CN109829442A - A kind of method and system of the human action scoring based on camera - Google Patents
A kind of method and system of the human action scoring based on camera Download PDFInfo
- Publication number
- CN109829442A CN109829442A CN201910132276.8A CN201910132276A CN109829442A CN 109829442 A CN109829442 A CN 109829442A CN 201910132276 A CN201910132276 A CN 201910132276A CN 109829442 A CN109829442 A CN 109829442A
- Authority
- CN
- China
- Prior art keywords
- key point
- movement
- picture
- human
- human body
- 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
Landscapes
- Image Analysis (AREA)
Abstract
The method and system for the human action scoring based on camera that the invention discloses a kind of, it is characterized in that, it can be according to the markup information of human region and key point in data set, information extraction model is set using deep neural network training human body key point, pass through the model, the key point location information of movement picture and standard operation picture to be scored, the key point positional information calculation similarity acted to two can be extracted, to realize movement scoring.This method and system are capable of the execution scoring of efficiently and accurately, do not depend on expert, it is only necessary to shoot picture or video by camera.
Description
Technical field
The present invention relates to computer deep learning fields, more particularly to a kind of human action scoring based on camera
Method and system.
Background technique
In real life, there are many study course standard operations, such as dancing, body-building, weight lifting, long-jump, high jump, to these
The scoring of movement, error correcting generally requires expert just can be carried out, and has many amateurs, usually only can be against study course mark
Quasi- action video is practiced, and the guidance of expert is lacked, be difficult to assess oneself movement whether standard.
The present invention has carried out the design of method and system aiming at the problem that acting scoring, crucial according to the human body in data set
Point position, movement to be scored can be extracted by the model by extracting model using deep neural network training human body key point
With the key point position of study course standard operation, the key point acted to two calculates similarity, to realize movement scoring.The party
Method and system are capable of the execution scoring of efficiently and accurately, do not depend on expert, it is only necessary to shoot picture or video by camera.
Summary of the invention
The technical problem to be solved by the present invention is to overcome the deficiencies of the prior art and provide a kind of human body based on camera
Act the method and system of scoring.
In order to solve the above technical problems, the present invention provides a kind of method of human action scoring based on camera, it is special
Sign is, includes the following steps:
Step 1: prepare human action image data collection, and acquisition mark letter is labeled to human region and key point
Breath, the human region and key point refer to the specified region on human body and key point in the area；
Step 2: the markup information of the human region and key point concentrated according to human action image data utilizes depth
Neural metwork training deep learning model extraction key point location information；Model is extracted for human body key point location information, it is defeated
Enter for a picture Image, exports as { KeyPointki, one deep learning model of training, so that F (Image)=
{KeyPointki}。
Step 3: by camera capture include movement to be scored picture, for movement to be scored picture with
The picture of standard operation calls the human body key point location information in two pictures of deep learning model extraction respectively；
Step 4: the human body key point location information of two movements of comparison calculates similarity, carries out movement scoring.
In the step 1, the human action image data, which is concentrated, includes at least 100,000 movement pictures, marks every
Human region and the location information of key point in picture, the number and title corresponding relationship in location information are as follows: 1- left eye, 2-
Right eye, 3- mouth, 4- neck, the left shoulder of 5-, the left elbow of 6-, 7- left finesse, 8- left hand, the right shoulder of 9-, the right elbow of 10-, 11- right finesse, 12-
The left buttocks of the right hand, 13-, 14- left knee, 15- left ankle, 16- left foot point, 17- right hips, the right knee of 18-, 19- right ankle,
20- right crus of diaphragm point, the markup information that final every figure obtains are { Bodyk,KeyPointki, wherein k indicates k-th of people in figure,
BodykIndicate that the zone position information of k-th of people, i indicate i-th of key point, KeyPointkiIndicate k-th of people, i-th of key
The position coordinates of point.
In the step 2, training deep learning model further includes following steps:
Step 1: using deep neural network one human region detection model of training, so that F1(Image)={ Bodyk,
For detecting human region position in picture, wherein human region is rectangle, and the location information of rectangular area is the rectangle upper left corner
With the apex coordinate in the lower right corner.
Step 2: model is extracted using the key point of deep neural network one human region of training, so that F2(Bodyk)=
{KeyPointki, for extracting the key point position in human region picture.
Step 3: above-mentioned two models coupling is got up, obtain human body key point location information extract model F (Image)=
F2(F1(Image))={ KeyPointki, for extracting the human body key point location information in picture；
In the step 3, movement picture to be scored is Image1, corresponding key point location information is F (Image1)
={ KeyPoint1i, standard operation picture is Image2, corresponding key point location information is F (Image2)=
{KeyPoint2i}。
In the step 4, the final scoring of movement are as follows:
S(Image1,Image2)=Similarity ({ KeyPoint1i},{KeyPoint2i})
Wherein Similarity is the similarity calculated between two movement key points, and the method used is chosen for calculating
The average value of cosine similarity between the vector that two neighboring key point is formed.
In the step 4, the vector Vector of 18 descriptions movement is chosent, wherein t value is 1-18, and number corresponds to
Vector be respectively as follows: 1- left eye -> right eye, 2- mouth -> left eye, 3- mouth -> neck, 4- neck -> left shoulder, the left shoulder -> left side 5-
Elbow, the left elbow -> left finesse of 6-, 7- left finesse -> left hand, 8- neck -> right shoulder, the right shoulder of 9- -> right elbow, the right elbow -> right finesse of 10-,
11- right finesse -> the right hand, 12- neck -> left buttocks, the left buttocks -> left knee of 12-, 13- left knee -> left ankle, 14- left foot
Ankle -> left foot point, 15- neck -> right hips, 16- right hips -> right knee, the right knee -> right ankle of 17-, the 18- right ankle -> right side
Tiptoe, then the movement picture finally scores are as follows:
In the step 3, when the movement wait score derives from action video, each frame is dynamic in interception action video
Make picture, and every movement picture is scored to obtain appraisal result, average value is calculated to all picture appraisal results, specifically
Are as follows:
Video to be scored is Video1, { Image is combined by the pictures that frame intercepts1s, call human body key point to extract
The set of keypoints that model obtains is { KeyPoint1si, obtained vector set is combined into { Vector1st, corresponding standard view
Frequency is Video2, pictures are combined into { Image2s, set of keypoints is { KeyPoint2si, vector set is combined into { Vector2st,
Wherein s indicates video frame number serial number, and i indicates key point number, and t indicates vector numbers, then the final scoring of the video are as follows:
A kind of system of the human action marking based on camera, it is characterised in that: include: the data source mould being sequentially connected
Block, model training module and movement scoring modules.
The data source modules are specifically included that for training human body key point to extract the data set of preparation required for model
The markup information of human action image data collection, human region and key point.
The model training module, for utilizing data set and deep neural network training deep learning model.
The movement scoring modules receive the figure of movement to be given a mark and standard operation for the interface of offer movement marking
Piece finally returns that marking as a result, the submodule being sequentially connected including four: movement to be given a mark and standard by calling model
It acts picture receiving submodule, human body key point location information extracting sub-module, human body key point confidence and ceases similarity calculation
Submodule, movement marking result return to submodule；The movement to be given a mark and standard operation picture receiving submodule, for receiving
The pictorial information of movement to be given a mark and standard operation；The human body key point location information extracting sub-module, for be scored
Movement and standard operation call the location information of the human body key point of deep learning model extraction two movements respectively；The people
Body key point confidence ceases similarity computational submodule, and the key point location information acted to two calculates similarity, moved
It gives a mark；The movement marking result returns to submodule, and the result finally given a mark is returned.
Advantageous effects of the invention: utilizing deep neural network according to the human body key point position in data set
Training human body key point, which extracts model, can extract the key point position of movement and standard operation to be given a mark by the model,
Then similarity is calculated by the key point acted to two, to realize movement marking.This method and system can be efficiently quasi-
True execution marking, does not depend on expert, it is only necessary to shoot picture or video by camera.
Detailed description of the invention
Fig. 1 is the flow diagram of the human action scoring method based on camera in exemplary embodiment of the present invention；
Fig. 2 is the structural schematic diagram of the human action scoring system based on camera in exemplary embodiment of the present invention.
Specific embodiment
The present invention is further illustrated with exemplary embodiment with reference to the accompanying drawing:
As shown in Figure 1, a kind of method that the present invention discloses human action marking based on camera, comprising:
Step 11: preparing human action image data collection, and acquisition mark letter is labeled to human region and key point
Breath.The present embodiment prepares data set by taking dance movement as an example as follows.
Step 111: collecting 100 dance movement videos, extract picture by frame, obtain 100,000 dancings by artificial screening
Picture is acted, covers different movements as far as possible.
Step 112: movement picture manually being marked, the position of human region and key point in every picture is marked and believes
It ceases, the number and title corresponding relationship in location information are as follows: 1- left eye, 2- right eye, 3- mouth, 4- neck, the left shoulder of 5-, 6- are left
Elbow, 7- left finesse, 8- left hand, the right shoulder of 9-, the right elbow of 10-, 11- right finesse, the 12- right hand, the left buttocks of 13-, 14- left knee, 15- are left
Ankle, 16- left foot point, 17- right hips, the right knee of 18-, 19- right ankle, 20- right crus of diaphragm point.It can be obtained by more key points
More detailed motion characteristic, help more accurately to carry out motion characteristic scoring, here key point can according to concern to
Scoring movement position is flexibly chosen, and method and system disclosed by the invention are applicable.The mark that final every figure obtains
Information is { Bodyk,KeyPointki, wherein k indicates k-th of people in figure, BodykIndicate the rectangular area position letter of k-th of people
The breath coordinate of bottom right angular vertex (the rectangle upper left corner), i indicate i-th of key point, KeyPointkiIndicate i-th of k-th of people pass
The position coordinates of key point.
Step 12: according to human region and the markup information of key point, utilizing deep neural network training human body key point
Location information extracts model.Model is extracted for human body key point location information, is inputted as a picture Image, exports and is
{KeyPointki, in accordance with the following steps, one deep learning model of training, so that F (Image)={ KeyPointki}。
Step 121: using deep neural network one human region detection model of training, so that F1(Image)=
{Bodyk, for detecting human region position in picture.
Step 122: model is extracted using the key point of deep neural network one human region of training, so that F2(Bodyk)
={ KeyPointki, for extracting the key point position in human region picture.
Step 123: above-mentioned two models coupling being got up, human body key point location information is obtained and extracts model F (Image)
=F2(F1(Image))={ KeyPointki, eventually for the human body key point location information extracted in picture.
This method be capable of it is more fast and accurate positioning human body key point location information, facilitate it is subsequent accurately
Carry out movement scoring.
Step 13: for movement picture and standard operation picture to be scored, calling human body key point to extract model respectively
Extract the human body key point of two movements.Movement picture to be scored is Image1, corresponding key point is F (Image1)=
{KeyPoint1i, study course standard operation picture is Image2, corresponding key point is F (Image2)={ KeyPoint2i}。
Step 14: the key point of two movements of comparison calculates similarity, carries out movement marking.It finally gives a mark and is
S(Image1,Image2)=Similarity ({ KeyPoint1i},{KeyPoint2i})
Wherein Similarity is the similarity calculated between two movement key points, and the method used is chosen for calculating
The average value of cosine similarity between the vector that two neighboring key point is formed.The present embodiment is chosen such as by taking dance movement as an example
The vector Vector of lower 18 descriptions dance movementt, wherein t value is 1-18, numbers corresponding vector and is respectively as follows: 1- left eye-
> right eye, 2- mouth -> left eye, 3- mouth -> neck, 4- neck -> left shoulder, the left shoulder of 5- -> left elbow, the left elbow -> left finesse of 6-, 7- are left
Wrist -> left hand, 8- neck -> right shoulder, the right shoulder of 9- -> right elbow, the right elbow -> right finesse of 10-, the 11- right finesse -> right hand, 12- neck-
> left buttocks, the left buttocks -> left knee of 12-, 13- left knee -> left ankle, 14- left ankle -> left foot point, 15- neck -> right stern
Portion, 16- right hips -> right knee, the right knee -> right ankle of 17-, 18- right ankle -> right crus of diaphragm is sharp, then the movement picture final score
For
This method can both carry out whole marking to movement, can also prompt the lower office of score by the similarity of single vector
Portion acts (such as head, shoulder, hand, leg, foot), and user can be helped targetedly to improve, and score is relatively low to be moved
Make.Here vector can flexibly be chosen according to the key point information that the transfer to be evaluated of concern makees position and mark, and the present invention is public
The method and system opened are applicable.
When the movement wait score derives from action video, the movement picture of each frame in action video is intercepted, and to every
It opens movement picture to be scored to obtain appraisal result, average value is calculated to all picture appraisal results.Assuming that video to be scored
For Video1, { Image is combined by the pictures that frame intercepts1s, call human body key point to extract the set of keypoints that model obtains
For { KeyPoint1si, obtained vector set is combined into { Vector1st, corresponding study course normal video is Video2, pictures
It is combined into { Image2s, set of keypoints is { KeyPoint2si, vector set is combined into { Vector2st, wherein s indicates video frame number
Serial number, i indicate key point number, and t indicates vector numbers, then the final score of the video are as follows:
It is scored with this standard of action video, scoring accuracy obtained is high and evaluation speed is fast, a frame
Interval can be adapted for most of scene, can also be with flexible adaptation special action by adjusting the frame number interval of interception
The case where rate or video length, has wide applicability.
As shown in Fig. 2, the system that the present invention discloses a kind of human action marking based on camera, specifically includes that successively
Connected data source modules 21, model training module 22 and movement scoring modules 23.
The data source modules 21, it is main to wrap for training human body key point to extract the data set of preparation required for model
It includes: the markup information of human action image data collection, human region and key point.
The model training module 22, for utilizing data set and deep neural network training pedestrian's body key point confidence
Breath extracts model.
The movement scoring modules 23 receive movement to be given a mark and standard operation for the interface of offer movement marking
Picture finally returns that marking as a result, the submodule being sequentially connected including four: movement to be given a mark and mark by calling model
Quasi- movement picture receiving submodule 231, human body key point location information extracting sub-module 232, human body key point confidence manner of breathing
Submodule 234 is returned like degree computational submodule 233, movement marking result；
The movement to be given a mark and standard operation picture receiving submodule 231, it is dynamic for receiving movement to be given a mark and standard
The pictorial information of work；
The human body key point location information extracting sub-module 232, for movement to be scored and standard operation, is adjusted respectively
The human body key point location information of information extraction model extraction two movements is set with human body key point；
The human body key point confidence ceases similarity computational submodule 233, the key point location information acted to two,
Similarity is calculated, movement marking is carried out；
The movement marking result returns to submodule 234, and the result finally given a mark is returned.
Present invention is mainly used for a kind of method and system of human action marking based on camera are provided, according to data set
In human body key point position, extracting model using deep neural network training human body key point can be extracted by the model
The key point position of movement and standard operation to be given a mark, the key point acted to two calculates similarity, to realize movement
Marking.This method and system are capable of the execution marking of efficiently and accurately, do not depend on expert, it is only necessary to pass through camera shooting figure
Piece or video.
Above embodiments do not limit the present invention in any way, all to be made in a manner of equivalent transformation to above embodiments
Other improvement and application, belong to protection scope of the present invention.
Claims (7)
1. a kind of method of the human action scoring based on camera, which comprises the steps of:
Step 1: prepare human action image data collection, and acquisition markup information, institute are labeled to human region and key point
It states human region and key point refers to the specified region on human body and key point in the area；
Step 2: the markup information of the human region and key point concentrated according to human action image data utilizes depth nerve
Network training human body key point location information extracts model extraction key point location information；Human body key point location information is mentioned
Modulus type is inputted as a picture Image, is exported as { KeyPointki, one deep learning model of training, so that F
(Image)={ KeyPointki}；
Step 3: the picture comprising movement to be scored is captured by camera, for the picture and standard of movement to be scored
The picture of movement calls human body key point location information to extract the human body key point confidence in two pictures of model extraction respectively
Breath；
Step 4: the human body key point location information of two movements of comparison calculates similarity, carries out movement scoring.
2. a kind of method of human action scoring based on camera as described in claim 1, it is characterised in that: the step
In one, the human action image data, which is concentrated, includes at least 100,000 movement pictures, mark in every picture human region and
The markup information of key point, the number and title corresponding relationship in markup information are as follows: 1- left eye, 2- right eye, 3- mouth, 4- neck
Son, the left shoulder of 5-, the left elbow of 6-, 7- left finesse, 8- left hand, the right shoulder of 9-, the right elbow of 10-, 11- right finesse, the 12- right hand, the left buttocks of 13-,
14- left knee, 15- left ankle, 16- left foot point, 17- right hips, the right knee of 18-, 19- right ankle, 20- right crus of diaphragm point, it is final every
Opening the markup information that figure obtains is { Bodyk,KeyPointki}；Wherein k indicates k-th of people in figure, BodykIndicate k-th of people's
Zone position information, i indicate i-th of key point, KeyPointkiIndicate the position coordinates of k-th of people, i-th of key point.
3. a kind of method of human action scoring based on camera as claimed in claim 2, it is characterised in that: the step
In two, training deep learning model further includes following steps:
Step 1: using deep neural network one human region detection model of training, so that F1(Image)={ Bodyk, it is used for
Human region position in picture is detected, wherein human region is rectangle, and the location information of rectangular area is the rectangle upper left corner and the right side
The apex coordinate of inferior horn；
Step 2: model is extracted using the key point of deep neural network one human region of training, so that F2(Bodyk)=
{KeyPointki, for extracting the key point position in human region picture；
Step 3: above-mentioned two models coupling being got up, human body key point location information is obtained and extracts model F (Image)=F2(F1
(Image))={ KeyPointki, for extracting the human body key point location information in picture；
In the step 3, movement picture to be scored is Image1, corresponding key point location information is F (Image1)=
{KeyPoint1i, standard operation picture is Image2, corresponding key point location information is F (Image2)=
{KeyPoint2i}。
4. a kind of method of human action scoring based on camera as claimed in claim 3, it is characterised in that: the step
In four, the final scoring of movement are as follows:
S(Image1,Image2)=Similarity ({ KeyPoint1i},{KeyPoint2i})
Wherein Similarity is the similarity calculated between two movement key points, and the method used is chosen adjacent for calculating
The average value of cosine similarity between the vector that two key points are formed.
5. a kind of method of human action scoring based on camera as claimed in claim 4, it is characterised in that: the step
In four, the vector Vector of 18 descriptions movement is chosent, wherein t value is 1-18, numbers corresponding vector and is respectively as follows: the left side 1-
Eye -> right eye, 2- mouth -> left eye, 3- mouth -> neck, 4- neck -> left shoulder, the left shoulder of 5- -> left elbow, the left elbow -> left finesse of 6-,
7- left finesse -> left hand, 8- neck -> right shoulder, the right shoulder of 9- -> right elbow, the right elbow -> right finesse of 10-, the 11- right finesse -> right hand, 12-
Neck -> left buttocks, the left buttocks -> left knee of 12-, 13- left knee -> left ankle, 14- left ankle -> left foot point, 15- neck ->
Right hips, 16- right hips -> right knee, the right knee -> right ankle of 17-, 18- right ankle -> right crus of diaphragm point, then the movement picture is final
Scoring are as follows:
6. a kind of method of human action scoring based on camera as claimed in claim 5, which is characterized in that the step
In three, when the movement wait score derives from action video, the movement picture of each frame in action video is intercepted, and dynamic to every
It is scored to obtain appraisal result as picture, average value is calculated to all picture appraisal results, specifically:
Video to be scored is Video1, { Image is combined by the pictures that frame intercepts1s, call human body key point to extract model
Obtained set of keypoints is { KeyPoint1si, obtained vector set is combined into { Vector1st, corresponding normal video is
Video2, pictures are combined into { Image2s, set of keypoints is { KeyPoint2si, vector set is combined into { Vector2st, wherein s
Indicate video frame number serial number, i indicates key point number, and t indicates vector numbers, then the final scoring of the video are as follows:
7. a kind of system of the human action marking based on camera of operation of one of -6 the methods according to claim 1,
It is characterized in that: including: the data source modules being sequentially connected, model training module and movement scoring modules.
The data source modules specifically include that human body for training human body key point to extract the data set of preparation required for model
Act the markup information of image data collection, human region and key point.
The model training module, for being extracted using data set and deep neural network training pedestrian's body key point location information
Model.
The movement scoring modules receive the picture of movement to be given a mark and standard operation for the interface of offer movement marking, lead to
Crossing calling model finally returns that marking as a result, the submodule being sequentially connected including four: movement to be given a mark and standard operation
Picture receiving submodule, human body key point location information extracting sub-module, human body key point confidence cease similarity calculation submodule
Block, movement marking result return to submodule；The movement to be given a mark and standard operation picture receiving submodule, for receiving wait beat
Transfer makees and the pictorial information of standard operation；The human body key point location information extracting sub-module, for movement to be scored
With standard operation, human body key point location information is called to extract the position letter of the human body key point of model extraction two movements respectively
Breath；The human body key point confidence ceases similarity computational submodule, and the key point location information acted to two calculates similar
Degree, carries out movement marking；The movement marking result returns to submodule, and the result finally given a mark is returned.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910132276.8A CN109829442A (en) | 2019-02-22 | 2019-02-22 | A kind of method and system of the human action scoring based on camera |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910132276.8A CN109829442A (en) | 2019-02-22 | 2019-02-22 | A kind of method and system of the human action scoring based on camera |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109829442A true CN109829442A (en) | 2019-05-31 |
Family
ID=66864122
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910132276.8A Pending CN109829442A (en) | 2019-02-22 | 2019-02-22 | A kind of method and system of the human action scoring based on camera |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109829442A (en) |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110222977A (en) * | 2019-06-03 | 2019-09-10 | 张学志 | One kind movement sport methods of marking based on computer vision and device |
| CN110781857A (en) * | 2019-11-05 | 2020-02-11 | 北京沃东天骏信息技术有限公司 | Motion monitoring method, device, system and storage medium |
| CN111259822A (en) * | 2020-01-19 | 2020-06-09 | 杭州微洱网络科技有限公司 | Method for detecting key point of special neck in E-commerce image |
| CN111314665A (en) * | 2020-03-07 | 2020-06-19 | 上海中科教育装备集团有限公司 | Key video segment extraction system and method for video post-scoring |
| CN111767768A (en) * | 2019-07-31 | 2020-10-13 | 北京京东尚科信息技术有限公司 | Image processing method, device and equipment |
| CN111797778A (en) * | 2020-07-08 | 2020-10-20 | 龙岩学院 | Automatic scoring method for breaking street dance anchor and wheat dance |
| CN111986260A (en) * | 2020-09-04 | 2020-11-24 | 北京小狗智能机器人技术有限公司 | Image processing method and device and terminal equipment |
| CN112348942A (en) * | 2020-09-18 | 2021-02-09 | 当趣网络科技(杭州)有限公司 | Body-building interaction method and system |
| CN112381035A (en) * | 2020-11-25 | 2021-02-19 | 山东云缦智能科技有限公司 | Motion similarity evaluation method based on motion trail of skeleton key points |
| CN112990011A (en) * | 2021-03-15 | 2021-06-18 | 上海工程技术大学 | Body-building action recognition and evaluation method based on machine vision and deep learning |
| CN113095248A (en) * | 2021-04-19 | 2021-07-09 | 中国石油大学(华东) | Technical action correction method for badminton |
| CN113137923A (en) * | 2020-01-17 | 2021-07-20 | 上海淡竹体育科技有限公司 | Standing long jump sport result measuring method |
| CN114783046A (en) * | 2022-03-01 | 2022-07-22 | 北京赛思信安技术股份有限公司 | CNN and LSTM-based human body continuous motion similarity scoring method |
| CN116392800A (en) * | 2023-04-23 | 2023-07-07 | 电子科技大学 | Based on target detection and image processing standing long jump distance measuring method and system |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107219925A (en) * | 2017-05-27 | 2017-09-29 | 成都通甲优博科技有限责任公司 | Pose detection method, device and server |
| CN107392086A (en) * | 2017-05-26 | 2017-11-24 | 深圳奥比中光科技有限公司 | Apparatus for evaluating, system and the storage device of human body attitude |
| CN108615055A (en) * | 2018-04-19 | 2018-10-02 | 咪咕动漫有限公司 | A kind of similarity calculating method, device and computer readable storage medium |
| CN109191588A (en) * | 2018-08-27 | 2019-01-11 | 百度在线网络技术（北京）有限公司 | Move teaching method, device, storage medium and electronic equipment |
-
2019
- 2019-02-22 CN CN201910132276.8A patent/CN109829442A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107392086A (en) * | 2017-05-26 | 2017-11-24 | 深圳奥比中光科技有限公司 | Apparatus for evaluating, system and the storage device of human body attitude |
| CN107219925A (en) * | 2017-05-27 | 2017-09-29 | 成都通甲优博科技有限责任公司 | Pose detection method, device and server |
| CN108615055A (en) * | 2018-04-19 | 2018-10-02 | 咪咕动漫有限公司 | A kind of similarity calculating method, device and computer readable storage medium |
| CN109191588A (en) * | 2018-08-27 | 2019-01-11 | 百度在线网络技术（北京）有限公司 | Move teaching method, device, storage medium and electronic equipment |
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110222977A (en) * | 2019-06-03 | 2019-09-10 | 张学志 | One kind movement sport methods of marking based on computer vision and device |
| CN111767768A (en) * | 2019-07-31 | 2020-10-13 | 北京京东尚科信息技术有限公司 | Image processing method, device and equipment |
| CN110781857B (en) * | 2019-11-05 | 2022-09-06 | 北京沃东天骏信息技术有限公司 | Motion monitoring method, device, system and storage medium |
| CN110781857A (en) * | 2019-11-05 | 2020-02-11 | 北京沃东天骏信息技术有限公司 | Motion monitoring method, device, system and storage medium |
| CN113137923A (en) * | 2020-01-17 | 2021-07-20 | 上海淡竹体育科技有限公司 | Standing long jump sport result measuring method |
| CN111259822A (en) * | 2020-01-19 | 2020-06-09 | 杭州微洱网络科技有限公司 | Method for detecting key point of special neck in E-commerce image |
| CN111314665A (en) * | 2020-03-07 | 2020-06-19 | 上海中科教育装备集团有限公司 | Key video segment extraction system and method for video post-scoring |
| CN111797778A (en) * | 2020-07-08 | 2020-10-20 | 龙岩学院 | Automatic scoring method for breaking street dance anchor and wheat dance |
| CN111797778B (en) * | 2020-07-08 | 2023-06-02 | 龙岩学院 | Automatic scoring method for break-in street dance and wheat-linking dancing |
| CN111986260A (en) * | 2020-09-04 | 2020-11-24 | 北京小狗智能机器人技术有限公司 | Image processing method and device and terminal equipment |
| CN112348942A (en) * | 2020-09-18 | 2021-02-09 | 当趣网络科技(杭州)有限公司 | Body-building interaction method and system |
| CN112348942B (en) * | 2020-09-18 | 2024-03-19 | 当趣网络科技(杭州)有限公司 | Body-building interaction method and system |
| CN112381035A (en) * | 2020-11-25 | 2021-02-19 | 山东云缦智能科技有限公司 | Motion similarity evaluation method based on motion trail of skeleton key points |
| CN112990011A (en) * | 2021-03-15 | 2021-06-18 | 上海工程技术大学 | Body-building action recognition and evaluation method based on machine vision and deep learning |
| CN113095248B (en) * | 2021-04-19 | 2022-10-25 | 中国石油大学(华东) | Technical action correcting method for badminton |
| CN113095248A (en) * | 2021-04-19 | 2021-07-09 | 中国石油大学(华东) | Technical action correction method for badminton |
| CN114783046A (en) * | 2022-03-01 | 2022-07-22 | 北京赛思信安技术股份有限公司 | CNN and LSTM-based human body continuous motion similarity scoring method |
| CN116392800A (en) * | 2023-04-23 | 2023-07-07 | 电子科技大学 | Based on target detection and image processing standing long jump distance measuring method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109829442A (en) | A kind of method and system of the human action scoring based on camera | |
| CN104517102B (en) | Student classroom notice detection method and system | |
| CN106022213B (en) | A kind of human motion recognition method based on three-dimensional bone information | |
| WO2018120964A1 (en) | Posture correction method based on depth information and skeleton information | |
| WO2021164283A1 (en) | Clothing color recognition method, device and system based on semantic segmentation | |
| JP6448223B2 (en) | Image recognition system, image recognition apparatus, image recognition method, and computer program | |
| CN109299659A (en) | A kind of human posture recognition method and system based on RGB camera and deep learning | |
| CN109919141A (en) | A kind of recognition methods again of the pedestrian based on skeleton pose | |
| CN106980852B (en) | Based on Corner Detection and the medicine identifying system matched and its recognition methods | |
| US11945125B2 (en) | Auxiliary photographing device for dyskinesia analysis, and control method and apparatus for auxiliary photographing device for dyskinesia analysis | |
| CN110008913A (en) | Pedestrian re-identification method based on fusion of attitude estimation and viewpoint mechanism | |
| CN102800126A (en) | Method for recovering real-time three-dimensional body posture based on multimodal fusion | |
| CN104794449B (en) | Gait energy diagram based on human body HOG features obtains and personal identification method | |
| CN111027432A (en) | Gait feature-based visual following robot method | |
| CN102567703A (en) | Hand motion identification information processing method based on classification characteristic | |
| CN108921881A (en) | A kind of across camera method for tracking target based on homography constraint | |
| CN109766796A (en) | A kind of depth pedestrian detection method towards dense population | |
| CN110472473A (en) | The method fallen based on people on Attitude estimation detection staircase | |
| CN110084192A (en) | Quick dynamic hand gesture recognition system and method based on target detection | |
| CN113537019B (en) | Detection method for identifying wearing of safety helmet of transformer substation personnel based on key points | |
| CN109840478A (en) | A kind of movement appraisal procedure, device, mobile terminal and readable storage medium storing program for executing | |
| KR102377767B1 (en) | Handwriting and arm movement learning-based sign language translation system and method | |
| CN106909890A (en) | A kind of Human bodys' response method based on position cluster feature | |
| Yang et al. | Human exercise posture analysis based on pose estimation | |
| CN109766782A (en) | Real-time body action identification method based on SVM |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190531 |