linear mapping relation.The method for employing piecemeal modeling improves life
Into manikin precision, while employing actual scanning human body model data as sample data, improve generation human body
The validity of model, makes its profile be more nearly real body shape.Generate after initialization model, error is carried out most to model
Smallization accurate adjustment, can generate more multifarious manikin, further reduce the gap of input size and moulded dimension.This hair
Bright method can more accurately generate manikin according to simplified input information and be spent without significantly increasing the time.
4th, illustrate
The example manikin piecemeal of accompanying drawing 1 and training process
Accompanying drawing 2 generates individual face body Model process according to human dimension
5th, embodiment
The present invention is further described below in conjunction with the accompanying drawings.
A kind of method that Case-based Reasoning generates individual face body Model according to anthropometric data, comprises the following steps:
The first step, anthropological measuring size is obtained and human body piecemeal:According to《GB10000-88 Chinese adult human dimensions》
The definition of measurement standard, obtains the key point of anthropological measuring size from each instance model, and then from instance model, that is, sweeps
Extracted in the three-dimensional grid model model for retouching acquisition by the size set of 26 anthropological measuring size compositions.Meanwhile, according to these
Instance model is divided into 16 rigid regions by key point with critical size;Numerical lineardependence analysis and input size by size
Type choose, set up can express 7 critical sizes to remaining 19 minor dimension numerical relation BP neural network with linearly
Mapping model;
Second step, the statistical information of block analysis body shape:After each manikin is divided into 16 rigid regions,
It is determined that the control dimensional parameters type in each region, and obtain the deformation parameter (such as Fig. 1) of all rigid regions;Modeled segments are adopted
The new method of customizing model can be accurately generated with a kind of:Learn its local deformation parameter and language to each human body
Linear mapping relation between adopted parameter, i.e., with linear regression method statistical analysis by dimensionality reduction each rigid position form parameter with
The mapping relations of sized data parameter;
Fig. 1 describes the handling process of first step and second step, and wherein first step is used as the pretreated of model
Journey, second step is learning process, the preprocessing process in the present invention merging into the first step and second step before generation model.
The process only needs to carry out once in scan model example set, need not be repeatedly during personalized model is generated
Call.
3rd step, personalized model generation, its flow is as shown in Figure 2:7 crucial dimensional values of user are inputted, are passed through
Train linear relation model generation all 26 anthropometric data set between obtained BP neural network and size;And root
The Linear Mapping parameter obtained according to step 2 calculates the corresponding human body Stiff Block deformation parameter of user, and each Stiff Block is spelled
Connect, obtain initial model;The model of generation is finely adjusted in the position on crucial girth size portion faces summit, makes generation
Final mask dimensionally more conforms to the data of input, reduces error.The step meets personalized size human body as generation
The core procedure of model, needs once to be called when often generating a manikin.
Claims (9)
1. a kind of method that Case-based Reasoning generates individual face body Model according to anthropometric data, its general characteristic is, wrap
Include the following steps:
The first step, anthropological measuring size is obtained and human body piecemeal:According to《GB10000-88 Chinese adult human dimensions》Measurement
The definition of standard, obtains the key point of anthropological measuring size, and then obtain from instance model, i.e. scanning from each instance model
Three-dimensional grid model model in extract by the size set of 26 anthropological measuring size compositions.Meanwhile, according to these key points
Instance model is divided into 16 rigid regions with critical size;Selected by the numerical lineardependence analysis and input Dimension Types of size
Take, set up the BP neural network that can express 7 critical sizes to remaining 19 minor dimension numerical relation and Linear Mapping mould
Type;
Second step, the statistical information of block analysis body shape:After each manikin is divided into 16 rigid regions, it is determined that
The control dimensional parameters type in each region, and obtain the deformation parameter of all rigid regions;Modeled segments can using one kind
Accurately generate the new method of customizing model:Each human body is learnt between its local deformation parameter and semantic parameter
Linear mapping relation, i.e., joined with linear regression method statistical analysis by each rigid position form parameter of dimensionality reduction with sized data
Several mapping relations;
3rd step, personalized model generation:7 crucial dimensional values of user are inputted, by training obtained BP neural network
Generation all 26 anthropometric data set;And the Linear Mapping parameter obtained according to step 2 calculates the corresponding human body of user
Stiff Block deformation parameter, and each Stiff Block is spliced, obtain initial model;To the model of generation in crucial girth size portion
The position of position surface vertices is finely adjusted, and the final mask of generation is dimensionally more conformed to the data of input, is reduced error.
2. the method that Case-based Reasoning according to claim 1 generates individual face body Model according to anthropometric data, its
It is characterised by, 26 described in step 1 size is respectively:Height, neck circumference, shoulder enclose, bust, waistline, hip enclose, arm length, shoulder
Width, shoulder height, chest breadth, breastheight, waist are wide, waist is high, hip is wide, hip is high, upperarm length, forearm length, elbow enclose, wrist is enclosed, leg length, thigh circumference,
Knee encloses, ankle encloses, thigh length, lower-leg length, head circumference;16 rigid regions of manikin are respectively:Head, shoulder-chest
Portion, waist-belly, lower abdomen, left upper arm, right upper arm, left forearm, right forearm, left hand, the right hand, left thigh, right thigh, left leg,
Right leg, left foot, right crus of diaphragm.
3. the method that Case-based Reasoning according to claim 1 generates individual face body Model according to anthropometric data, its
It is characterised by, statistical analysis is carried out to the linear relationship between human dimension in step one, chooses the chi of strong linear correlation
It is very little to as inputoutput pair, input size and the linear relationship of Output Size are obtained using the method for linear regression, choosing
While inputting size, it should be noted that choose the size easily measured, and ensure the linear relationship inputted between size relatively
It is weak.Obtain with input size have the size of strong linear relationship after, by BP neural network set up 7 key input sizes with
The mapping relations of remaining non-strong linear relationship size.Therefore 7 crucial sizes are respectively:Height, neck circumference, shoulder enclose, bust, waist
Enclose, hip encloses, arm is long;19 minor dimensions are respectively:Shoulder breadth, shoulder height, chest breadth, breastheight, waist are wide, waist is high, hip is wide, hip is high, on
Brachium, forearm length, elbow are enclosed, wrist is enclosed, leg length, thigh circumference, knee enclose, ankle encloses, thigh length, lower-leg length, head circumference.
4. the method that Case-based Reasoning according to claim 1 generates individual face body Model according to anthropometric data, its
It is characterised by, in step 2, it is determined that the control dimensional parameters at each position are as follows:Control head dimensional parameters include head circumference,
Neck circumference;The dimensional parameters of control shoulder-chest enclose including neck circumference, shoulder, shoulder breadth, shoulder height, bust, chest breadth, breastheight;Control waist-belly
Dimensional parameters include that bust, chest breadth, breastheight, waistline, waist be wide, waist is high;Underbelly dimensional parameters are controlled to include waistline, waist
Wide, waist is high, hip encloses, hip is wide, hip is high;The dimensional parameters of control left upper arm and right upper arm including arm length, upperarm length, elbow with enclosing;
The dimensional parameters of control left forearm and right forearm include arm length, forearm length, wrist and enclose length;The size of left hand and the right hand is controlled to join
Number includes arm length, wrist and encloses length;The dimensional parameters of control left thigh and right thigh enclose including leg length, thigh circumference, knee, thigh
Length;The dimensional parameters of control left leg and right leg include leg length, knee encloses, ankle encloses, lower-leg length;Control left foot and the right side
The dimensional parameters of pin include leg length, ankle and enclosed.
5. the method that Case-based Reasoning according to claim 1 generates individual face body Model according to anthropometric data, its
It is characterised by, in step 2, using a template model as reference model, the parametrization for completing each Stiff Block of model is represented:
By the triangle gridding k on each Stiff Block of template to the corresponding triangle gridding deformation of object module, with 3*3 affine transformation
Matrix MkWeigh.Next, by matrix MkIt is rewritten as a 9*1 vectorial Vk, all triangle V in the positionkMerge into the rigidity
The deformation parameter of block.
6. the method that Case-based Reasoning according to claim 1 generates individual face body Model according to anthropometric data, its
It is characterised by, the method for deformation parameter dimensionality reduction is PCA (Principal Component in step 2
Analysis, PCA).Linear mapping relation between each block models spot size and the deformation parameter of its dimensionality reduction was learning
Obtained in journey by least square method, its Linear Mapping matrix is expressed as M(i), wherein i represents that the matrix is i-th Stiff Block
Control the Linear Mapping matrix of dimensional parameters and deformation parameter.
7. the method that Case-based Reasoning according to claim 1 generates individual face body Model according to anthropometric data, its
It is characterised by, comprises the following steps in step 3:
Input the crucial dimensional values set S of user 7input, the sized data based on input, the chi for having very strong linear relationship
It is very little to that can be calculated by following linear formula:fnew=finputβ+ε, the formula is calculated using least square method, is obtaining these
After size, all sizes can be added S by usknownIn set, then equally adopt with the aforedescribed process, calculate and SknownCollection
There is the size of strong linear relationship in conjunction, and constantly update SknownSet, until all and SknownSize has strong linear pass in set
The size of system is finished by calculating;
According to 7 critical sizes, by training obtained BP neural network and linear mapping relation generation to remove SknownGather it
Outer anthropometric data, obtains 26 complete anthropometric data set;
According to the parameter sets of 26 anthropological measuring sizes, according to the local deformation parameter of each human body obtained in step 2
Linear mapping relation obtains the deformation parameter of each human body Stiff Block between semantic parameter;
The average value that position triangle deformation parameter is abutted by calculating adjacent block completes the splicing of adjacent block, so as to complete
The smooth integration of whole manikin, obtains initial manikin;
The method minimized using error function is finely adjusted to generation model in crucial girth size portion faces vertex position,
Make all errors and minimize.Error function includes length dimension error term Edist, girth size error term Egirth, deformation miss
Poor item Edeform。
8. the method that Case-based Reasoning according to claim 3 generates individual face body Model according to anthropometric data, its
It is characterised by, the linear relationship between statistical analysis human dimension uses Pearson correlation coefficient (Pearson
Correlation Coefficient), formula is:
<mrow>
<msub>
<mi>Cor</mi>
<mrow>
<mi>k</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>=</mo>
<mfrac>
<mrow>
<munderover>
<mi>&Sigma;</mi>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>n</mi>
</munderover>
<mrow>
<mo>(</mo>
<mrow>
<msubsup>
<mi>f</mi>
<mi>k</mi>
<mi>i</mi>
</msubsup>
<mo>-</mo>
<msub>
<mover>
<mi>f</mi>
<mo>&OverBar;</mo>
</mover>
<mi>k</mi>
</msub>
</mrow>
<mo>)</mo>
</mrow>
<mrow>
<mo>(</mo>
<mrow>
<msubsup>
<mi>f</mi>
<mi>j</mi>
<mi>i</mi>
</msubsup>
<mo>-</mo>
<msub>
<mover>
<mi>f</mi>
<mo>&OverBar;</mo>
</mover>
<mi>j</mi>
</msub>
</mrow>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<msqrt>
<mrow>
<munderover>
<mi>&Sigma;</mi>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>n</mi>
</munderover>
<msup>
<mrow>
<mo>(</mo>
<mrow>
<msubsup>
<mi>f</mi>
<mi>k</mi>
<mi>i</mi>
</msubsup>
<mo>-</mo>
<msub>
<mover>
<mi>f</mi>
<mo>&OverBar;</mo>
</mover>
<mi>k</mi>
</msub>
</mrow>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
</mrow>
</msqrt>
<msqrt>
<mrow>
<munderover>
<mi>&Sigma;</mi>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>n</mi>
</munderover>
<msup>
<mrow>
<mo>(</mo>
<mrow>
<msubsup>
<mi>f</mi>
<mi>j</mi>
<mi>i</mi>
</msubsup>
<mo>-</mo>
<msub>
<mover>
<mi>f</mi>
<mo>&OverBar;</mo>
</mover>
<mi>j</mi>
</msub>
</mrow>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
</mrow>
</msqrt>
</mrow>
</mfrac>
</mrow>
Wherein,WithIt is the average value of kth number and jth size, n is the quantity of instance model.The scope of coefficient correlation for-
Between 1 to 1, when coefficient is close to 1 or -1, the linear relationship between size k and size j is stronger, conversely, working as coefficient close to 0
When, there is no linear relationship between size k and size j or there was only very weak linear relationship.Strong linear relationship defined in the present invention
Incidence coefficient threshold value be 0.9 and -0.9, choose coefficient correlation more than 0.9 or the size pair less than -0.9, and therefrom choose
The size relatively easily measured is used as input information.
9. the method that Case-based Reasoning according to claim 7 generates individual face body Model according to anthropometric data, its
It is characterised by, three error functions are respectively length dimension error term Edist, girth size error term Egirth, deformation error term
EdeformIt is defined respectively as:Length dimension error:Wherein fdIt is human body length dimension
Set, dist () is the function for calculating two key points or key central point Euclidean distance;Girth size error:
<mrow>
<msub>
<mi>E</mi>
<mrow>
<mi>g</mi>
<mi>i</mi>
<mi>r</mi>
<mi>t</mi>
<mi>h</mi>
</mrow>
</msub>
<mo>=</mo>
<munder>
<mi>&Sigma;</mi>
<mrow>
<mi>g</mi>
<mi>i</mi>
<mi>r</mi>
<mo>&Element;</mo>
<msub>
<mi>f</mi>
<mi>g</mi>
</msub>
</mrow>
</munder>
<msub>
<mi>w</mi>
<mi>i</mi>
</msub>
<msup>
<mrow>
<mo>(</mo>
<mrow>
<mi>c</mi>
<mi>i</mi>
<mi>r</mi>
<mrow>
<mo>(</mo>
<mrow>
<msub>
<mi>m</mi>
<mrow>
<mi>n</mi>
<mi>e</mi>
<mi>w</mi>
</mrow>
</msub>
<mo>,</mo>
<mi>i</mi>
</mrow>
<mo>)</mo>
</mrow>
<mo>-</mo>
<msub>
<mi>girth</mi>
<mi>i</mi>
</msub>
</mrow>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
</mrow>
Wherein fgIt is the set of each human body girth size, cir () is to calculate i-th of crucial cross section convex closure perpendicular to skeleton
The function of length;Define deformation error as follows, explain the change of body shape spatial shape state:
<mrow>
<msub>
<mi>E</mi>
<mrow>
<mi>d</mi>
<mi>e</mi>
<mi>f</mi>
<mi>o</mi>
<mi>r</mi>
<mi>m</mi>
</mrow>
</msub>
<mo>=</mo>
<munder>
<mo>&Sigma;</mo>
<mrow>
<mrow>
<mo>(</mo>
<mrow>
<msub>
<mi>p</mi>
<mi>i</mi>
</msub>
<mo>,</mo>
<msub>
<mi>p</mi>
<mi>j</mi>
</msub>
</mrow>
<mo>)</mo>
</mrow>
<mo>&Element;</mo>
<mi>e</mi>
<mi>d</mi>
<mi>g</mi>
<mi>e</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>m</mi>
<mrow>
<mi>in</mi>
<mi>i</mi>
<mi>t</mi>
</mrow>
</msub>
<mo>)</mo>
</mrow>
</mrow>
</munder>
<msup>
<mrow>
<mo>(</mo>
<msub>
<mi>&Delta;p</mi>
<mi>i</mi>
</msub>
<mo>-</mo>
<msub>
<mi>&Delta;p</mi>
<mi>j</mi>
</msub>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
</mrow>
In this, Δ is minitUpper summit and mnewThe difference on upper summit, the purpose of this error term is to minimize deformation and most
Relative position change between smallization key point.The error term can prevent the appearance of local deformity and distortion situation.Finally
Majorized function can be expressed as:E=α Edist+βEgirth+γEdeform.W in preceding two formulaiIt is the power of i-th of size
Weight.The weight of 7 critical sizes of initial input is larger, because the accuracy of these sizes can be guaranteed.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610119075.0A CN107154071A (en) | 2016-03-02 | 2016-03-02 | The method that Case-based Reasoning generates individual face body Model according to anthropological measuring size data |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610119075.0A CN107154071A (en) | 2016-03-02 | 2016-03-02 | The method that Case-based Reasoning generates individual face body Model according to anthropological measuring size data |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107154071A true CN107154071A (en) | 2017-09-12 |
Family
ID=59792496
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610119075.0A Pending CN107154071A (en) | 2016-03-02 | 2016-03-02 | The method that Case-based Reasoning generates individual face body Model according to anthropological measuring size data |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107154071A (en) |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108280476A (en) * | 2018-01-22 | 2018-07-13 | 东华大学 | Based on the human upper limb of principal component analysis and hierarchical cluster band typoiogical classification method |
CN108564651A (en) * | 2018-02-26 | 2018-09-21 | 盎锐(上海)信息科技有限公司 | Body scan data device and data creation method with data systematic function |
CN110135078A (en) * | 2019-05-17 | 2019-08-16 | 上海凌笛数码科技有限公司 | A kind of human parameters automatic generation method based on machine learning |
CN110264310A (en) * | 2019-05-30 | 2019-09-20 | 肖伯祥 | A kind of clothing pattern making method based on human body big data |
WO2019178886A1 (en) * | 2018-03-22 | 2019-09-26 | 香港纺织及成衣研发中心 | Intelligent bionic human body part model detection device and method for manufacturing same |
CN110838179A (en) * | 2019-09-27 | 2020-02-25 | 深圳市三维人工智能科技有限公司 | Body modeling method and device based on body measurement data and electronic equipment |
CN110909464A (en) * | 2019-11-19 | 2020-03-24 | 大连工业大学 | Method for manufacturing standard ready-made clothes mannequin |
CN111047407A (en) * | 2019-12-13 | 2020-04-21 | 南京中略信息技术有限公司 | Clothing personalized size customization method using variational multidimensional regression |
CN115631322A (en) * | 2022-10-26 | 2023-01-20 | 钰深(北京)科技有限公司 | User-oriented virtual three-dimensional fitting method and system |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103886117A (en) * | 2012-12-20 | 2014-06-25 | 上海工程技术大学 | Method for improving virtual human modeling accuracy in 3D clothing fitting software |
-
2016
- 2016-03-02 CN CN201610119075.0A patent/CN107154071A/en active Pending
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103886117A (en) * | 2012-12-20 | 2014-06-25 | 上海工程技术大学 | Method for improving virtual human modeling accuracy in 3D clothing fitting software |
Non-Patent Citations (2)
Title |
---|
BON-YEOL KOO, EUN-JOO PARK, DONG-KWON CHOI, ETC: "Example-based statistical framework for parametric modeling of human body shapes", 《COMPUTERS IN INDUSTRY》 * |
冯明辉: "基于单目视觉照相法的人体参数快速测量研究", 《中国优秀硕士学位论文全文数据库基础科学辑》 * |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108280476A (en) * | 2018-01-22 | 2018-07-13 | 东华大学 | Based on the human upper limb of principal component analysis and hierarchical cluster band typoiogical classification method |
CN108564651A (en) * | 2018-02-26 | 2018-09-21 | 盎锐(上海)信息科技有限公司 | Body scan data device and data creation method with data systematic function |
WO2019178886A1 (en) * | 2018-03-22 | 2019-09-26 | 香港纺织及成衣研发中心 | Intelligent bionic human body part model detection device and method for manufacturing same |
US11860051B2 (en) | 2018-03-22 | 2024-01-02 | The Hong Kong Research Institute Of Textiles And Apparel Limited | Intelligent bionic human body part model detection device and method for manufacturing same |
CN110135078B (en) * | 2019-05-17 | 2023-03-14 | 浙江凌迪数字科技有限公司 | Human body parameter automatic generation method based on machine learning |
CN110135078A (en) * | 2019-05-17 | 2019-08-16 | 上海凌笛数码科技有限公司 | A kind of human parameters automatic generation method based on machine learning |
CN110264310A (en) * | 2019-05-30 | 2019-09-20 | 肖伯祥 | A kind of clothing pattern making method based on human body big data |
CN110264310B (en) * | 2019-05-30 | 2021-09-03 | 肖伯祥 | Clothing pattern making method based on human body big data |
CN110838179A (en) * | 2019-09-27 | 2020-02-25 | 深圳市三维人工智能科技有限公司 | Body modeling method and device based on body measurement data and electronic equipment |
CN110838179B (en) * | 2019-09-27 | 2024-01-19 | 深圳市三维人工智能科技有限公司 | Human body modeling method and device based on body measurement data and electronic equipment |
CN110909464A (en) * | 2019-11-19 | 2020-03-24 | 大连工业大学 | Method for manufacturing standard ready-made clothes mannequin |
CN111047407A (en) * | 2019-12-13 | 2020-04-21 | 南京中略信息技术有限公司 | Clothing personalized size customization method using variational multidimensional regression |
CN115631322A (en) * | 2022-10-26 | 2023-01-20 | 钰深(北京)科技有限公司 | User-oriented virtual three-dimensional fitting method and system |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107154071A (en) | The method that Case-based Reasoning generates individual face body Model according to anthropological measuring size data | |
CN105006014B (en) | The realization method and system that virtual clothing Fast simulation is tried on | |
CN101751689B (en) | Three-dimensional facial reconstruction method | |
EP1160732B1 (en) | Virtual shape generation through free-form deformation | |
CN110211196A (en) | A kind of virtually trying method and device based on posture guidance | |
Wuhrer et al. | Estimating 3D human shapes from measurements | |
CN110264310B (en) | Clothing pattern making method based on human body big data | |
CN114202629A (en) | Human body model establishing method, system, equipment and storage medium | |
US20140333614A1 (en) | System and method for simulating realistic clothing | |
CN103810750B (en) | Human body section ring based parametric deformation method | |
KR101072944B1 (en) | System for creating 3d human body model and method therefor | |
CN109829971B (en) | Method and device for creating human body virtual model | |
CN101271589B (en) | Three-dimensional mannequin joint center extraction method | |
CN114119907A (en) | Fitting method and device of human body model and storage medium | |
CN114119905A (en) | Virtual fitting method, system, equipment and storage medium | |
CN116797699A (en) | Intelligent animation modeling method and system based on three-dimensional technology | |
CN206825428U (en) | A kind of real-time acquisition device of three-dimensional point cloud | |
Jianhua et al. | Human skin deformation from cross-sections | |
CN100595795C (en) | A human model design method based on hybrid interpolation parameterization | |
Meixner et al. | Development of a method for an automated generation of anatomy-based, kinematic human models as a tool for virtual clothing construction | |
CN106228417A (en) | A kind of bra fit detection method | |
Fengyi et al. | 3D Garment Design Model Based on Convolution Neural Network and Virtual Reality | |
CN106652035A (en) | Human body modeling method based on deformable spiral line model | |
CN114758039A (en) | Sectional driving method and equipment of human body model and storage medium | |
CN110399656A (en) | Parameters design is saved based on the lower dress waist of fuzzy logic and neural network |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20170912 |
|
WD01 | Invention patent application deemed withdrawn after publication |