CN103218824A - Motion key frame extracting method based on distance curve amplitudes - Google Patents
Motion key frame extracting method based on distance curve amplitudes Download PDFInfo
- Publication number
- CN103218824A CN103218824A CN2012105660916A CN201210566091A CN103218824A CN 103218824 A CN103218824 A CN 103218824A CN 2012105660916 A CN2012105660916 A CN 2012105660916A CN 201210566091 A CN201210566091 A CN 201210566091A CN 103218824 A CN103218824 A CN 103218824A
- Authority
- CN
- China
- Prior art keywords
- motion
- key frame
- frame
- distance
- key
- 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
- Processing Or Creating Images (AREA)
Abstract
The invention discloses a motion key frame extracting method based on distance curve amplitudes. The motion key frame extracting method based on the distance curve amplitudes is used for extracting key postures of motion to describe an original motion sequence. The motion key frame extracting method based on the distance curve amplitudes includes the steps: firstly, selecting a group of new key frame distances to be as distance characteristics; secondly, carrying out dimensionality reduction on the distance characteristics through a principal component analysis method (PCA method for short), extracting first-dimension principal components, removing noises through smoothing filtering, and a characteristic curve is achieved, wherein substantive characteristics of an original motion can be well reflected; thirdly, acquiring initial key frames through extraction of local extreme points on the characteristic curve; evenly inserting corresponding numbers of frames among initial adjacent key frames according to the differences of amplitudes of the characteristic curve, then combining overstocked key frames, and final key frame collection is achieved. A lot of experimental data show that the motion key frame extracting method based on the distance curve amplitudes can not only satisfy good effects on visual generality of the motion, but also has low a compression ratio and an error ratio, and a processing procedure can be automatically finished without manual interference.
Description
Technical field
The present invention relates to the human body motion capture technology, more particularly, relate to the key-frame extraction of human motion.
Background technology
In nearly decades, the human body motion capture technical development is rapid, and motion-captured importance also increases thereupon.Be extensive use of motion capture system in film and the recreation.Be accompanied by problem of existence,, how handle the focus that huge movement capturing data has become domestic and international research because the huge scale of motion capture database that causes of seizure data is also very huge.
The key frame technology is a kind of effective solution, and the frame of selecting most important most critical in the motion is represented whole motion sequence as key frame, and other non-key frame and important not as these key frames can be calculated through interpolation algorithm by key frame.Because the key-frame extraction technology occupies irreplaceable effect in the expression of whole motion sequence, not only can accelerate processing speed of data, but also the storage of movement capturing data, compress, browse and reuse aspect remarkable advantages is arranged.
That is to say that " key-frame extraction of human motion " is at one section motion sequence, extracts the critical movements attitude of some automatically, motion has a vision generality preferably to this section, simultaneously can carry out restructure from motion again, the reduction original motion keeps a lower error rate.
Key-frame extraction should satisfy following needs, and on the one hand, the key frame under certain compressibility can effectively be summarized the original motion sequence.On the other hand, key frame can be used for rebuilding as far as possible accurately the original motion sequence.Quality for key-frame extraction also has two evaluation criterias except subjective vision is judged: error rate and ratio of compression.
Up to the present, extraction method of key frame has uniform sampling, and curve is simplified, based on the method for cluster etc.
The human body motion capture data are to adopt the BVH form, and its exercise data is made up of the exercise data of a frame one frame, and the data of each frame are comprising the pose information of motion, and each posture is made up of all articulation points of human body.Skeleton model shown among the BVH data importing MATLAB is seen accompanying drawing 2, this model comprises 31 articulation points (articulation point that wherein will use is marked), each articulation point adopts tree structure, root node (root) is the root node of tree-like human skeleton, successively extends to form each subtree of root node to each terminal joint of human skeleton from the root articulation point.Wherein, the root joint is represented that by 3 translational movements and 3 rotation amounts each is represented other non-root articulation points by 3 rotation amounts.One has 96 degree of freedom.The current location of the translation of root decision human motion, the rotation decision human body of root towards; The rotation of other each articulation points is illustrated in the direction of this articulation point under the local coordinate system at his father's articulation point place, and they determine the human posture jointly.
The data of human body motion capture are the human posture's sequences that is obtained by the discrete time point sampling, and each sampled point is a frame, and the posture of each frame is determined jointly by 31 articulation points.Like this, i at any time, the human posture is expressed as
Wherein
And
Position and the direction of representing the root articulation point respectively, i.e. translational movement and rotation amount,
J=2 ..., the direction of the non-root articulation point of 31 expressions.
Summary of the invention
The objective of the invention is to: proposed based on a kind of new aitionastic essential characteristic of distance feature curve, amplitude according to characteristic curve is carried out key-frame extraction twice, solve emphatically at one section motion sequence, can extract the critical movements attitude of some automatically, motion has a vision generality preferably for this section, simultaneously rebuild the original motion sequence again, keep a lower error rate.
The present invention is based on the method that curve is simplified, a kind of motion key-frame extraction based on the distance Curve amplitude is provided, comprise the steps:
S1, one group of joint distance of selection are as distance feature;
S2, employing PCA method are extracted the first dimension major component, adopt smothing filtering to remove noise, obtain characteristic curve;
S3, the Local Extremum of passing through to extract on the characteristic curve obtain initial key frame;
S4, between adjacent described initial key frame, the characteristic curve amplitude difference of calculating, and then insert uniformly additional key frame according to uniform sampling;
S5, overstocked described additional key frame and the described initial key frame of merging stay final key frame set.
Wherein, step S1, the selection of described joint distance, the logical semantics mode of following table is made:
| D1: the degree of crook of left leg | D4: left arm degree of crook | D7: the degree of bowing/come back/shake the head |
| D2: the degree of crook of right leg | D5: right arm degree of crook | D8: the degree of waving of left arm |
| D3: the distance between the bipod | D6: the degree of bending over | D9: the degree of waving of right arm |
Then a motion frame is represented to become: θ=(d1, d2, d3, d4, d5, d6, d7, d8, d9), d1 wherein .., d9 are the joint distance of choosing in the S1 step with logical semantics.
The method of PCA described in the step S2 is divided into following five steps:
S21: the mean value that makes up 9 distance feature samples shown in the S1
I=1 wherein ... 9 represent this 9 distance feature samples, and T is the total frame number of this motion;
S22: the difference of calculating the original value and the mean value of these 9 distance feature
Make up the matrix of differences D=[Δ θ of distance feature then
1.... and Δ θ
9];
S23: calculate covariance matrix C=DD
T, D wherein
TTransposition for D;
S24: calculate the eigenvalue of covariance matrix, and corresponding proper vector L;
S25: extract proper vector
So just rebuild one group of 9 new dimension major component
Its value is by contribution rate series arrangement from big to small; Extract the first dimension major component then, maximum contribution rate just
This motion M and then can represent to become:
N wherein
FrameTotalframes for motion;
In addition, select the Lowess smothing filtering for use among the step S2.
Under the optimal way, the implementation procedure of step S4 is:
S41: the amplitude difference of calculating the characteristic curve between the adjacent described initial key frame;
S42 a: threshold value is set; If described amplitude difference 〉=described threshold value is then inserted the described additional key frame of a frame or multiframe.
Under the optimal way, the implementation method of step S5:, adopt the method for the frame number between the adjacent key frame of restriction to merge at described additional key frame and the overstocked situation of described initial key frame.
The present invention compared with prior art has the following advantages:
1, selected one group of new distance feature, the characteristic curve of extraction can better react the essential characteristic of human motion.
2, adopt the method for twice key-frame extraction, extract the border key frame as initial key frame according to the Local Extremum of characteristic curve for the first time, adopt the automatic interleave of uniform sampling algorithm and merge overstocked frame according to characteristic curve amplitude difference for the second time, the appointment sampling frame number that automatism need not be artificial is preferably arranged, and can better adapt to motion, extract less key frame in mild motion place, extract more crucial frame number in violent motion place.Can well summarize original motion, a lower error rate and compressibility are arranged again simultaneously.
Description of drawings
Fig. 1 process flow diagram of the present invention.
The skeleton illustraton of model that Fig. 2 shows in MATLAB.
Nine joint distance feature presentation graphs that Fig. 3 selects, θ=(d1, d2, d3, d4, d5, d6, d7, d8, d9).
Fig. 4 ' plays football ' motion characteristics curve and key frame distribution plan; Here circle is represented initial key frame, and asterisk is expressed as final key frame.
Fig. 5 ' plays football ' motion different extraction method of key frame comparative result exploded views; Wherein, (a) the inventive method; (b) uniform sampling method, the oval over-sampling of representing is sampled with owing; (c) curve is simplified, and the oval over-sampling of representing is sampled with owing; (d) has only the hypercomplex number distance method.
The key-frame extraction comparison diagram of Fig. 6 similar movement type (is example to walk motion); Wherein, (a) big the walking of arm swing; (b) little the walking of arm swing; (c) cheerful and light-hearted walking; (d) brutal walking.
Fig. 7 adopts the compressibility comparison diagram of the inventive method to six kinds of different motion types (play football, jump, run-stop, walking, dance, walk-jump-walk).
Four kinds of distinct methods of Fig. 8 (be the inventive method successively, the hypercomplex number distance method, curve is simplified, the uniform sampling method) to six kinds of reconstruction error comparison diagrams of sampling and moving; (a) error (extracting 33 key frames) of playing football; (b) error of Tiaoing (extracting 24 key frames); The error of (c) running-stopping (extracting 11 key frames); (d) error of walking (extracting 16 key frames); (e) error of Tiao Wuing (extracting 37 key frames); (f) error of walking-jump-walking (extracting 50 key frames).
Embodiment
Technical scheme of the present invention is: this paper invention is based on the motion extraction method of key frame of distance Curve amplitude, comprise and select one group of new joint distance as distance feature, adopt the PCA method to extract the first dimension major component, adopt smothing filtering to remove noise, obtain characteristic curve, obtain initial key frame by the Local Extremum of extracting on the characteristic curve, between initial adjacent key frame, the characteristic curve amplitude difference of calculating, and then, remerge overstocked key frame and obtain final key frame set according to the corresponding frame number of the uniform insertion of uniform sampling.Accompanying drawing 1 is depicted as algorithm flow chart of the present invention, and it specifically comprises following sport technique segment:
1. select one group of new joint distance as distance feature.
The present invention selects nine human synovial distances with logical semantics as distance feature, reflects the essential characteristic of this motion, θ=(d1, d2, d3, d4, d5, d6, d7, d8 d9), sees accompanying drawing 3.Its logical semantics sees Table 1.So just 96 original dimension data are dropped to 9 dimensions.
The logical semantics of nine distance feature that table 1. extracts
| D1: the degree of crook of left leg | D4: left arm degree of crook | D7: the degree of bowing/come back/shake the head |
| D2: the degree of crook of right leg | D5: right arm degree of crook | D8: the degree of waving of left arm |
| D3: the distance between the bipod | D6: the degree of bending over | D9: the degree of waving of right arm |
Wherein, the human synovial distance can be determined according to the needs of concrete human body motion capture, nine optimal selections that are characterized as present embodiment that the present invention selects, it also is the optimal selection that most human body molar behaviors are caught, step for activities, as the action of arm, should specifically select the articulation point of arm to be achieved.
2. adopt the PCA method that these nine distance feature are analyzed, further dimensionality reduction extracts its first dimension major component, adopts the Lowess smothing filtering, relevant parameters is set characteristic curve is carried out denoising, obtains the distance feature curve.
PCA analyzes can be divided into following five steps:
Step 1: the mean value that makes up 9 distance feature samples shown in the S1
I=1 wherein ... 9 represent this 9 distance feature samples, and T is the total frame number of this motion;
Step 2: the difference of calculating the original value and the mean value of these 9 distance feature
Make up the matrix of differences D=[Δ θ of distance feature then
1.... and Δ f θ
9];
Step 3: calculate covariance matrix C=DD
T, D wherein
TTransposition for D;
Step 4: calculate the eigenvalue of covariance matrix, and corresponding proper vector L;
Step 5: extract proper vector
So just rebuild one group of 9 new dimension major component
Its value is by contribution rate series arrangement from big to small; Extract the first dimension major component then, maximum contribution rate just
This motion M and then can represent to become:
N wherein
FrameTotalframes for motion;
Adopt the Lowess smothing filtering, relevant parameters is set curve is carried out denoising, obtained a characteristic curve, can better reflect original motion.
3. for one section motion fragment, key deposits initial key frame.Extract the Local Extremum loc on the characteristic curve, think the border attitude of moving, thereby obtain initial key frame key.Concrete steps are described below:
Step 1: we use " findpeak " function to obtain local maximum in MATLAB;
Step 2: data are reversed, continue to use " findpeak " function to obtain local minimum, and then the data inverted is returned, continue after being convenient to use;
Step 3: merge local maximum and local minimum, obtain Local Extremum loc() according to frame number order from small to large.With the Local Extremum that obtains as initial key frame key.
These initial key frames are a little borders attitudes, and we do not consider the intensity of moving.If for requiring low error requirement and summarize human motion preferably, then need between adjacent initial key frame, insert a frame or multiframe more.We think that if the characteristic curve amplitude between the adjacent initial key frame has a bigger difference then the strong attitude of the motion between this adjacent key frame changes greatly, think that if difference is less the mild attitude of corresponding motion changes less.The split-and-merge algorithm of the distance feature profile amplitude below adopting further extracts key frame wherein.
4. for one section motion fragment, key deposits initial key frame, and Lkey deposits final key frame.Between initial key frame key,, divide or the merging key frame, and then obtain final key frame Lkey then based on the distance feature profile amplitude.Specifically describe as follows:
Step 1: according to characteristic curve, calculate adjacent initial key frame key1, the amplitude difference vary of the characteristic curve between the key2;
Step 2: a threshold value threshold is set.If this vary＜threshold, it is inviolent to illustrate that initial adjacent key frame attitude changes, and does not need interleave, otherwise, change violently between the adjacent initial key frame, wherein need to insert a frame or multiframe more.For example, this is changed to vary=0.4, and we do not need to insert more multiframe; If this is changed to vary=10, we need adopt the mode of uniform sampling to insert 10 frames between key1 and key2, are increased among the initial key frame key.Can obtain a key frame set like this;
Step 3: but the adjacent key frame in the key frame that the obtains set may be overstocked, therefore needs to merge overstocked key frame and obtain final key frame.We adopt the method for the frame number between the adjacent key frame of restriction to merge.If the frame number between the adjacent key frame is greater than a threshold value, we then preserve this key frame in final key frame Lkey.
In the implementation procedure of the present invention, the value of concrete parameter determines that as required that is: span, threshold value etc. all do not have concrete the qualification, and according to circumstances value gets final product.
Below, with specific embodiment of specific implementation of the present invention.Embodiments of the invention are being to implement under the prerequisite with the technical solution of the present invention, provided detailed embodiment and concrete operating process, but protection scope of the present invention are not limited to following embodiment.The example sample is to choose in the CMU database (being the motion capture database that create in graphics laboratory, U.S. Carnegie Mellon University).
Concrete implementation step is:
Step 1: from the CMU database, select a representative motion, ' playing football ' motion (this total frame length that moves is 802 frames).Extract nine joints distance as distance feature θ=(d1, d2, d3, d4, d5, d6, d7, d8, d9).
Step 2: adopt the PCA method that this nine dimensions distance feature is analyzed further dimensionality reduction.Extract its first dimension major component, adopt the Lowess smothing filtering, relevant parameters is set characteristic curve is carried out denoising, obtain our required distance feature curve and then carry out key-frame extraction.
Step 3: adopt among the MATLAB ' findpeak ' function finds the Local Extremum loc on the characteristic curve, thinks the border attitude of moving to obtain initial key frame key.See accompanying drawing 4.
Step 4: based on the division and the merging key frame of distance feature profile amplitude, Lkey deposits final key frame.See accompanying drawing 4.We have extracted Local Extremum and have been represented by circle as initial key frame key, and we obtain final key frame Lkey and are represented by asterisk simultaneously.By research ' playing football ' motion, we find that motion is non-regular movement, and the slow relatively attitude of its first half motion changes little, and the violent attitude of the posterior part of motion changes greatly.When motion when being slow, the characteristic of correspondence profile amplitude is little, do not need especially to insert again key frame, yet, violent when moving, just the key frame that need make new advances according to the division of characteristic curve amplitude difference between adjacent initial key frame because initial key frame can not well react this variation, merges overstocked key frame then.By this method, we can effectively obtain final key frame set.
Step 5: the comparison of different key-frame extraction algorithms.We have adopted diverse ways: the inventive method, and uniform sampling, the method for hypercomplex number distance is simplified and had only to curve, extracts the key frame frame number (ratio of compression is identical) of similar number from one ' playing football ' motion.Distinct methods extracts the comparative result of key frame and sees accompanying drawing 5.We have extracted 33 key frames, have well summarized motion, have avoided over-sampling and have owed the sampling problem.
Step 6: adopt the inventive method that the key frame of similar movement is compared.We are chosen in the motion of walking of four types of certain frame scopes interior (130 frame) and extract key frame.See accompanying drawing 6.We can be from Fig. 6 (a)-(d), and the key frame that the present invention extracts can obviously be distinguished similar movement.
Step 7: adopt six kinds of dissimilar motion sequences of the inventive method test,, jump, run-stop, walking, dance, walk-jump-walk comprising playing football.See accompanying drawing 7.Find that from table 2 compressibility that the inventive method obtains is in 10%.
The ratio of compression of six kinds of type of sports of table 2. relatively
| Type | Play football | Jump | Run-stop | Walk | Dance | Walk-jump- |
| Sum | ||||||
| 802 | 439 | 239 | 343 | 1033 | 1200 | |
| Key frame | 33 | 24 | 11 | 16 | 37 | 50 |
| Ratio of compression (%) | 4.1 | 5.5 | 4.6 | 4.6 | 3.5 | 4.3 |
Then, we calculate absolute average error with following formula:
E=[∑(F(n)-F'(n))
2]/N
Here, F (n) is the original motion data, F'(n) is the exercise data after corresponding the reconstruction, and N multiply by 96 degree of freedom for this motion totalframes.
Can see that from table 3 we adopt the absolute average error after six sampling motion calculation are rebuild, wherein, the number of the data represented extraction key frame of bracket.Accompanying drawing 8, the error ratio of having showed six kinds of sampling motions.
The error rate of four kinds of methods of table 3. relatively
We adopt four kinds of methods to remove to obtain identical crucial frame number.Afterwards, we rebuild motion sequence by approach based on linear interpolation and obtain rebuilding the back error.We can find to use the inventive method lower than the error rate that other method obtains under identical ratio of compression.So this is because it has extracted important key frame and has reached lower error.No matter from table, can find that the inventive method all has a remarkable advantages in the non-regular movement or the motion of rule.
The above; only be the preferable embodiment of the present invention; but protection scope of the present invention is not limited thereto; anyly be familiar with those skilled in the art in the technical scope that the present invention discloses; be equal to replacement or change according to technical scheme of the present invention and inventive concept thereof, all should be encompassed within protection scope of the present invention.
Claims (5)
1. the motion key-frame extraction based on the distance Curve amplitude is characterized in that, comprises the steps:
S1, one group of joint distance of selection are as distance feature;
S2, employing PCA method are extracted the first dimension major component, adopt smothing filtering to remove noise, obtain characteristic curve;
S3, the Local Extremum of passing through to extract on the characteristic curve obtain initial key frame;
S4, between adjacent described initial key frame, the characteristic curve amplitude difference of calculating, and then insert uniformly additional key frame according to uniform sampling;
S5, overstocked described additional key frame and the described initial key frame of merging stay final key frame set.
2. according to the described motion key-frame extraction of claim 1, it is characterized in that among the step S1 based on the distance Curve amplitude, the selection of described joint distance, the logical semantics mode of following table is made:
Then a motion frame is represented to become: θ=(d1, d2, d3, d4, d5, d6, d7, d8, d9), d1 wherein .., d9 are the joint distance of choosing in the S1 step with logical semantics.
3. according to claim 1 or 2 described motion key-frame extraction, it is characterized in that the method for PCA described in the step S2 is divided into following five steps based on the distance Curve amplitude:
S21: the mean value that makes up 9 distance feature samples shown in the S1
I=1 wherein ... 9 represent this 9 distance feature samples, and T is the total frame number of this motion;
S22: the difference of calculating the original value and the mean value of these 9 distance feature
Make up the matrix of differences D=[Δ θ of distance feature then
1.... and Δ θ
9];
S23: calculate covariance matrix C=DD
T, D wherein
TTransposition for D;
S24: calculate the eigenvalue of covariance matrix, and corresponding proper vector L;
S25: extract proper vector
So just rebuild one group of 9 new dimension major component
Its value is by contribution rate series arrangement from big to small; Extract the first dimension major component then, maximum contribution rate just
This motion M and then can represent to become:
N wherein
FrameTotalframes for motion;
In addition, select the Lowess smothing filtering for use among the step S2.
4. according to the described motion key-frame extraction of claim 3, it is characterized in that the implementation procedure of step S4 is based on the distance Curve amplitude:
S41: the amplitude difference of calculating the characteristic curve between the adjacent described initial key frame;
S42 a: threshold value is set; If described amplitude difference 〉=described threshold value is then inserted the described additional key frame of a frame or multiframe.
5. according to the described motion key-frame extraction of claim 4, it is characterized in that the implementation method of step S5 based on the distance Curve amplitude:
At described additional key frame and the overstocked situation of described initial key frame, adopt the method for the frame number between the adjacent key frame of restriction to merge.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2012105660916A CN103218824A (en) | 2012-12-24 | 2012-12-24 | Motion key frame extracting method based on distance curve amplitudes |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2012105660916A CN103218824A (en) | 2012-12-24 | 2012-12-24 | Motion key frame extracting method based on distance curve amplitudes |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN103218824A true CN103218824A (en) | 2013-07-24 |
Family
ID=48816567
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2012105660916A Pending CN103218824A (en) | 2012-12-24 | 2012-12-24 | Motion key frame extracting method based on distance curve amplitudes |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103218824A (en) |
Cited By (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103679747A (en) * | 2013-11-15 | 2014-03-26 | 南昌大学 | Key frame extraction method of motion capture data |
| CN103927776A (en) * | 2014-03-28 | 2014-07-16 | 浙江中南卡通股份有限公司 | Animation curve optimization method |
| CN103942778A (en) * | 2014-03-20 | 2014-07-23 | 杭州禧颂科技有限公司 | Fast video key frame extraction method of principal component characteristic curve analysis |
| CN104331904A (en) * | 2014-10-30 | 2015-02-04 | 大连大学 | Three-dimensional human motion key frame extracting method based on fuse of improved LLE and PCA |
| CN104331911A (en) * | 2014-11-21 | 2015-02-04 | 大连大学 | Improved second-order oscillating particle swarm optimization based key frame extraction method |
| CN104463788A (en) * | 2014-12-11 | 2015-03-25 | 西安理工大学 | Human motion interpolation method based on motion capture data |
| CN106504267A (en) * | 2016-10-19 | 2017-03-15 | 东南大学 | A kind of motion of virtual human data critical frame abstracting method |
| CN106791480A (en) * | 2016-11-30 | 2017-05-31 | 努比亚技术有限公司 | A kind of terminal and video skimming creation method |
| CN104200455B (en) * | 2014-06-13 | 2017-09-15 | 北京工业大学 | A kind of key poses extracting method based on movement statistics signature analysis |
| CN105931270B (en) * | 2016-04-27 | 2018-03-27 | 石家庄铁道大学 | Video key frame extracting method based on gripper path analysis |
| CN107993275A (en) * | 2016-10-26 | 2018-05-04 | 佳能株式会社 | Method and system for the dynamic sampling of the slow-action curve of animation |
| CN108121973A (en) * | 2017-12-25 | 2018-06-05 | 江苏易乐网络科技有限公司 | Key frame extraction method of motion capture data based on principal component analysis |
| CN109241956A (en) * | 2018-11-19 | 2019-01-18 | Oppo广东移动通信有限公司 | Method, apparatus, terminal and the storage medium of composograph |
| CN110415336A (en) * | 2019-07-12 | 2019-11-05 | 清华大学 | High-precision human posture method for reconstructing and system |
| CN110532837A (en) * | 2018-05-25 | 2019-12-03 | 九阳股份有限公司 | Image processing method and household appliance in a kind of article fetching process |
| CN112698196A (en) * | 2021-03-24 | 2021-04-23 | 深圳市三和电力科技有限公司 | High-voltage switch mechanical characteristic monitoring device |
| CN113327228A (en) * | 2021-05-26 | 2021-08-31 | Oppo广东移动通信有限公司 | Image processing method and device, terminal and readable storage medium |
| CN113407371A (en) * | 2020-12-03 | 2021-09-17 | 腾讯科技（深圳）有限公司 | Data anomaly monitoring method and device, computer equipment and storage medium |
| CN114550071A (en) * | 2022-03-22 | 2022-05-27 | 北京壹体科技有限公司 | Method, device and medium for automatically identifying and capturing track and field video action key frames |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1967525A (en) * | 2006-09-14 | 2007-05-23 | 浙江大学 | Extraction method of key frame of 3d human motion data |
-
2012
- 2012-12-24 CN CN2012105660916A patent/CN103218824A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1967525A (en) * | 2006-09-14 | 2007-05-23 | 浙江大学 | Extraction method of key frame of 3d human motion data |
Non-Patent Citations (5)
| Title |
|---|
| 吴刚: "《隐式多项式曲线理论与应用》", 31 December 2011, article "隐式多项式曲线理论与应用", pages: 39-42 * |
| 彭淑娟: "基于中心距离特征的人体运动序列关键帧提取", 《系统仿真学报》, vol. 24, no. 3, 31 March 2012 (2012-03-31), pages 565 - 569 * |
| 杨跃东等: "基于几何特征的人体运动捕获数", 《系统仿真学报》, vol. 19, no. 10, 31 May 2007 (2007-05-31), pages 2229 - 2234 * |
| 沈军行等: "从运动捕获数据中提取关键帧", 《计算机辅助设计与图形学学报》, vol. 16, no. 5, 31 May 2004 (2004-05-31), pages 719 - 723 * |
| 肖俊等: "三维人体运动特征可视化与交互式分割", 《软件学报》, vol. 19, no. 8, 31 August 2008 (2008-08-31), pages 1995 - 2003 * |
Cited By (26)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103679747A (en) * | 2013-11-15 | 2014-03-26 | 南昌大学 | Key frame extraction method of motion capture data |
| CN103679747B (en) * | 2013-11-15 | 2016-08-17 | 南昌大学 | A kind of key frame extraction method of motion capture data |
| CN103942778A (en) * | 2014-03-20 | 2014-07-23 | 杭州禧颂科技有限公司 | Fast video key frame extraction method of principal component characteristic curve analysis |
| CN103927776A (en) * | 2014-03-28 | 2014-07-16 | 浙江中南卡通股份有限公司 | Animation curve optimization method |
| CN104200455B (en) * | 2014-06-13 | 2017-09-15 | 北京工业大学 | A kind of key poses extracting method based on movement statistics signature analysis |
| CN104331904A (en) * | 2014-10-30 | 2015-02-04 | 大连大学 | Three-dimensional human motion key frame extracting method based on fuse of improved LLE and PCA |
| CN104331911A (en) * | 2014-11-21 | 2015-02-04 | 大连大学 | Improved second-order oscillating particle swarm optimization based key frame extraction method |
| CN104463788A (en) * | 2014-12-11 | 2015-03-25 | 西安理工大学 | Human motion interpolation method based on motion capture data |
| CN104463788B (en) * | 2014-12-11 | 2018-02-16 | 西安理工大学 | Human motion interpolation method based on movement capturing data |
| CN105931270B (en) * | 2016-04-27 | 2018-03-27 | 石家庄铁道大学 | Video key frame extracting method based on gripper path analysis |
| CN106504267A (en) * | 2016-10-19 | 2017-03-15 | 东南大学 | A kind of motion of virtual human data critical frame abstracting method |
| CN106504267B (en) * | 2016-10-19 | 2019-05-17 | 东南大学 | A kind of motion of virtual human data critical frame abstracting method |
| CN107993275A (en) * | 2016-10-26 | 2018-05-04 | 佳能株式会社 | Method and system for the dynamic sampling of the slow-action curve of animation |
| CN106791480A (en) * | 2016-11-30 | 2017-05-31 | 努比亚技术有限公司 | A kind of terminal and video skimming creation method |
| CN108121973A (en) * | 2017-12-25 | 2018-06-05 | 江苏易乐网络科技有限公司 | Key frame extraction method of motion capture data based on principal component analysis |
| CN110532837A (en) * | 2018-05-25 | 2019-12-03 | 九阳股份有限公司 | Image processing method and household appliance in a kind of article fetching process |
| CN109241956A (en) * | 2018-11-19 | 2019-01-18 | Oppo广东移动通信有限公司 | Method, apparatus, terminal and the storage medium of composograph |
| CN109241956B (en) * | 2018-11-19 | 2020-12-22 | Oppo广东移动通信有限公司 | Method, device, terminal and storage medium for synthesizing image |
| CN110415336A (en) * | 2019-07-12 | 2019-11-05 | 清华大学 | High-precision human posture method for reconstructing and system |
| CN110415336B (en) * | 2019-07-12 | 2021-12-14 | 清华大学 | High-precision human body posture reconstruction method and system |
| CN113407371A (en) * | 2020-12-03 | 2021-09-17 | 腾讯科技（深圳）有限公司 | Data anomaly monitoring method and device, computer equipment and storage medium |
| CN112698196B (en) * | 2021-03-24 | 2021-06-08 | 深圳市三和电力科技有限公司 | High-voltage switch mechanical characteristic monitoring device |
| CN112698196A (en) * | 2021-03-24 | 2021-04-23 | 深圳市三和电力科技有限公司 | High-voltage switch mechanical characteristic monitoring device |
| CN113327228A (en) * | 2021-05-26 | 2021-08-31 | Oppo广东移动通信有限公司 | Image processing method and device, terminal and readable storage medium |
| CN113327228B (en) * | 2021-05-26 | 2024-04-16 | Oppo广东移动通信有限公司 | Image processing method and device, terminal and readable storage medium |
| CN114550071A (en) * | 2022-03-22 | 2022-05-27 | 北京壹体科技有限公司 | Method, device and medium for automatically identifying and capturing track and field video action key frames |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103218824A (en) | Motion key frame extracting method based on distance curve amplitudes | |
| Zhang et al. | V4d: 4d convolutional neural networks for video-level representation learning | |
| Scetbon et al. | Deep k-svd denoising | |
| CN109685831B (en) | Target tracking method and system based on residual layered attention and correlation filter | |
| Lou et al. | Example-based human motion denoising | |
| CN102521843B (en) | Three-dimensional human body motion analysis and synthesis method based on manifold learning | |
| Wang et al. | 3D human motion editing and synthesis: A survey | |
| CN103164694A (en) | Method for recognizing human motion | |
| CN110135386B (en) | Human body action recognition method and system based on deep learning | |
| CN104834893A (en) | Front-view pedestrian gait period detection method | |
| CN109567789A (en) | Division processing method, device and the readable storage medium storing program for executing of ECG data | |
| CN103679747B (en) | A kind of key frame extraction method of motion capture data | |
| CN104331911A (en) | Improved second-order oscillating particle swarm optimization based key frame extraction method | |
| CN107169988A (en) | A kind of extraction method of key frame based on COS distance hierarchical clustering | |
| US20150187114A1 (en) | Method and apparatus for editing 3d character motion | |
| CN111862167A (en) | Rapid robust target tracking method based on sparse compact correlation filter | |
| CN101241601B (en) | Graphic processing joint center parameter estimation method | |
| CN111598822A (en) | Image fusion method based on GFRW and ISCM | |
| Xiao et al. | An efficient keyframe extraction from motion capture data | |
| CN111210394A (en) | Image enhancement technology based on deep decomposition synthesis network | |
| CN110517336B (en) | Human motion data compression method and device based on main force joint point | |
| CN102819549B (en) | Based on the human motion sequences segmentation method of Least-squares estimator characteristic curve | |
| CN106504267B (en) | A kind of motion of virtual human data critical frame abstracting method | |
| CN115496935A (en) | Hyperspectral image classification method based on deformable convolution | |
| KR101547208B1 (en) | Apparatus and method for reconstructing whole-body motion using wrist trajectories |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C05 | Deemed withdrawal (patent law before 1993) | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20130724 |