CN101136105A - Freely differences calculus and deformable contour outline extracting system - Google Patents
Freely differences calculus and deformable contour outline extracting system Download PDFInfo
- Publication number
- CN101136105A CN101136105A CNA2007101043926A CN200710104392A CN101136105A CN 101136105 A CN101136105 A CN 101136105A CN A2007101043926 A CNA2007101043926 A CN A2007101043926A CN 200710104392 A CN200710104392 A CN 200710104392A CN 101136105 A CN101136105 A CN 101136105A
- Authority
- CN
- China
- Prior art keywords
- calculus
- differences
- freely
- outline extracting
- profile
- 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
Images
Landscapes
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention comprises: a image memory 1, a component for executing the high level vision 2, a basic polygon transform component 3, an orthogonal location transform component 4, a free difference operation component 5, a target outline extracting component 6, and an energy function minimizing component 7.
Description
Technical field
The present invention relates to Flame Image Process, particularly implement the method that deformable contour outline extracts in the method for enforcement rim detection and the image segmentation field in the Flame Image Process.
Background technology
In various image applications, if relate to image object extract, measurement etc. all be unable to do without image segmentation.Image segmentation is meant utilizes that Partial Feature extracts target area in the image or area-of-interest in the image information.Usually, target area or area-of-interest are called foreground area, remainder is called the background area.
Image segmentation is the committed step that is entered into graphical analysis by Flame Image Process.With regard to concrete image segmentation algorithm, existing thousands of kinds of relevant research document.Two kinds of following patterns have mainly been taked in these researchs: first kind, the processing procedure of " bottom-up " comprises that rim detection, edge thinning and edge such as are connected at several stages.This is a process that is pushed into high layer information by bottom knowledge.The research method of the order of this strictness is regarded each link of image segmentation as relatively independent stage, can adopt the module with relatively independent function to calculate, but simultaneously high level arrived in the error propagation of bottom, has no chance to revise.Simultaneously, owing to ignored the high layer information of target, for example the connectivity of target causes profile to extract difficulty.Second kind, the processing procedure of " top-down ", it is Snake model (being called movable contour model, active contour model, deformable model again), it is a kind of profile extraction algorithm based on deformable model, its remarkable advantage is that whole geometry information with objective contour is as guidance, in the evolutionary process of energy trace, with its shape of internal force constraint, its behavior of external force guiding is dragged it to significant characteristics of image with image force.The Snake model is converted into a kind of top-down optimal profile approximate procedure to the edge extracting process with the gradient information of the variation of image grayscale foundation as edge extracting.Like this, when seeking significant characteristics of image, higher layer mechanism can be carried out with model by characteristics of image being pushed to a suitable Local Extremum alternately.
Current, contour extraction method based on deformable model has obtained extensive studies and application in fields such as image segmentation, machine vision, according to statement and implementation different, mainly be divided into two classes: parametric type Snake model, geometric type Snake model.Parametric type Snake model is the parametric form of Lagrangian type, the curve or the curved surface of explicitly statement deformation, and result of calculation has compactness.Geometric type Snake model is the level set form of Euler's type, implicitly explains the curve or the curved surface of deformation.With regard to present present Research, the major advantage of parametric type Snake model is to have topological preferably retentivity.Weak point is when calculating, and need select elastic parameter rightly.Simultaneously, it does not exist the division of handling curve and the natural criterion of merging, topology automatically ability to transform a little less than, in application, relatively be suitable for allowing mutual situation.The major advantage of geometric type Snake model is that topological adaptivity is strong, and weak point is not have good topological retentivity, and computing velocity is slow, is unsuitable for real-time application.
No matter take above-mentioned any pattern to carry out image segmentation, the rim detection link that all is absolutely necessary.At present, the main method of implementing rim detection is to carry out the template computing by means of differentiating operator, differentiating operator that some are classical such as gradient operator, Robert operator, Sobel operator, Prewitts operator, Laplace operator, Canny operator, LOG operator etc.The local extremum detection method that they provide can provide absolute amplitude and the directional information that pixel grey scale changes.
Summary of the invention
An object of the present invention is to provide a kind of computing method of implementing rim detection, it not only can calculate the absolute change value about pixel grey scale, and can calculate relative changing value about pixel grey scale, this method can be from the overall situation and the regional area of image, embody the rule that pixel grey scale changes and distributes, have feasibility, simplicity, the extensibility of calculating.
Another object of the present invention provides a kind of deformable contour outline extracting system, and on the one hand, this system can set up the system of a complete closure between bottom-up information and high-rise vision, rather than the open cycle system of " top-down " or " bottom-up "; On the other hand, this system is with high-rise visual knowledge quantification as much as possible, to give full play to the directive function that high-rise visual knowledge extracts profile.
For achieving the above object, the invention provides a kind of freely differences calculus and deformable contour outline extracting system, it comprises: image storage part spare 1, the parts 2 of implementing the directive function of high-rise vision, basic polygonal shape transform component 3, orthogonal line position transform component 4, freely differences calculus parts 5, objective contour extract parts 6, the minimized parts 7 of energy function.
The several rules of parts 2 by providing of the directive function of high-rise vision is provided, the people is carried out quantification as much as possible to the understanding of the objective contour of needs extraction, with the execution of instructing profile to extract, these rules comprise the account form that position, initial profile evolution mode, the pixel grey scale of initial profile change, basic each bar limit concatenate rule in evolutionary process of polygon etc.The account form that changes about pixel grey scale has adopted freely differences calculus parts 5 provided by the invention to realize, it provides a kind of method of relative value that can the calculating pixel grey scale change, certainly, and the absolute value that it also can the calculating pixel grey scale change.Objective contour extracts the optimal value that parts 6 adopt the method ferret out profile of dynamic programming.The method that the contour curve feature that the minimized parts 7 of energy function provide tolerance to be extracted is approached the objective contour curvilinear characteristic.
Freely differences calculus provided by the invention is used to implement rim detection, embodies the rule that pixel grey scale changes and distributes more fully; Deformable contour outline extracting system provided by the invention has been set up the system of a closure between high-rise vision and bottom knowledge, like this, under the directive function of high-rise visual knowledge, this contour outline extracting system is a self-adjusting system that is in mobile equilibrium.It not only can improve speed and the quality of cutting apart effectively, also has adaptability widely simultaneously, and for example, this technology can be applicable in the multimedia equipment of interactive image editing device etc.The present invention has important significance for theories and using value in fields such as image segmentation.
Description of drawings
Fig. 1 utilizes freely differences calculus and deformable contour outline extracting system to carry out the information flow chart of image segmentation.
Fig. 2 is the control system that realizes process shown in Figure 1.
Fig. 3 is the process flow diagram of the sequence of operation of first embodiment.
Fig. 4 is the people's face image that provides among first embodiment.
Fig. 5 is the given initial profile of user.
Fig. 6 is in first embodiment, the people's face mask curve that adopts method of the present invention to extract.
Fig. 7 is when the calculating pixel grey scale change, the mode that gradient direction changes to both sides from initial position.
Fig. 8 is when the calculating pixel grey scale change, gradient direction mode from inside to outside.
Fig. 9 is when the calculating pixel grey scale change, the mode of gradient direction ecto-entad.
Figure 10 is when the calculating pixel grey scale change, the mode that gradient direction changes to initial position from both sides.
Figure 11 shows that the people's face mask curve that in embodiment two, extracts.
Figure 12 shows that head part MR image among the embodiment three, its middle conductor AB is the given initial profile of user, C, the D obligatory point for providing in addition.
Figure 13 shows that among the embodiment three head part MR image being implemented profile extracts the result.
Figure 14 extracts the application examples 1 of profile of people's face, eye, mouth.
Figure 15 extracts the application examples 2 of profile of people's face, eye, mouth.
Figure 16 extracts the application examples 3 of profile of people's face, eye, mouth.
Embodiment
Core concept of the present invention is: utilize freely differences calculus to carry out rim detection, utilize deformable contour outline extracting system to carry out the extraction of objective contour.
For ease of clearly understanding above-mentioned purpose of the present invention, advantage, feature and purposes, describe image segmentation algorithm of the present invention in detail below in conjunction with accompanying drawing.
Accompanying drawing 1 is a core frame of the present invention.Label 1 is that image storage part spare, label 2 are that freely differences calculus parts, label 6 for objective contour extract parts, label 7 be energy function minimized parts for basic polygonal shape transform component, label 4 for orthogonal line position transform component, label 5 for parts, the label 3 of the directive function of the high-rise vision of enforcement.Dotted line means that this contour outline extracting system is a self-adjusting system that is in mobile equilibrium among Fig. 1, and dot-and-dash line means that the parts 2 of the directive function of implementing high-rise vision can determine whether to participate in the self-regulating process of contour outline extracting system according to practical application request.If participate in, will be after the execution unit 6 according to the direction execution unit 2 in path 1; If do not participate in, will be after the execution unit 6 according to the direction execution unit 3 in path 2.
In accompanying drawing 2, label 8 is the control assembly that links to each other with parts 1-7 among Fig. 1, is used for controlling the start and stop of the operation of each parts or the transmission of the data between them.Control assembly 8 is a computing machine that comprises storer and CPU, and as shown in Figure 2, is connected with imaging device 9, display device 10 and input media 11 through each interface.
Embodiment one: extract people's face mask.
The operation of first embodiment of the invention is then described.Fig. 3 carries out the operational flowchart of image segmentation according to first embodiment for expression.For the ease of understanding the operation of first embodiment, Fig. 4 shows the image of a width of cloth people face.This facial image is of a size of 92 * 112, and gray level is 256 grades.The purpose of present embodiment is to extract the profile of people face.This means that the present invention can be applied in people's the extraction system of face contour.Mainly comprise the steps:
Step 1: read in the two dimensional image raw data;
Step 2: image is carried out pre-service;
Step 3: the user imports initial point range;
Step 4: the initialization of the parameter of deformable contour outline extracting system;
Step 5: construct basic polygon;
Step 6: calculate the orthogonal line;
Step 7: calculate and embody the gradient that pixel grey scale changes;
Step 8: adopt the method for dynamic programming to calculate objective contour;
Step 9: judge that whether objective contour meets the condition of convergence, if meet, then withdraws from; Otherwise, the parameter of renewal deformable contour outline extracting system, repeating step five to nine.
The detailed method of operation of each step is as follows:
Step 1: read in the two dimensional image raw data.
Step 2: image is carried out pre-service.In the present embodiment, do not need image is carried out special pre-service.
Step 3: the given initial point range of user.According to the position of objective contour in image-region,, might as well be made as V here along along the given successively a series of initial points of (or contrary) clockwise
i(i=0,1 ..., n), i represents to connect order.With V
i(i=0,1 ..., n) connect one section broken line can forming a closed polygon or non-closure in turn, be called basic polygon or basic sets of line segments.
Step 4: the initialization of the parameter of deformable contour outline extracting system.
Need the given initial parameter of user to comprise:
(1) initial profile evolution mode has following five kinds of selections:
(1.1) can be to inside and outside two changes of direction;
(1.2) expand outwardly;
(1.3) inwardly shrink;
(1.4) be split into a closed or non-closed curve by initial segment, the end points of initial segment keeps;
(1.5) be split into a closed or non-closed curve by initial segment, the end points of initial segment does not keep;
(2) account form of pixel grey scale variation:
(2.1) gradient direction changes to both sides from initial position;
(2.2) gradient direction from inside to outside;
(2.3) gradient direction ecto-entad;
(2.4) gradient direction changes to initial position from both sides.
These four kinds of modes are respectively shown in accompanying drawing 7,8,9,10.In Fig. 7～10, dotted line is represented objective contour, and solid line is represented basic polygon, v
I, 0For on the AB of limit a bit, v
I, M, v
I ,-MBe a process point v
I, 0The orthogonal line of limit AB on point, wherein, v
I, MBe positioned at the outside of polygon ABCD, V
I ,-MBe positioned at the inside of polygon ABCD.Arrow shows for the AB of limit, the direction of graded.
(3) adopt segmentation to develop or continuous evolution mode between basic each bar limit of polygon;
(4) get the outward flange or the inward flange of objective contour;
(5) get the absolute difference or the relative difference of graded;
(6) Initial Hurdle of gradient.
In the present embodiment, above-mentioned every selection is as follows:
(1) initial profile evolution mode: expand outwardly;
(2) account form of pixel grey scale variation: gradient direction from inside to outside;
(3) adopt continuous evolution mode between basic each bar limit of polygon;
(4) get the outward flange of objective contour;
(5) relative difference of the plain graded of capture;
(6) Initial Hurdle of gradient: 10.
Step 5: construct basic polygon.With V
i(i=0,1 ..., n-1) be docile and obedient (or contrary) clockwise and connect successively.
Step 6: calculate the orthogonal line.
For polygonal substantially each bar limit, as V
iV
I+1(i=0,1 ..., n-1), based on the Bresenham line algorithm, calculate this limit the pixel of process.Calculate orthogonal line respectively through this limit of each pixel.
Step 7: calculate and embody the gradient that pixel grey scale changes.Relative difference according to the direction calculating grey scale change from inside to outside of step 4 appointment.
Step 8: adopt the method for dynamic programming to calculate objective contour.Adopt the least energy curve of the method calculating of dynamic programming corresponding to formula (1).If corresponding to V
iV
I+1(i=0,1 ..., n-1), V
n=V
0The least energy curve be respectively
Connect into the curve of a closure according to the continuous evolution mode of step 4 appointment.
Wherein every implication is as follows:
I: representative image;
v
k(k=0,1 ...): represent the profile candidate point, and v
n=v
0
D (v
k, v
K+1): some v
k, v
K+lBetween distance;
I (v
k): v in the image I
kThe gray scale of individual pixel;
I
Character: after image I carried out certain conversion, certain characteristic image that obtains;
I
Character(v
k): characteristic image I
CharacterIn v
kThe eigenwert of individual pixel;
I (v
k Constrain): in image I, have the some v of constraint condition
k
I
Character(v
k Constrain): in image I
CharacterIn, the some v with constraint condition
k
w
i(i=1,2 ..., 7): every weights.
Step 9: judge that whether objective contour meets the condition of convergence, if meet, then withdraws from; Otherwise, upgrade the parameter of deformable contour outline extracting system, the curve that obtains for step 8, from more arbitrarily, be docile and obedient (or contrary) clockwise and get every certain step-length and a bit constitute a new basic polygon, repeating step five to nine is till convergence.The condition of convergence finishes for when gross energy cyclic fluctuation occurs or reaches maximum iteration time.
Accompanying drawing 5 shows the initial profile of user's appointment.The image of accompanying drawing 6 is the face mask curve through obtaining after 3 iteration.
Embodiment two: extract people's face mask.
The operation of second embodiment of the invention and the operation of first embodiment unique different be the absolute difference of in step 4, getting graded, accompanying drawing 11 shows the face mask curve of extraction.Be not difficult to find out with the contrast of embodiment two by embodiment one, in deformable contour outline extracting system provided by the invention, during based on freely differences calculus calculating pixel grey scale change, can select the pattern of calculating neatly according to the characteristics of image.
Embodiment three: extract people's limbic brain.
A kind of embodiments of the present invention of the constraint condition that provides according to the characteristics of objective contour are provided present embodiment.Figure 12 shows that head part MR image, picture size is 253 * 275, and gray level is 256 grades.Initial profile is one section straight line of A, B point-to-point transmission, and the scope of putting on the orthogonal line of choosing is limited in A, C, 4 rectangular areas that constituted of B, D.The account form that pixel grey scale changes be gradient direction from inside to outside, get relative difference.Evolutionary process is that initial profile is expanded to both sides respectively, and is split into two curves, and these two curves locate keep connection status A, B at 2, form the curve of a closure; Subsequently, more arbitrarily, choose a bit every certain step-length along clockwise direction from the curve, constitute a new closed polygon, continue to develop, up to convergence.The result that Figure 13 obtains for 2 times for iteration.
We are at DELL Inspiron (TM) 640m n-Series notebook computer (CPU:T2050 1.60GHz, the 1GB internal memory), operating system is that Windows XP, development language are to have realized algorithm described in the invention in the development environment of VC++2005 and MATLAB7.1, and tests on the different data set of several classes.Figure 14, Figure 15, Figure 16 show the application example of the present invention in the extraction system of the main geometric properties curve of human face.In these three groups of application examples, extracted people's face mask, mouth profile, eye contour curve, wherein, the eye profile comprises the profile of eye socket and the profile of eyeball.
These several groups of examples have fully proved the validity of our algorithms.
Claims (15)
1. freely differences calculus and deformable contour outline extracting system is characterized in that comprising:
Image storage part spare;
Implement the parts of the directive function of high-rise vision;
Basic polygonal shape transform component;
Orthogonal line position transform component;
The freely differences calculus parts;
Objective contour extracts parts;
The minimized parts of energy function.
2. freely differences calculus as claimed in claim 1 and deformable contour outline extracting system, the parts that it is characterized in that implementing the directive function of high-rise vision comprise by the rule of formulating people's subjective vision are carried out quantification to the understanding of profile.
3. freely differences calculus as claimed in claim 2 and deformable contour outline extracting system is characterized in that comprising the candidate point and the order of connection thereof of the initial profile that the characteristics according to objective contour provide.
4. freely differences calculus as claimed in claim 2 and deformable contour outline extracting system is characterized in that comprising that profile extracts is the outward flange or the inward flange of target.
5. as any described freely differences calculus and deformable contour outline extracting system in the claim 2,3,4, it is characterized in that comprising the constraint condition that the characteristics according to objective contour provide.
6. as any described freely differences calculus and deformable contour outline extracting system in the claim 2,3, it is characterized in that comprising the evolution mode of the profile that provides, comprise that profile expands outwardly, profile inwardly shrinks, profile is split into the curve isotype to inside and outside two changes of direction, profile by initial straight line segment simultaneously.
7. freely differences calculus as claimed in claim 1 and deformable contour outline extracting system is characterized in that described basic polygonal shape transform component stores the profile candidate point of the described skeleton pattern of a plurality of formations, and the order that the profile candidate point is coupled together.
8. freely differences calculus as claimed in claim 1 and deformable contour outline extracting system is characterized in that orthogonal line position transform component comprises the orthogonal line of calculating corresponding to the selected point on each limit of basic polygon.
9. as claim 1,2 described freely differences calculus and deformable contour outline extracting systems, it is characterized in that comprising the executive mode that freely differences calculus is provided.
10. as claim 1,2,9 described freely differences calculus and deformable contour outline extracting systems, it is characterized in that the freely differences calculus parts comprise for polygonal substantially a certain limit, have the mode of four kinds of calculating pixel grey scale change (hereinafter referred to as gradient): gradient direction is inboard by the outside sensing of basic polygon, gradient direction is changed by the initial position on the basic polygon of two side direction limit to both sides variation, gradient direction by the initial position on basic polygon limit by the inboard directed outside of basic polygon, gradient direction.
11., it is characterized in that comprising absolute difference or the relative difference of choosing freely differences calculus as claim 2,8,9,10 described freely differences calculus and deformable contour outline extracting systems.
12. freely differences calculus as claimed in claim 1 and deformable contour outline extracting system is characterized in that objective contour extracts parts and comprises the method extraction objective contour that adopts dynamic programming.
13., it is characterized in that objective contour extracts parts and comprises each bar limit concatenate rule in evolutionary process of basic polygon as claim 1,2,7,12 described freely differences calculus and deformable contour outline extracting systems.
14. freely differences calculus as claimed in claim 1 and deformable contour outline extracting system, it is characterized in that energy function shape that the minimized parts of energy function comprise as:
Implication every in the formula is as follows:
I: representative image;
v
k(k=0,1 ...): represent the profile candidate point, and v
n=v
0
D (v
k, v
K+1): some v
k, v
K+1Between distance;
I (v
k): v in the image I
kThe gray scale of individual pixel;
P I (v
k): some v
kFreely differences calculus in image I;
I
Character: after image I carried out certain conversion, certain characteristic image that obtains;
I
Character(v
k): characteristic image I
CharacterIn v
kThe eigenwert of individual pixel;
P I
Character(v
k): some v
kAt I
CharacterIn freely differences calculus;
I (v
Kconstrain): in image I, have the some v of constraint condition
k
I
Character(v
Kconstrain): in image I
CharacterIn, the some v with constraint condition
k
w
i(i=1,2 ..., 7): every weights.
15., it is characterized in that adopting basic polygonal shape transform component for the situation of objective contour curve for closure as described freely differences calculus of claim 1-14 and deformable contour outline extracting system; For the objective contour curve is the situation of non-closure, then adopts basic sets of line segments transform component.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNA2007101043926A CN101136105A (en) | 2007-05-11 | 2007-05-11 | Freely differences calculus and deformable contour outline extracting system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNA2007101043926A CN101136105A (en) | 2007-05-11 | 2007-05-11 | Freely differences calculus and deformable contour outline extracting system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN101136105A true CN101136105A (en) | 2008-03-05 |
Family
ID=39160192
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNA2007101043926A Pending CN101136105A (en) | 2007-05-11 | 2007-05-11 | Freely differences calculus and deformable contour outline extracting system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101136105A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106056598A (en) * | 2016-05-27 | 2016-10-26 | 哈尔滨工业大学 | Line segment detection and image segmentation fusion-based satellite high-resolution image building contour extraction method |
| CN108509866A (en) * | 2018-03-12 | 2018-09-07 | 华南理工大学 | A kind of facial contour extraction method |
-
2007
- 2007-05-11 CN CNA2007101043926A patent/CN101136105A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106056598A (en) * | 2016-05-27 | 2016-10-26 | 哈尔滨工业大学 | Line segment detection and image segmentation fusion-based satellite high-resolution image building contour extraction method |
| CN108509866A (en) * | 2018-03-12 | 2018-09-07 | 华南理工大学 | A kind of facial contour extraction method |
| CN108509866B (en) * | 2018-03-12 | 2020-06-19 | 华南理工大学 | Face contour extraction method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Amor et al. | Action recognition using rate-invariant analysis of skeletal shape trajectories | |
| Despinoy et al. | Unsupervised trajectory segmentation for surgical gesture recognition in robotic training | |
| JP4625129B2 (en) | Monocular tracking of 3D human motion using coordinated mixed factor analysis | |
| Ibraheem et al. | Survey on various gesture recognition technologies and techniques | |
| CN101964064B (en) | Human face comparison method | |
| CN102592112B (en) | Method for determining gesture moving direction based on hidden Markov model | |
| Aarno et al. | Motion intention recognition in robot assisted applications | |
| CN109508686B (en) | Human behavior recognition method based on hierarchical feature subspace learning | |
| CN104573621A (en) | Dynamic gesture learning and identifying method based on Chebyshev neural network | |
| Song et al. | Robot complex motion learning based on unsupervised trajectory segmentation and movement primitives | |
| Sahoo et al. | Hand gesture recognition using densely connected deep residual network and channel attention module for mobile robot control | |
| El-Sawah et al. | A framework for 3D hand tracking and gesture recognition using elements of genetic programming | |
| Lahiani et al. | Real Time Static Hand Gesture Recognition System for Mobile Devices. | |
| CN101136105A (en) | Freely differences calculus and deformable contour outline extracting system | |
| Li | Design of human-computer interaction system using gesture recognition algorithm from the perspective of machine learning | |
| Herzog et al. | Motion imitation and recognition using parametric hidden markov models | |
| Aleotti et al. | Arm gesture recognition and humanoid imitation using functional principal component analysis | |
| Itkarkar et al. | A study of vision based hand gesture recognition for human machine interaction | |
| Chen et al. | A dynamic gesture interface for virtual environments based on Hidden Markov Models | |
| Aleotti et al. | Functional principal component analysis for recognition of arm gestures and humanoid imitation | |
| Panduranga et al. | Dynamic hand gesture recognition system: a short survey | |
| Ho et al. | An adaptive visual attentive tracker for human communicational behaviors using HMM-based TD learning with new state distinction capability | |
| Rao | Shape vectors: an efficient parametric representation for the synthesis and recognition of hand script characters | |
| Zhao et al. | Research on the application of body posture action feature extraction and recognition comparison | |
| Wang et al. | Model for deep learning-based skill transfer in an assembly process |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Open date: 20080305 |