CN108734764A - A kind of method and system of clothes alignment - Google Patents
A kind of method and system of clothes alignment Download PDFInfo
- Publication number
- CN108734764A CN108734764A CN201810449157.0A CN201810449157A CN108734764A CN 108734764 A CN108734764 A CN 108734764A CN 201810449157 A CN201810449157 A CN 201810449157A CN 108734764 A CN108734764 A CN 108734764A
- Authority
- CN
- China
- Prior art keywords
- image
- model
- clothes
- alignment
- clothing
- 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
- 238000000034 method Methods 0.000 title claims abstract description 26
- 238000001514 detection method Methods 0.000 claims abstract description 19
- 238000010801 machine learning Methods 0.000 claims abstract description 5
- 230000009466 transformation Effects 0.000 claims description 6
- 239000011159 matrix material Substances 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 3
- 238000000844 transformation Methods 0.000 claims description 3
- 210000003739 neck Anatomy 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T15/00—3D [Three Dimensional] image rendering
- G06T15/04—Texture mapping
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T15/00—3D [Three Dimensional] image rendering
- G06T15/10—Geometric effects
- G06T15/20—Perspective computation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Graphics (AREA)
- General Physics & Mathematics (AREA)
- Computing Systems (AREA)
- Geometry (AREA)
- Image Analysis (AREA)
Abstract
A kind of method of clothes alignment, which is characterized in that including geometry patch will be added on model；Detect the structural key point of model's image；Image of clothing is aligned with nude model's image；The final garment consumption picture obtained after deformation.The present invention provides a kind of method and system of clothes alignment, using in such a way that model adds geometry and detects its locations of structures automatically, it is accurately positioned key point coordinate position, reach error in contiguous range for the image of resolution ratio, solve the problems, such as accurately to obtain mannequin's image, at the same avoid it is inconvenient for operation using scaling board scheme, using machine learning detection human joint points scheme be easily blocked influence accuracy of detection the problems such as.
Description
Technical field
The present invention relates to a kind of methods that computer graphics and technical field of computer vision more particularly to clothes are aligned
And system.
Background technology
The matching of clothes and human body image migrates, and is exactly before unobvious change the information such as clothes illumination, profile, texture
It puts, the garment data under different shooting environmentals is through on same human body image, matched with image of clothing with the proviso that obtaining
Model's image.Traditional method has the side that human joint points are detected using the method for scaling board and machine learning or deep learning
Method, but it is cumbersome in actual use using the method for scaling board, extra work amount is brought to photographer, and utilize
The method of machine learning or deep learning detection human joint points is influenced very big by types of garments, is easy to cause testing result not
Accurately.The present invention proposes a kind of alignment techniques of entity model shooting, using the addition geometry on entity model and certainly
The dynamic mode for detecting its locations of structures, solves the problems, such as accurately to obtain model's image under clothes, avoids scaling board operation not
Convenient, human joint points are easily blocked the problems such as influencing accuracy of detection.
Invention content
A kind of method of clothes alignment, which is characterized in that including,
Geometry patch will be added on model；
Detect the structural key point of model's image；
Image of clothing is aligned with nude model's image；
The final garment consumption picture obtained after deformation.
Further, it is described will on model add geometry patch, including, model front, the back side neck top
The positions such as portion, limb end increase geometry patch.
Further, the shape of the geometry patch is the geometries such as triangle, rectangle, circle, huge.
Further, the structural key point in detection model's image, by marking sample, machine learning is examined
Survey model.
Further, the detection model, including, the mould of key point in automatic and accurate detection body model's image
The structure of type, model includes training and two stages of prediction：In the training stage, to mode input label data, finally obtain to
A few trained model；In forecast period pass is directly acquired to model's image of trained mode input geometry mechanism information
The coordinate position of key point.
Further, the image of clothing and nude model's image automatic aligning, including,
Image of clothing is tentatively aligned with model's image；
It exposes the false reparation.
Further, the image of clothing is tentatively aligned with model's image, including, image of clothing and model's image are preliminary
Alignment, obtain clothes photographed data key point coordinate position, by ransac stochastical samplings consistency algorithm with
The mode that similarity similarity transformations combine, seeks the transformation matrix from clothes to model's image.Further, described to wear
Side is repaired, including,
The profile point set for obtaining clothes and model respectively, calculates the normal direction each put；
Using distance between points the matching point set between two profiles is calculated as constraint with normal angle；
Mesh generation is carried out to image of clothing, by adding characteristic point, triangle interior relative stiffness and straight-line segment
Equal constraint solvings target gridding apex coordinate position, obtains the image of clothing after deformation.
The present invention provides a kind of system of clothes alignment, including,
Geometry module will add geometry patch on model；
Detection module detects the structural key point in model's image；
Image of clothing is aligned by alignment module with nude model's image；
Deformation module, the final garment consumption picture obtained after deformation.
Advantageous effect
The present invention provides a kind of method and system of clothes alignment, using adding geometry in model and detect it automatically
The mode of locations of structures is accurately positioned key point coordinate position, reaches error in contiguous range for the image of resolution ratio, solution
Determined accurate acquisition mannequin's image the problem of, while avoid it is inconvenient for operation using scaling board scheme, using engineering
Practise the problems such as detection human joint points scheme is easily blocked influence accuracy of detection.
Description of the drawings
Fig. 1 is geometry schematic diagram
Fig. 2 is critical point detection flow chart in body model's image
Fig. 3 is clothes schematic diagram of exposing the false
Fig. 4 is garment consumption after repairing
Specific implementation mode
The implementation of technical solution is described in further detail below in conjunction with the accompanying drawings.The present embodiment provides a kind of clothes pair
Neat method, including,
Geometry patch will be added on model；
Detect the structural key point of model's image；
Image of clothing is aligned with nude model's image；
The final image of clothing obtained after deformation.
Preferred embodiment, described in the present embodiment will on model add geometry patch, including, model front,
The positions such as the neck top at the back side, limb end increase geometry patch.
The shape of preferred embodiment, the geometry patch described in the present embodiment is triangle, rectangle, circle.It is preferred that
Embodiment, the structural key point in detection model's image described in the present embodiment, including,
Training sample production method；
Build critical point detection model in body model's image.
Preferred embodiment, the training sample production method described in the present embodiment, including,
The nude model's image for shooting multiple backgrounds obtains model's mask and its joint by the manual, mode of automatically extracting
Point coordinates；
The body joint point coordinate position with the matched body model of existing garment data is chosen, two model's profiles of calculating
Match point set；
Mesh generation is carried out to image of clothing, by adding characteristic point, triangle interior relative stiffness and straight-line segment
Equal constraint solvings target gridding apex coordinate position, obtains the garment data after deformation, by the data and nude model's image into
Row map overlay, obtains label data.
Preferred embodiment, critical point detection model in structure body model's image described in the present embodiment, including, one
Automatic and accurate detects the model of key point in body model's image, and the structure of model includes training and two stages of prediction：It is instructing
The white silk stage finally obtains the model of at least one training to mode input label data；In forecast period, to trained model
The model's image for inputting geometry mechanism information, directly acquires the coordinate position of key point.
Preferred embodiment, the image of clothing described in the present embodiment and nude model's image automatic aligning, including,
Image of clothing is tentatively aligned with model's image；
It exposes the false reparation.
Preferred embodiment, the image of clothing described in the present embodiment are tentatively aligned with model's image, including, image of clothing with
Model's image is tentatively aligned, and is being obtained clothes photographed data key point coordinate position, is being calculated by ransac stochastical sampling consistency
The mode that method is combined with similarity similarity transformations seeks the transformation matrix from clothes to model's image.
Preferred embodiment, the reparation of exposing the false described in the present embodiment, including,
The profile point set for obtaining clothes and model respectively, calculates the normal direction each put；
Using distance between points the matching point set between two profiles is calculated as constraint with normal angle；
Mesh generation is carried out to image of clothing, by adding characteristic point, triangle interior relative stiffness and straight-line segment
Equal constraint solvings target gridding apex coordinate position, obtains the image of clothing after deformation.
The present embodiment provides a kind of systems of clothes alignment, including,
Geometry module will add geometry patch on model；
Detection module detects the structural key point in model's image；
Image of clothing is aligned by alignment module with nude model's image；
Deformation module, the final image of clothing obtained after deformation.
Foregoing description has a wide range of applications.For example, while it is disclosed that example may focus on clothes
In the method and system for filling alignment, it will be appreciated that concept disclosed in this invention can be applied equally to other wearable necks
Domain.Similarly, although being discussed to each embodiment in combination with the method and system of clothes alignment, clothes alignment
Any independent characteristic of method and system can be used alone or be integrated together use.Therefore, only to the discussion of any embodiment
It is intended as example, is not intended to propose by the scope limitation of the disclosure (including claims) in these examples.
Claims (9)
1. a kind of method of clothes alignment, which is characterized in that including,
Geometry patch will be added on model；
Detect the structural key point of model's image；
Image of clothing is aligned with nude model's image；
The final garment consumption picture obtained after deformation.
2. a kind of method of clothes alignment as described in claim 1, which is characterized in that described to add geometry knot on model
Structure patch, including, model front, the back side neck at the top of, the positions such as limb end increase geometry patch.
3. a kind of method of clothes alignment as claimed in claim 2, which is characterized in that the shape of the geometry patch
It is the geometries such as triangle, rectangle, circle, huge.
4. a kind of method of clothes alignment as described in claim 1, which is characterized in that the knot in detection model's image
Structure key point, by marking sample, machine learning obtains detection model.
5. a kind of method of clothes alignment as claimed in claim 4, which is characterized in that the detection model, including, one
Automatic and accurate detects the model of key point in body model's image, and the structure of model includes training and two stages of prediction：It is instructing
The white silk stage finally obtains the model of at least one training to mode input label data；In forecast period, to trained model
The model's image for inputting geometry mechanism information, directly acquires the coordinate position of key point.
6. a kind of method of clothes alignment as described in claim 1, which is characterized in that the image of clothing and nude model
Image automatic aligning, including,
Image of clothing is tentatively aligned with model's image；
It exposes the false reparation.
7. a kind of method of clothes alignment as claimed in claim 6, which is characterized in that the image of clothing and model's image
Preliminary alignment, including, image of clothing is tentatively aligned with model's image, is being obtained clothes photographed data key point coordinate position, is being led to
The mode that ransac stochastical samplings consistency algorithm is combined with similarity similarity transformations is crossed, is sought from clothes to model's image
Transformation matrix.
8. a kind of method of clothes alignment as claimed in claim 6, which is characterized in that the reparation of exposing the false, including,
The profile point set for obtaining clothes and model respectively, calculates the normal direction each put；
Using distance between points the matching point set between two profiles is calculated as constraint with normal angle；
Mesh generation is carried out to image of clothing, by adding characteristic point, triangle interior relative stiffness and straight-line segment etc. about
Beam solves target gridding apex coordinate position, obtains the image of clothing after deformation.
9. a kind of system of clothes alignment, which is characterized in that including,
Geometry module will add geometry patch on model；
Detection module detects the structural key point in model's image；
Image of clothing is aligned by alignment module with nude model's image；
Deformation module, the final garment consumption picture obtained after deformation.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810449157.0A CN108734764A (en) | 2018-05-11 | 2018-05-11 | A kind of method and system of clothes alignment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810449157.0A CN108734764A (en) | 2018-05-11 | 2018-05-11 | A kind of method and system of clothes alignment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108734764A true CN108734764A (en) | 2018-11-02 |
Family
ID=63937249
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810449157.0A Pending CN108734764A (en) | 2018-05-11 | 2018-05-11 | A kind of method and system of clothes alignment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108734764A (en) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20020178061A1 (en) * | 2002-07-12 | 2002-11-28 | Peter Ar-Fu Lam | Body profile coding method and apparatus useful for assisting users to select wearing apparel |
| CN102982578A (en) * | 2012-10-31 | 2013-03-20 | 北京航空航天大学 | Estimation method for dressed body 3D model in single character image |
| US20140176565A1 (en) * | 2011-02-17 | 2014-06-26 | Metail Limited | Computer implemented methods and systems for generating virtual body models for garment fit visualisation |
| US9245201B1 (en) * | 2013-03-15 | 2016-01-26 | Excelis Inc. | Method and system for automatic registration of images |
| CN106096560A (en) * | 2016-06-15 | 2016-11-09 | 广州尚云在线科技有限公司 | A kind of face alignment method |
| CN107507217A (en) * | 2017-08-17 | 2017-12-22 | 北京觅己科技有限公司 | Preparation method, device and the storage medium of certificate photo |
| WO2018011649A1 (en) * | 2016-07-14 | 2018-01-18 | Moda-Match Ltd. | Fitting clothing articles to human images |
| CN107767335A (en) * | 2017-11-14 | 2018-03-06 | 上海易络客网络技术有限公司 | A kind of image interfusion method and system based on face recognition features' point location |
-
2018
- 2018-05-11 CN CN201810449157.0A patent/CN108734764A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20020178061A1 (en) * | 2002-07-12 | 2002-11-28 | Peter Ar-Fu Lam | Body profile coding method and apparatus useful for assisting users to select wearing apparel |
| US20140176565A1 (en) * | 2011-02-17 | 2014-06-26 | Metail Limited | Computer implemented methods and systems for generating virtual body models for garment fit visualisation |
| CN102982578A (en) * | 2012-10-31 | 2013-03-20 | 北京航空航天大学 | Estimation method for dressed body 3D model in single character image |
| US9245201B1 (en) * | 2013-03-15 | 2016-01-26 | Excelis Inc. | Method and system for automatic registration of images |
| CN106096560A (en) * | 2016-06-15 | 2016-11-09 | 广州尚云在线科技有限公司 | A kind of face alignment method |
| WO2018011649A1 (en) * | 2016-07-14 | 2018-01-18 | Moda-Match Ltd. | Fitting clothing articles to human images |
| CN107507217A (en) * | 2017-08-17 | 2017-12-22 | 北京觅己科技有限公司 | Preparation method, device and the storage medium of certificate photo |
| CN107767335A (en) * | 2017-11-14 | 2018-03-06 | 上海易络客网络技术有限公司 | A kind of image interfusion method and system based on face recognition features' point location |
Non-Patent Citations (1)
| Title |
|---|
| 杨阿华等: "由航空影像自动重建大范围3维地形", 《中国图象图形学报》 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107490398B (en) | A kind of gauge pointer automatic identifying method | |
| Melo et al. | A new solution for camera calibration and real-time image distortion correction in medical endoscopy–initial technical evaluation | |
| Lin et al. | Automated body feature extraction from 2D images | |
| CN100585329C (en) | Location system of video finger and location method based on finger tip marking | |
| CN108272154B (en) | A kind of garment dimension measurement method and device | |
| GB2584033A (en) | Object tracking in zoomed video | |
| CN109766758A (en) | A kind of vision SLAM method based on ORB feature | |
| JP2011185872A5 (en) | ||
| JP2002092647A (en) | Information presentation system and model error detection system | |
| CN110074788B (en) | Body data acquisition method and device based on machine learning | |
| CN104766309A (en) | Plane feature point navigation and positioning method and device | |
| CN107862690A (en) | The circuit board element localization method and positioner of a kind of feature based Point matching | |
| CN108628306A (en) | Robot ambulation disorder detection method, device, computer equipment and storage medium | |
| CN107357426A (en) | A kind of motion sensing control method for virtual reality device | |
| CN116051631A (en) | Light spot labeling method and system | |
| Yang et al. | Auto-recognition Method for Pointer-type Meter Based on Binocular Vision. | |
| JP2019045989A (en) | Information processing apparatus, information processing method and computer program | |
| CN108734764A (en) | A kind of method and system of clothes alignment | |
| TWI270812B (en) | Method of calculating sub-pixel movement and position tracking sensor using the same | |
| CN106645168B (en) | A kind of detection method of crane boom cylinder concave-convex surface defect | |
| Lin et al. | Semi-automatic extraction of ribbon roads from high resolution remotely sensed imagery by T-shaped template matching | |
| CN113435412B (en) | Cement distribution area detection method based on semantic segmentation | |
| CN110501339A (en) | A kind of cloth cover localization method under complex environment | |
| JP2021009557A (en) | Information processing device, information processing method, and program | |
| TWI764393B (en) | Manufacturing method of pressure garment |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181102 |
|
| RJ01 | Rejection of invention patent application after publication |