CN108564596A - A kind of the intelligence comparison analysis system and method for golf video - Google Patents
A kind of the intelligence comparison analysis system and method for golf video Download PDFInfo
- Publication number
- CN108564596A CN108564596A CN201810170508.4A CN201810170508A CN108564596A CN 108564596 A CN108564596 A CN 108564596A CN 201810170508 A CN201810170508 A CN 201810170508A CN 108564596 A CN108564596 A CN 108564596A
- Authority
- CN
- China
- Prior art keywords
- sportsman
- golf
- frame
- sequence
- key point
- 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.)
- Granted
Links
- 238000004458 analytical method Methods 0.000 title claims abstract description 100
- 238000000034 method Methods 0.000 title claims description 16
- 238000001514 detection method Methods 0.000 claims abstract description 83
- 238000009877 rendering Methods 0.000 claims abstract description 24
- 230000003068 static effect Effects 0.000 claims abstract description 23
- 230000033001 locomotion Effects 0.000 claims description 47
- 238000000605 extraction Methods 0.000 claims description 33
- 230000009471 action Effects 0.000 claims description 28
- 230000008859 change Effects 0.000 claims description 17
- 238000012549 training Methods 0.000 claims description 14
- 230000007547 defect Effects 0.000 claims description 13
- 210000000245 forearm Anatomy 0.000 claims description 13
- 230000003542 behavioural effect Effects 0.000 claims description 12
- 239000000284 extract Substances 0.000 claims description 12
- 238000012216 screening Methods 0.000 claims description 5
- 210000000707 wrist Anatomy 0.000 claims description 5
- 238000005457 optimization Methods 0.000 claims description 3
- 238000002203 pretreatment Methods 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 2
- 238000005191 phase separation Methods 0.000 claims description 2
- 230000033764 rhythmic process Effects 0.000 claims description 2
- 101100366707 Arabidopsis thaliana SSL11 gene Proteins 0.000 claims 1
- 101100366711 Arabidopsis thaliana SSL13 gene Proteins 0.000 claims 1
- 101100366561 Panax ginseng SS11 gene Proteins 0.000 claims 1
- 101100366562 Panax ginseng SS12 gene Proteins 0.000 claims 1
- 230000001483 mobilizing effect Effects 0.000 claims 1
- 238000000926 separation method Methods 0.000 claims 1
- 238000010586 diagram Methods 0.000 description 10
- 238000005516 engineering process Methods 0.000 description 6
- 230000007257 malfunction Effects 0.000 description 4
- 230000000386 athletic effect Effects 0.000 description 3
- 210000000988 bone and bone Anatomy 0.000 description 3
- 210000002414 leg Anatomy 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 210000003423 ankle Anatomy 0.000 description 2
- 230000007812 deficiency Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 210000003128 head Anatomy 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 230000005764 inhibitory process Effects 0.000 description 2
- 210000003127 knee Anatomy 0.000 description 2
- 239000003550 marker Substances 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000012552 review Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 240000006409 Acacia auriculiformis Species 0.000 description 1
- 230000037147 athletic performance Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000000903 blocking effect Effects 0.000 description 1
- 210000005069 ears Anatomy 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 210000000887 face Anatomy 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
- G06T7/248—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments involving reference images or patches
-
- A—HUMAN NECESSITIES
- A63—SPORTS; GAMES; AMUSEMENTS
- A63B—APPARATUS FOR PHYSICAL TRAINING, GYMNASTICS, SWIMMING, CLIMBING, OR FENCING; BALL GAMES; TRAINING EQUIPMENT
- A63B71/00—Games or sports accessories not covered in groups A63B1/00 - A63B69/00
- A63B71/06—Indicating or scoring devices for games or players, or for other sports activities
- A63B71/0619—Displays, user interfaces and indicating devices, specially adapted for sport equipment, e.g. display mounted on treadmills
- A63B71/0622—Visual, audio or audio-visual systems for entertaining, instructing or motivating the user
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/41—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items
- G06V20/42—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items of sport video content
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/48—Matching video sequences
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N23/00—Cameras or camera modules comprising electronic image sensors; Control thereof
- H04N23/90—Arrangement of cameras or camera modules, e.g. multiple cameras in TV studios or sports stadiums
-
- A—HUMAN NECESSITIES
- A63—SPORTS; GAMES; AMUSEMENTS
- A63B—APPARATUS FOR PHYSICAL TRAINING, GYMNASTICS, SWIMMING, CLIMBING, OR FENCING; BALL GAMES; TRAINING EQUIPMENT
- A63B71/00—Games or sports accessories not covered in groups A63B1/00 - A63B69/00
- A63B71/06—Indicating or scoring devices for games or players, or for other sports activities
- A63B71/0619—Displays, user interfaces and indicating devices, specially adapted for sport equipment, e.g. display mounted on treadmills
- A63B71/0622—Visual, audio or audio-visual systems for entertaining, instructing or motivating the user
- A63B2071/0638—Displaying moving images of recorded environment, e.g. virtual environment
- A63B2071/0644—Displaying moving images of recorded environment, e.g. virtual environment with display speed of moving landscape controlled by the user's performance
-
- A—HUMAN NECESSITIES
- A63—SPORTS; GAMES; AMUSEMENTS
- A63B—APPARATUS FOR PHYSICAL TRAINING, GYMNASTICS, SWIMMING, CLIMBING, OR FENCING; BALL GAMES; TRAINING EQUIPMENT
- A63B71/00—Games or sports accessories not covered in groups A63B1/00 - A63B69/00
- A63B71/06—Indicating or scoring devices for games or players, or for other sports activities
- A63B71/0619—Displays, user interfaces and indicating devices, specially adapted for sport equipment, e.g. display mounted on treadmills
- A63B2071/0647—Visualisation of executed movements
-
- A—HUMAN NECESSITIES
- A63—SPORTS; GAMES; AMUSEMENTS
- A63B—APPARATUS FOR PHYSICAL TRAINING, GYMNASTICS, SWIMMING, CLIMBING, OR FENCING; BALL GAMES; TRAINING EQUIPMENT
- A63B71/00—Games or sports accessories not covered in groups A63B1/00 - A63B69/00
- A63B71/06—Indicating or scoring devices for games or players, or for other sports activities
- A63B71/0619—Displays, user interfaces and indicating devices, specially adapted for sport equipment, e.g. display mounted on treadmills
- A63B2071/0658—Position or arrangement of display
- A63B2071/0661—Position or arrangement of display arranged on the user
- A63B2071/0663—Position or arrangement of display arranged on the user worn on the wrist, e.g. wrist bands
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Signal Processing (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Human Computer Interaction (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Physical Education & Sports Medicine (AREA)
- Image Analysis (AREA)
Abstract
The present invention discloses a kind of intelligence comparison analysis system of golf video, including：Infrared detecting device, the Openpose critical point detection devices being connect with the infrared detecting device, the key point information sequence pretreatment unit being connect with the Openpose critical point detections device, the golf swing sequence detecting apparatus being connect with the key point information sequence pretreatment unit, the golf swing key frame detection device being connect with the golf swing sequence detecting apparatus and the key point information sequence pretreatment unit, the intelligence being connect with the golf swing key frame detection device and the key point information sequence pretreatment unit compares analytical equipment, the rendering device that analytical equipment and the infrared detecting device are connect is compared with the intelligence.The present invention also proposes that a kind of intelligence of golf video compares analysis method, and each feature is extracted by dynamic+static, front+side multi-angle comparison analysis.
Description
Technical field
The invention belongs to Human bodys' response technical fields, and in particular to the detection method of golf swing sequence,
The intelligence of golf swing Video Key frame detection method more particularly to a kind of golf video compares analysis system
And method.
Background technology
Traditional athletic training is all that one-to-one or one-to-many coach personally instructs training, not only receives time and location
Limitation, and training cost is high, and coach's resource is nervous.In order to solve this problem, more and more professional sport training forces or
Training forces of person country start motion capture application in athletic training.Motion capture system assists training expert from multiple visual angles
The athletic performance of trainer is observed, and obtains the multiple kinematic parameters and physical signs parameter of trainer, is scientific training
Reliable foundation is provided.
For golf sports, the performance when posture of sportsman and action play golf for sportsman has critically important
Influence.Sometimes a subtle attitudes vibration can all influence the direction played ball, dynamics etc..Many researchers attempt
Swing when being played golf come study movement person by the posture of human body, because naked eyes can not accurately be observed in a flash
The action of sportsman whether standard.
There are many technologies to be all applied in golf sports capture at present, mainly there is mechanical motion-captured, electromagnetic type
It is motion-captured, acoustics formula is motion-captured, optical motion captures and acquires the motion information data of student.Mechanical motion captures
Precision is higher, and real-time is good, can capture multiple objects simultaneously, but mechanical device is bigger to the action constraint for capturing object,
It is inconvenient to use.Electromagnetic type movement capturing technology is ripe, of low cost, it can be achieved that real-time capture, but can only capture low frequency movement,
To environmental requirement height, cannot around there are magnetic field, multiple capture that there can be interference.Acoustics formula is motion-captured to be can be very good to solve people
Body is cheap from occlusion issue, but there are larger delay and lag, precision is not high, and error is larger.Optical motion is caught
It catches, to Moving Objects, there is no limit low in hardware cost, later data processing is big, and algorithm is complicated, requires light high, precision
It is to be improved.
Currently, the golf analyzer of market is all the data for acquiring sportsman by sensor mostly, due to needing to wear
Wear profession equipment or need marker is placed with, come track human body each position movable information, to a certain degree
On limit the movement of sportsman, be widely applied so hardly resulting in.On the other hand, in computer vision research field, greatly
Part is all to shoot the action of sportsman by the depth camera as Kinect or clapped by common camera
Video is taken the photograph by the method for Attitude estimation come each bone key point information of predicted motion person.The movement of sportsman is obtained at present
Information just only has both technologies, but the movable information for only obtaining these sportsmen is also not enough to identify the action of sportsman
Whether mistake.Major part golf analyzer all can only analyze golf video by people at present, or review and regard
Frequency is analyzed, or by motion information data come analysis of key kinematic parameter to provide corresponding guidance, but cannot be as special
Industry trains the place for equally directly pointing out critical error.
Invention content
For golf motion analysis, the identification granularity of traditional action recognition is thicker, can only distinguish race, walk,
The obvious action such as embrace, and golf motion analysis needs to identify whether posture and action satisfy the criteria, and identifies
Granularity is very fine, is to be unable to reach effect by traditional action recognition.Golf analyzer at present on the market is most of
It is the variation of the body parameter in the motion process by sportsman's carry sensors to record sportsman, but these parameters can only
The mistake for reflecting some actions of sportsman on certain angle, cannot accurately position the place of stroke defect at the beginning, as
The place that specialized coaches equally point out faults in time.And carry sensors can bring big inconvenience to sportsman, be unfavorable for
Sportsman plays.
In order to solve the above technical problems, technological deficiency of the existing technology, the present invention is overcome to adopt the following technical scheme that：
A kind of intelligence comparison analysis system of golf video, including：
The infrared detecting device being connect with front, side camera, the swinging for sportsman for exporting front, side regard
Frequently；
The Openpose critical point detection devices being connect with infrared detecting device, the fortune for front, side to input
The front of swing video and the preprepared complete coach that mobilize swing the key of human body in video detection video
Point information, the key point information of disengaging movement person's target person and non athlete's target person, the front of output sportsman's target,
The key point information sequence of side and the positive key point information sequence of coach；
The key point information sequence pretreatment unit being connect with Openpose critical point detection devices, for the fortune to input
The positive key point information sequence in the front of target, the key point information sequence of side and coach is mobilized to carry out missing values
Filling, the removal of exceptional value filling, the key point information sequence in the front of sportsman's target, side after output pretreatment with
And the positive key point information sequence of coach；
The golf swing sequence detecting apparatus being connect with key point information sequence pretreatment unit, for input
Sportsman front key point information sequence and coach front key point information sequence between carry out similarity mode, extraction fortune
Complete golf swing sequence in positive key point information sequence is mobilized, by golf swing sequence and Fei Gaoer
Husband swings action sequence phase separation, complete golf swing sequence in the key point information sequence of output sportsman front
Sart point in time and end time point；
The golf being connect with golf swing sequence detecting apparatus and key point information sequence pretreatment unit is waved
Bar acts key frame detection device, for the sportsman front key point information sequence and the complete golf of sportsman to input
It swings at the beginning of action sequence and the end time carries out golf swing key frame detection, golf swing is closed
Key frame is divided into 8 frames, is respectively：Erect-position frame draws bar frame, upper boom frame, vertex frame, discharges frame, frame of batting, send bar frame, frame of putting away sports equipment.
The time point of 8 key operations frames of this section of sportsman front key point information sequence is exported respectively；
The intelligence being connect with golf swing key operations frame detection device and key point information sequence pretreatment unit
Analytical equipment can be compared, for 8 key operations frames to the sportsman front key point information sequence of input time point with
And the key point information sequence progress intelligence comparison analysis in the front, side of sportsman, using golf professional knowledge and combine
The comparison analysis rule that actual conditions are formulated compares analysis to carry out intelligence, and output compares the analysis result of analysis；
The rendering device that analytical equipment and infrared detecting device are connect is compared with intelligence, for the intelligence to input than team point
The video that swings in the front, side of analysing result and sportsman renders, and will compare analysis result and is rendered into the front of movement, side
In the video that swings in face, the video that swings in sportsman front, side after output rendering.
As a kind of preferred embodiment, infrared detecting device includes：
Infrared detection unit：It detects whether golf is knocked off by infrared ray, if hitting, output signal and hits
Ball time point；
The video clipping unit being connect with infrared detection unit：Once receiving the signal from infrared detection unit and batting
Time point exports 6 seconds long fronts, flank movement persons and waves respectively immediately according to 3 seconds before and after hitting time point come editing video
Bar video.
As a kind of preferred embodiment, the Openpose critical point detection devices include：
Key point extraction unit：It is swung video and pre- to front, the side of sportsman using Openpose algorithm models
First the front of ready coach swings video extraction key point information；
The target person key point information extraction unit being connect with key point extraction unit：Most using intermediate region and area
Big criterion removes the key point information of non athlete's target person, extracts the key point information of sportsman's target person.
As a kind of preferred embodiment, the key point information sequence pretreatment unit includes：
Missing Data Filling unit：Key point letter for the target person for filling the output of Openpose critical point detection devices
Missing values in breath, although the precision of Openpose algorithm models is very high, there is also the feelings that some key points are not previously predicted out
Condition, it is due to the blocking etc. of target person itself to cause such case.In order to fill the missing values in key point information, I
Using human motion continuity, the missing values in key point information are filled using the method for linear interpolation；
The exceptional value removal fills unit being connect with Missing Data Filling unit：Although the precision of Openpose algorithm models
The case where very high, but there are prediction errors, to reduce the influence that these error-critical point information work to later analysis, I
Fluctuate very big key point information by the way of sliding window to traverse the removal of key point information sequence, and using linear
The method of interpolation refills.
As a kind of preferred embodiment, the golf swing sequence detecting apparatus includes：
Key point information peak detection unit：Peak point for the ordinate information sequence for extracting key point, i.e., it is crucial
The peak point of the height change curve of point.Purpose is the big position approximate of the golf swing sequence in order to detect sportsman
It sets；
The golf swing sequence context detection unit being connect with key point information peak detection unit：For passing through
The peak point of key point information peak detection unit output detects the range of golf swing sequence.
As a kind of preferred embodiment, the golf swing key frame detection device,
Including：Feature extraction unit：Movement locus feature for extracting sportsman；
The key frame detection unit being connect with feature extraction unit：For extracting 8 from the movement locus feature of sportsman
A golf key operations frame.
As a kind of preferred embodiment, the intelligence compares analytical equipment, including：
Dynamic analysis unit：Front, the side behavioral characteristics of sportsman are extracted, and are carried out according to the behavioral characteristics extracted
Compare analysis；
Static analysis unit：Front, the side static nature of sportsman are extracted, and is carried out according to the static nature extracted
Compare analysis.
As a kind of preferred embodiment, the rendering device, including：
Key frame picture rendering unit：For rendering the comparison analysis result of key frame in sportsman front, side 8
On key operations frame picture and export；
Video Rendering unit：It is regarded in the front of sportsman, swinging for side for rendering dynamic comparison analysis result
On frequency and export.
To achieve the purpose of the present invention, the present invention also proposes that a kind of intelligence of golf video compares analysis method,
Include the following steps：
Critical point detection step：To the front of sportsman, side golf video and preprepared coach
The front golf video of member carries out critical point detection, detects nose, neck, right shoulder, right elbow, right finesse, left shoulder, a left side
Elbow, left finesse, right stern, right knee, right ankle, left stern, left knee, left ankle, right eye, left eye, auris dextra, left ear totally 18 bones
Bone key point data；
Key point information sequence pre-treatment step：The crucial point data detected is pre-processed, precision higher is obtained
Key point information sequence；
Golf swing Sequence Detection step：Extract the peak value of the elevation information of sportsman front left finesse key point
Point carries out peak point position to slide back and forth search, finds possible golf swing sequence context, then pass through again
DTW algorithms (dynamic time warping) are dynamic come the golf for calculating possible golf swing sequence context and coach
Make the similarity of sequence, then carry out certain screening, merging treatment, obtains the golf swing sequence of final sportsman
Row；
Golf key operations frame detecting step：It is special that movement locus is extracted from the positive key point information sequence of sportsman
Sign is then based on movement locus feature and carries out the detection of golf key operations frame；
Intelligence compares analytical procedure：According to golf key operations frame, the sportsman front side key point letter detected
Breath sequence and video come into Mobile state+static state, and front+side compares analysis；
Render analysis result step：Intelligence is compared into the comparison analysis result that analytical procedure obtains and renders output.
As a kind of preferred embodiment, the golf swing Sequence Detection step further includes：
Step 3.1：Extract the time point of the peak point of the elevation information change sequence of sportsman front left finesse key point
Position；
Step 3.2：The time point for training frame overhead, this time point are found from the positive key point information sequence of coach
It can first pass through and manually decide in advance, then be used as template；
Step 3.3：Rough golf sequence context is extracted, according to the time point of peak point, before peak point
80 frames as rough golf sequence context at the beginning of point, rear 80 frame of peak point waves as rough golf
The end time point of bar sequence context；
Step 3.4：The peak point of the elevation information change sequence for the sportsman front left finesse key point extracted may
It is the 4th frame (crown frame) of golf key operations frame, it is also possible to the 8th frame (frame of putting away sports equipment), it is also possible to other situations,
The crown frame time point that the present invention chooses coach is matched with the peak point time point of sportsman, then from this point to the upper right corner
Inferior horn carries out traversal search to the left, and the path of search is indicated with W, and k-th of element definition of W is wk=(i, j), definition
The key point information Y of the rough golf sequence of sportsmanu, (1≤u≤m) and the complete golf sequence of coach
The key point information J of rowv, the mapping of (1≤v≤n)；
W=w1,w2,w3,……wk,….wK1≤K<m+n-1；
Searching route must satisfy following four constraint：
Initial conditions：Searching route must be from wk=(iStudent's peak point time point,jTrain crown frame time point) this point is to the upper right corner and to the left
Inferior horn carries out traversal search；
Continuity：If wk-1=(a ', b '), then for next point w in pathk=(a, b) needs to meet (a-a ')
≤ 1 and (b-b ')≤1, that is, can not possibly go to match across some point, point alignment that can only be adjacent with oneself；
Monotonicity：If wk-1=(a ', b '), then for next point w in pathk=(a, b) needs to meet (a-a ')
>=0 and (b-b ') >=0；
End condition：When inferior horn is searched for the left from initial conditions, as u=1 or v=1, terminates inferior horn to the left and searches for,
W at this time1=(istart,jstart)；.When being searched for from initial conditions to the upper right corner, as u=m or v=n, terminate to the upper right corner
Search, w at this timeK=(iend,jend)；When inferior horn traversal search all terminates from initial conditions to the upper right corner and to the left, at this time
istartAs point at the beginning of the more accurate golf sequence of sportsman, iendThe as preliminary height of sportsman
The end time point of golf swing sequence.Then by three above constraints respectively from left finesse, left hand elbow, right finesse,
The preliminary golf sequence of all possible sportsmen is searched in the horizontal direction change information sequence of right hand elbow.
Step 3.5：The preliminary golf sequence of all possible sportsman that step 3.4 extracts is carried out at screening
Reason.
Pass through Dis=DTW (seqy, seqj), seqyRepresent the key point information sequence of sportsman, seqjRepresent coach
Key point information sequence, Disy,jIt represents between the key point information sequence of sportsman and the key point information sequence of coach
Distance.Calculate front left finesse, the left hand of the preliminary golf sequence of all possible sportsman that step 3.4 extracts
Elbow, right finesse, right hand elbow elevation information respectively with the height of coach front left finesse, left hand elbow, right finesse, right hand elbow believe
Cease the distance between sequence Dislw、Disle、Disrw、Disre。DislwRepresentative be the preliminary golf sequence of sportsman just
The similarity of the elevation information sequence of the elevation information of face left finesse and the front left finesse of coach；DisleRepresentative is sportsman
The elevation information sequence of the elevation information of the front left hand elbow of preliminary golf sequence and the front left hand elbow of coach
Similarity；DisrwRepresentative is the front of the elevation information and coach of the front right finesse of the preliminary golf sequence of sportsman
The similarity of the elevation information sequence of right finesse；DisreRepresentative is the front left hand elbow of the preliminary golf sequence of sportsman
Elevation information is at a distance from the elevation information sequence of the front left hand elbow of coach.Then pass through Calculate the preliminary golf sequence and coach's golf sequence
Average distance between row apart from smaller, illustrates this using the average distance as preliminary motion person's sequence at a distance from coach
Two sequences are more similar.Calculate the preliminary golf sequence of all sportsmen and the golf sequence of coach
Distance Diss。
Then all preliminary golf sequence is subjected to non-minimum inhibition, the purpose that non-minimum inhibits be for
Taking out has sequence best in overlapping region that golf sequence, traversal is used as to detect the Gao Er after non-minimum inhibition
Whether husband's sequence that swings is spaced with the presence or absence of two adjacent sequences less than 120 frames, if being not present, after non-minimum is inhibited
Obtained sequence carries out threshold value screening, and threshold value Dis is arranged in the present inventionthreshold=3000, it is more than threshold value DisthresholdSequence
It directly removes, then using remaining sequence as final golf arrangement set.
If individually being taken out in the presence of by the two sequences, processing is merged, is because of some sportsmen the reason of doing so
The speed that swings is slower, may cause step 3.4 the phenomenon that searching route is divided into two pieces there are golf sequence, also may be used
Can be because sportsman has played golf twice within the very short time, hitting time point interval is very small twice, this
Merging treatment is exactly in order to avoid the situation of golf Sequence Detection mistake.
Merging treatment step：
It will be put as point at the beginning of collating sequence, by the latter sequence at the beginning of the previous sequence of two sequences
End time point of the end time point of row as collating sequence.Then sliding window traversal is carried out to collating sequence and calculates movement
The respective faces key point of member's golf front left finesse, left hand elbow, right finesse, right hand elbow four key point informations and coach
The average distance of information sequence.Take the sequence of minimum range as the ultimate sequence range of the collating sequence.If the minimum range
Less than threshold value DisthresholdlongseqWhen=4200, then using this sequence as the golf sequence in the collating sequence range
Row.If more than threshold value Disthresholdlongseq, then it is that speed is led slowly excessively because sportsman swings to illustrate the two sequences not
It causes, but sportsman plays second of golf rapidly again after having played a golf, this time batted twice
Point interval is very small.At this point, the two sequences are all added to final golf sequence as golf sequence
In set.
As a kind of preferred embodiment, golf swing key frame detection further includes：
Golf swing sequence one is divided into：Erect-position frame, draw bar frame, upper boom frame, vertex frame, release frame, batting frame,
Bar frame, frame of putting away sports equipment are sent, totally eight frames.
Step 4.1：The wrist of sportsman, the height change track spy of elbow are extracted according to sportsman front key point information
Sign, shape feature, shoulder breadth feature, right forearm and the horizontal sextant angle variation track feature of two-arm and shoulder, left upper arm are pressed from both sides with horizontal
Angle variation track feature, shoulder breadth feature；
Step 4.2：First of wrist, elbow height change track characteristic extraction golf sequence based on sportsman
Frame (erect-position frame), the 4th frame (vertex frame), the 6th frame (batting frame), the 8th frame (frame of putting away sports equipment)；
Step 4.3：Based on the shape feature of two-arm and shoulder optimization first frame (erect-position frame)；
Step 4.4：Based on the 8th frame of shoulder breadth characteristic optimization (frame of putting away sports equipment)；
Step 4.5：It is in first frame based on right forearm and horizontal sextant angle variation track feature and the second frame (drawing bar frame)
Feature between (erect-position frame) and the 4th frame (vertex frame), the second frame of extraction (drawing bar frame)；
Step 4.6：It is in the second frame based on left upper arm and horizontal sextant angle variation track feature and third frame (upper boom frame)
Feature between (drawing bar frame) and the 4th frame (vertex frame), extraction third frame (upper boom frame)；
Step 4.7：It is in the 4th frame based on right forearm and horizontal sextant angle variation track feature and the 5th frame (release frame)
Feature between (vertex frame) and the 6th frame (batting frame), the 5th frame of extraction (release frame)；
Step 4.8：It is in the 6th frame based on right forearm and horizontal sextant angle variation track feature and the 7th frame (sending bar frame)
The feature of (batting frame) between the 8th frame (frame of putting away sports equipment), the 7th frame of extraction (sending bar frame).
As a kind of preferred embodiment, the intelligence compares analytical procedure, further includes：
Step 5.1：Behavioral characteristics are extracted according to sportsman front key point information, including：The head position of sportsman becomes
Change feature, upper body body change in location feature, shoulder corner variation characteristic, swing rhythm characteristic；
Step 5.2：Static nature is extracted according to sportsman front key point information, including：The two-arm and shoulder of sportsman
Shape feature, left upper arm and left forearm angle feature, left leg stretch feature, shoulder and horizontal direction angle feature；
Step 5.3：Behavioral characteristics are extracted according to sportsman side key point information, including：The head position of sportsman becomes
Change feature, vertebral line feature；
Step 5.4：Static nature is extracted according to sportsman side key point information, including：The micro-bend feature of two legs, left upper arm
Feature is stretched with left forearm angle feature, right upper arm and horizontal direction angle feature, left leg；
Step 5.5：According to step 5.1 to 5.4 extract front behavioral characteristics, front static nature, side behavioral characteristics,
The rule that side static nature combination golf professional knowledge is formulated compares the action that swings of analysis sportsman, and output compares analysis
As a result.
As a kind of preferred embodiment, the rendering analysis result step further includes：
Step 6.1：The dynamic comparison analysis result that step 5.5 is obtained is rendered into the front of sportsman, side swings and regards
On frequency, the place of player motion mistake is outlined with red block can be seen certainly when sportsman sees video with this at a glance
The place of oneself stroke defect, and indicate stroke defect reason on the right of red block and correct the opinions and suggestions of mistake；
Step 6.2：The static comparison analysis result that step 5.5 is obtained, is rendered into the corresponding key operations frame of sportsman
On, the place of player motion mistake is outlined with red block come can be with one soon when sportsman sees key frame picture with this
To the place of the stroke defect of oneself, and indicates stroke defect reason on the right of red block and correct the opinion of mistake and build
View.
The beneficial effects of the invention are as follows：The present invention is directed to the coach that traditional athletic training is all one-to-one or one-to-many
Training is personally instructed, not only receives the limitation of time and location, and training cost is high, coach's resource is nervous, and the Gao Er of market
Husband's analyzer is all the data that sportsman is acquired by sensor mostly, due to needing to dress the equipment of profession or need in body
Upper placement marker, come track human body each position movable information, limit the movement of sportsman to a certain extent, so
It hardly results in and is widely applied.On the other hand, in computer vision research field, most of is all by as Kinect
Depth camera come shoot sportsman action or by common camera come shoot video by the method for Attitude estimation come
Each bone key point information of predicted motion person.The movable information for obtaining sportsman at present just only has both technologies, but only
Only obtain these sportsmen movable information be also not enough to identify sportsman action whether mistake, major part golf at present
Analyzer all can only analyze golf video by people, either review video to analyze or pass through movable information
Data carry out analysis of key kinematic parameter to provide corresponding guidance, but key mistake cannot be directly pointed out as specialized coaches
The technological deficiency in place accidentally, the intelligence that the present invention proposes a set of practicable golf video compare analysis system
And method, the golf video sequence in video point-device can be extracted, and can be very good to detect height
Golf swing key operations frame, and each spy extracted by dynamic+static, front+side multi-angle comparison analysis
Sign may finally position the place of sportsman's malfunction and can provide corresponding guiding opinion, Ke Yi great well
Golf sports universal greatly, save practice cost, have critically important application value.
Description of the drawings
Fig. 1 is the configuration diagram that analysis system is compared using the intelligence of the present invention.
The intelligence of the golf video of Fig. 2 present invention compares analysis system functional block diagram.
Fig. 3 is the infrared detecting device block diagram of the present invention.
Fig. 4 is the Openpose critical point detection device block diagrams of the present invention.
Fig. 5 is the key point information sequence pretreatment unit block diagram of the present invention.
Fig. 6 is the golf swing sequence detecting apparatus block diagram of the present invention.
Fig. 7 is the golf swing key frame detection device block diagram of the present invention.
Fig. 8 is that the intelligence of the present invention compares analytical equipment block diagram.
Fig. 9 is the rendering device block diagram of the present invention.
Specific implementation mode
The invention will be further described below in conjunction with the accompanying drawings.Following embodiment is only used for clearly illustrating the present invention
Technical solution, and not intended to limit the protection scope of the present invention.
It is to compare analysis system schematic diagram institute using the intelligence of the framework golf video of present system such as Fig. 1
Show：
It includes system administration computer, infrared detecting device and two that the intelligence of golf video, which compares analysis system,
Camera, wherein system administration computer include：The intelligence of golf video compares analysis system, coach front swings
Video database.
The intelligence of realization golf video of the present invention compares analysis system, and the specific method is as follows：The present invention utilizes red
Whether outer detection device detection golf is knocked off, if hitting, output signal and hitting time point, and according to hitting time
The front and back 3 seconds editing videos of point, the intelligence that the video of editing is passed to golf video by USB interface compare analysis system
System is compared the Openpose critical point detection devices of analysis system by the intelligence of golf video in the component computer
Front, the side of 2 couples of sportsmen swings video and the coach front taken out in swinging video database from coach front
The video that swings carries out critical point detection and extracts the key point information of sportsman's target person；Then pass through key point information sequence
Sportsman front, side and the positive key point information of coach are carried out filling and the exception of missing values by pretreatment unit 3
The removal of value is filled；Then complete golf swing sequence is detected by golf swing sequence detecting apparatus 4
Row；Then the detection of golf key operations frame is carried out by golf swing key frame detection device 5 again, detected
8 critical actions of golf swing；Analytical equipment 6, which is compared, then through intelligence carries out intelligence comparison analysis,
The place of the malfunction of main positioning sportsman, and provide the corresponding instruction for correcting mistake；It is filled finally by rendering
Analysis result will be compared by, which setting, renders on video and key frame picture, is fed back to video and the visual bandwagon effect of picture
Sportsman allows sportsman that can directly see very much place and the reason of the malfunction of oneself, helps sportsman faster
Malfunction is corrected, training effectiveness is improved.
As shown in the intelligence comparison analysis system of Fig. 2 golf videos of the present invention, including：(front regards camera 1
Frequently), camera 2 (side video), infrared detecting device 1, Openpose critical point detections device 2, key point information sequence are pre-
Processing unit 3, golf swing sequence detecting apparatus 4, golf swing key frame detection device 5, intelligence compare
Analytical equipment 6, rendering device 7, wherein：
The infrared detecting device 1 being connect with camera 1 (positive plane video) and camera 2 (side video), for by red
Whether outer detection detection golf is knocked off, if being knocked off, according to 3 seconds editing videos before and after hitting time point, output fortune
Front, the side of mobilization swing video；
The Openpose critical point detections device 2 being connect with infrared detecting device 1, for the front of input, side
The front of swing video and the complete coach taken out in swinging video database from coach front of sportsman swing
The key point information of human body, disengaging movement person's target person and the key point of non athlete's target person are believed in video detection video
Breath exports the front of sportsman's target, the key point information sequence and the positive key point information sequence of coach of side；
The key point information sequence pretreatment unit 3 being connect with Openpose critical point detections device 2, for input
The positive key point information sequence in the front of sportsman's target, the key point information sequence of side and coach carries out missing values
Filling, exceptional value removal filling, the key point information sequence in the front of sportsman's target, side after output pretreatment
And the positive key point information sequence of coach；
The golf swing sequence detecting apparatus 4 being connect with key point information sequence pretreatment unit 3, for defeated
Similarity mode is carried out between the sportsman front key point information sequence entered and coach front key point information sequence, is extracted
Complete golf swing sequence in the key point information sequence of sportsman front, by golf swing sequence and non-height
Golf swing action sequence is separated, complete golf swing sequence in the key point information sequence of output sportsman front
At the beginning of point and end time point；
The golf being connect with golf swing sequence detecting apparatus 4 and key point information sequence pretreatment unit 3
Swing action key frame detection device 5, for completely high to the sportsman front key point information sequence of input and sportsman
At the beginning of golf swing action sequence and the end time carries out golf swing key frame detection, and golf is dynamic
8 frames are divided into as key frame, are respectively：Erect-position frame draws bar frame, upper boom frame, vertex frame, discharges frame, frame of batting, send bar frame, receive
Bar frame.The time point of 8 key operations frames of this section of sportsman front key point information sequence is exported respectively；
It is connect with golf swing key operations frame detection device 5 and key point information sequence pretreatment unit 3
Intelligence compares analytical equipment 6, the time point of 8 key operations frames for the sportsman front key point information sequence to input
And the key point information sequence progress intelligence comparison analysis in the front, side of sportsman, using golf professional knowledge and tie
It closes the comparison analysis rule that actual conditions are formulated and compares analysis to carry out intelligence, output compares the analysis result of analysis；
The rendering device 7 that analytical equipment 6 and infrared detecting device 1 are connect is compared with intelligence, for the intelligent ratio to input
The video that swings in team's analysis result and the front of sportsman, side renders, and will compare analysis result and is being rendered into movement just
Face, side the video that swings in, output render after sportsman front, side the video that swings.
A program of device of the invention is write by C Plus Plus, and running environment windows10 needs allocation of computer USB to connect
Mouth is to connect infrared detecting device.
As shown in figure 3, infrared detecting device 1, is to be detected to whether golf is knocked off, is carried out if being knocked off
Corresponding processing, specific implementation step are as follows：
Infrared detection unit 11：Detect whether golf is knocked off by infrared ray, if hitting, output signal and
Hitting time point；
The video clipping unit 12 being connect with infrared detection unit：Once receiving the signal from infrared detection unit and hitting
Ball time point exports 6 seconds long fronts, flank movement persons respectively immediately according to 3 seconds before and after hitting time point come editing video
Swing video.
As shown in figure 4, Openpose critical point detections device 2, is sportsman front, side and the coach to input
The positive video that swings carries out critical point detection, and isolates the key point information sequence of sportsman's target person.
Key point extraction unit 21：Using Openpose algorithm models to front, the side of sportsman swing video and
The front of preprepared coach swings video extraction key point information；
The target person key point information extraction unit 22 being connect with key point extraction unit：Using intermediate region and area
Maximal criterion removes the key point information of non athlete's target person, target person and non-targeted personage are distinguished, will regard
Frequency is equally divided into 3 parts in the horizontal direction, takes middle section region, removes the key point letter of the personage in other two parts regions
Breath, then extracts the key point information of target person in the region of middle section, because clapping by area maximal criterion again
When taking the photograph the golf sports of sportsman, sportsman is typically in the intermediate region in video, and sportsman is also from camera
Nearest, and the entire body of complete sportsman can be taken, so region area shared by extraction key point is maximum,
The finally key point information of the key point information and coach of output sportsman's target person.
As shown in figure 5, key point information sequence pretreatment unit 3, front, side for sportsman's target to input
Key point information sequence and the positive key point information sequence of coach carries out the fillings of missing values, the removal of exceptional value is filled out
It fills, the front in the front of sportsman's target, the key point information sequence of side and coach after output pretreatment is crucial
Point information sequence, specific implementation step are as follows：
Missing Data Filling unit 31：Key point for the target person for filling the output of Openpose critical point detection devices
Missing values in information, we utilize the continuity of human motion, have key point information using the former frames or rear several frames of missing values
Linearly fill missing values.
The exceptional value removal fills unit 32 being connect with Missing Data Filling unit：Use sliding window length for the side of 10 frames
Formula fluctuates very big key point information to traverse the removal of key point information sequence, and is refilled using the method for linear interpolation.
As shown in fig. 6, golf swing sequence detecting apparatus 4, for detect sportsman swing it is complete in video
Golf swing sequence.
Key point information peak detection unit 41：The peak point of the ordinate information sequence of left finesse key point is extracted, i.e.,
The peak point of the height change curve of left finesse key point.Purpose is to detect the golf swing sequence of sportsman
Position Approximate.
The golf swing sequence context detection unit being connect with key point information peak detection unit：For passing through
The peak point of key point information peak detection unit output detects the range of golf swing sequence.
As shown in fig. 7, golf swing key frame detection device 5, the golf swing for detecting movement
8 golf key operations frames in sequence.
Feature extraction unit 51：The wrist of sportsman is extracted according to sportsman front key point information, the height of elbow becomes
Change shape feature, shoulder breadth feature, right forearm and horizontal sextant angle variation track feature, the left upper arm of track characteristic, two-arm and shoulder
With horizontal sextant angle variation track feature, shoulder breadth feature；
The key frame detection unit 52 being connect with feature extraction unit：It is all according to being extracted in feature extraction unit 51
Feature extracts 8 golf key operations frames.
As shown in figure 8, intelligence compares analytical equipment 6, swing the behavioral characteristics of action for sportsman to be compared
And static nature, and be compared according to the feature extracted, output compares analysis result.
As shown in figure 9, rendering device 7, the comparison analysis result for exporting intelligence comparison analytical equipment 6 is with intuitive
Mode renders output.
Key frame picture rendering unit：It is rendered in sportsman front, the key of side 8 for static state to be compared analysis result
On action action frame picture, the place of player motion mistake is outlined with red block come can be with one when sportsman sees video with this
The place of the stroke defect of oneself is arrived soon, and is indicated stroke defect reason on the right of red block and corrected the opinion of mistake
And suggestion, and export the picture after rendering；
Video Rendering unit：It is regarded in the front of sportsman, swinging for side for rendering dynamic comparison analysis result
On frequency, the place of player motion mistake is outlined with red block come can be when sportsman sees key frame picture with this at a glance
See the place of the stroke defect of oneself, and indicate on the right of red block stroke defect reason and correct mistake opinion and
It is recommended that and exporting the video after rendering.
The above is only a preferred embodiment of the present invention, it is noted that for the ordinary skill people of the art
For member, without departing from the technical principles of the invention, several improvement and deformations can also be made, these improvement and deformations
Also it should be regarded as protection scope of the present invention.
Claims (16)
1. a kind of intelligence of golf video compares analysis system, which is characterized in that including：
The infrared detecting device that is connect with front camera, side camera, the sportsman for exporting front, side swing
Video；
The Openpose critical point detection devices being connect with the infrared detecting device, the fortune for front, side to input
Swinging for mobilizing and swings video at video from the swing front of the complete coach that video database takes out of coach front
The separation of the key point information, sportsman's target person and non athlete's target person key point information of human body in video is detected,
Export the front of sportsman's target, the key point information sequence and the positive key point information sequence of coach of side；
The key point information sequence pretreatment unit being connect with the Openpose critical point detections device, for the fortune to input
The positive key point information sequence in the front of target, the key point information sequence of side and coach is mobilized to carry out missing values
Filling, the removal of exceptional value filling, the key point information sequence in the front of sportsman's target, side after output pretreatment with
And the positive key point information sequence of coach；
The golf swing sequence detecting apparatus being connect with the key point information sequence pretreatment unit, for input
Sportsman front key point information sequence and coach front key point information sequence between carry out similarity mode, extraction fortune
Complete golf swing sequence in positive key point information sequence is mobilized, by golf swing sequence and Fei Gaoer
Husband swings action sequence phase separation, complete golf swing sequence in the key point information sequence of output sportsman front
Sart point in time and end time point；
The Gao Er being connect with the golf swing sequence detecting apparatus and the key point information sequence pretreatment unit
Husband, which swings, acts key frame detection device, for completely high to the sportsman front key point information sequence of input and sportsman
At the beginning of golf swing action sequence and the end time carries out golf swing key frame detection, and golf is dynamic
8 frames are divided into as key frame, are respectively：Erect-position frame draws bar frame, upper boom frame, vertex frame, discharges frame, frame of batting, send bar frame, receive
Bar frame, and the time point of 8 key operations frames of this section of sportsman front key point information sequence is exported respectively；
The intelligence being connect with the golf swing key frame detection device and the key point information sequence pretreatment unit
Analytical equipment can be compared, for 8 action key frames to the sportsman front key point information sequence of input time points with
And the key point information sequence progress intelligence comparison analysis in the front, side of sportsman, using golf professional knowledge and combine
The comparison analysis rule that actual conditions are formulated compares analysis to carry out intelligence, and output compares the analysis result of analysis；
The rendering device that analytical equipment and the infrared detecting device are connect is compared with the intelligence, for the intelligent ratio to input
The video that swings in team's analysis result and the front of sportsman, side renders, and will compare analysis result and is being rendered into movement just
Face, side the video that swings in, output render after sportsman front, side the video that swings.
2. a kind of intelligence of golf video according to claim 1 compares analysis system, which is characterized in that described
Infrared detecting device further includes：
Infrared detection unit：Detect whether golf is knocked off by infrared ray, if hitting, output signal and square stance
Between point；
The video clipping unit being connect with the infrared detection unit：Once receive signal from the infrared detection unit and
Hitting time point exports 6 seconds long fronts, flank movements respectively immediately according to 3 seconds before and after hitting time point come editing video
Member swings video.
3. a kind of intelligence of golf video according to claim 1 compares analysis system, which is characterized in that described
Openpose critical point detection devices further include：
Key point extraction unit：It is swung video and accurate in advance to front, the side of sportsman using Openpose algorithm models
The front of the coach got ready swings video extraction key point information；
The target person key point information extraction unit being connect with the key point extraction unit：Most using intermediate region and area
Big criterion removes the key point information of non athlete's target person, extracts the key point information of sportsman's target person.
4. a kind of intelligence of golf video according to claim 1 compares analysis system, which is characterized in that described
Key point information sequence pretreatment unit further includes：
Missing Data Filling unit：In the key point information of target person for filling the output of Openpose critical point detection devices
Missing values fill the missing values in key point information using the continuity of human motion using the method for linear interpolation；
The exceptional value removal fills unit being connect with the Missing Data Filling unit：Key is traversed by the way of sliding window
Very big key point information is fluctuated in point information sequence removal, and is refilled using the method for linear interpolation.
5. a kind of intelligence of golf video according to claim 1 compares analysis system, which is characterized in that described
Golf swing sequence detecting apparatus further includes：
Key point information peak detection unit：Peak point for the ordinate information sequence for extracting key point, i.e. key point
The peak point of height change curve, it is therefore an objective to detect the position of the golf swing sequence of sportsman；
The golf swing sequence context detection unit being connect with the key point information peak detection unit：For passing through
The peak point of key point information peak detection unit output detects the range of golf swing sequence.
6. a kind of intelligence of golf video according to claim 1 compares analysis system, which is characterized in that described
Golf swing key frame detection device further includes：
Feature extraction unit：Movement locus feature for extracting sportsman；
The key frame detection unit being connect with the feature extraction unit：For extracting 8 from the movement locus feature of sportsman
A golf acts key frame.
7. a kind of intelligence of golf video according to claim 1 compares analysis system, which is characterized in that described
Intelligence compares analytical equipment, further includes：
Dynamic analysis unit：Front, the side behavioral characteristics of sportsman are extracted, and are compared according to the behavioral characteristics extracted
Analysis；
Static analysis unit：Front, the side static nature of sportsman are extracted, and is compared according to the static nature extracted
Analysis.
8. a kind of intelligence of golf video according to claim 1 compares analysis system, which is characterized in that described
Rendering device further includes：
Key frame picture rendering unit：For rendering the comparison analysis result of key frame in sportsman front, the action of side 8
On key frame picture and export；
Video Rendering unit：For dynamic comparison analysis result to be rendered the video that swings in the front of sportsman, side
And it exports.
9. a kind of method that the intelligence based on golf video described in claim 1 compares analysis system, feature exist
In comprising the following steps：
Critical point detection step：It swings video to the front of sportsman, side golf video and from coach front
The coach front golf video extracted in database carries out critical point detection；
Key point information sequence pre-treatment step：The crucial point data detected is pre-processed, the higher pass of precision is obtained
Key point information sequence；
Golf swing Sequence Detection step：The peak point of the elevation information of sportsman front left finesse key point is extracted,
Peak point position is carried out to slide back and forth search, possible golf swing sequence context is searched for, is then calculated again by DTW
Method calculates the similarity of the golf swing sequence of possible golf swing sequence context and coach, then
Certain screening, merging treatment are carried out, the golf swing sequence of final sportsman is obtained；
Golf key operations frame detecting step：Movement locus feature is extracted from the positive key point information sequence of sportsman,
It is then based on movement locus feature and carries out the detection of golf key operations frame；
Intelligence compares analytical procedure：According to detect golf key operations frame, sportsman front side key point information sequence
Row and video come into Mobile state+static state, and front+side compares analysis；
Render analysis result step：Intelligence is compared into the comparison analysis result that analytical procedure obtains and renders output.
10. a kind of intelligence of golf video according to claim 9 compares analysis method, which is characterized in that institute
Critical point detection step is stated, further includes：
Step SS11：It is swung video and preprepared to front, the side of sportsman using Openpose algorithm models
The front of coach swings video extraction key point information；
Step SS12：The key point information of non athlete's target person is removed using intermediate region and area maximal criterion, it will
Target person is distinguished with non-targeted personage, and video is equally divided into 3 parts in the horizontal direction, takes middle section region, removal
Then the key point information of the personage in other two parts regions is extracted by area maximal criterion in the region of middle section again
The key point information of target person, because when shooting the golf sports of sportsman, during sportsman is typically in video
Between region, and sportsman is also nearest from camera, and can take the entire body of complete sportsman, so
Region area is maximum shared by extraction key point, finally exports the key of the key point information and coach of sportsman's target person
Point information.
11. a kind of intelligence of golf video according to claim 9 compares analysis method, which is characterized in that institute
Key point information sequence pre-treatment step is stated, further includes：
Step SS21：Fill missing values in the key point information of the target person of Openpose critical point detection devices output, profit
With the continuity of human motion, there is key point information using the former frames or rear several frames of missing values linearly to fill missing values；
Step SS22：The removal of key point information sequence is traversed by the way of sliding window fluctuates very big key point letter
Breath, and refilled using the method for linear interpolation.
12. a kind of intelligence of golf video according to claim 9 compares analysis method, which is characterized in that institute
Golf swing Sequence Detection step is stated, further includes：
Step SS31：Extract the time point of the peak point of the elevation information change sequence of sportsman front left finesse key point
It sets；
Step SS32：The time point for training frame overhead is found from the positive key point information sequence of coach, this time point can
It is manually decided with first passing through in advance, is then used as template；
Step SS33：Rough golf sequence context is extracted, according to the time point of peak point, by 80 frames before peak point
Point at the beginning of as rough golf sequence context, rear 80 frame of peak point is as rough golf sequence
The end time point of row range；
Step SS34：The peak point of the elevation information change sequence for the sportsman front left finesse key point extracted may be high
Er Fu acts the vertex frame of key frame, it is also possible to frame of putting away sports equipment, it is also possible to which others action key frame chooses the crown of coach
Frame time point is matched with the peak point time point of sportsman, then from this point to the upper right corner and to the left inferior horn carry out traversal search
Rope, the path of search, as possible golf swing sequence；
Step SS35：The preliminary golf sequence of all possible sportsman that step SS34 is extracted is carried out at screening
Reason.
13. a kind of intelligence of golf video according to claim 9 compares analysis method, which is characterized in that institute
Golf key operations frame detecting step is stated, further includes：
Step SS41：It is extracted according to sportsman front key point information：The wrist of sportsman, the height change track of elbow are special
Sign, shape feature, the shoulder breadth feature of two-arm and shoulder, right forearm and horizontal sextant angle variation track feature, left upper arm are pressed from both sides with horizontal
Angle variation track feature, shoulder breadth feature；
Step SS42：The erect-position frame of wrist, elbow height change track characteristic extraction golf sequence based on sportsman,
Vertex frame, frame of batting, frame of putting away sports equipment；
Step SS43：Optimize erect-position frame based on the shape feature of two-arm and shoulder；
Step SS44：It is put away sports equipment frame based on shoulder breadth characteristic optimization；
Step SS45：Based on right forearm and horizontal sextant angle variation track feature and draws bar frame and be between erect-position frame and vertex frame
Feature, extraction draw bar frame；
Step SS46：It is in and is drawn between bar frame and vertex frame based on left upper arm and horizontal sextant angle variation track feature and upper boom frame
Feature, extract upper boom frame；
Step SS47：It is between vertex frame and batting frame based on right forearm and horizontal sextant angle variation track feature and release frame
Feature, extraction release frame；
Step SS48：Based on right forearm and horizontal sextant angle variation track feature and bar frame is sent to be between batting frame and frame of putting away sports equipment
Feature, extraction send bar frame.
14. a kind of intelligence of golf video according to claim 9 compares analysis method, which is characterized in that institute
It states intelligence and compares analytical procedure, further include：
Step SS51：Behavioral characteristics are extracted according to sportsman front key point information, including：The head position variation of sportsman is special
Sign, upper body body change in location feature, shoulder corner variation characteristic, swing rhythm characteristic；
Step SS52：Static nature is extracted according to sportsman front key point information, including：The two-arm of sportsman and the shape of shoulder
Shape feature, left upper arm and left forearm angle feature, left leg stretch feature, shoulder and horizontal direction angle feature；
Step SS53：Behavioral characteristics are extracted according to sportsman side key point information, including：The head position variation of sportsman is special
Sign, vertebral line feature；
Step SS54：Static nature is extracted according to sportsman side key point information, including：The micro-bend feature of two legs, left upper arm with
Left forearm angle feature, right upper arm and horizontal direction angle feature, left leg stretch feature；
Step SS55：It is special that front behavioral characteristics, front static nature, side dynamic are extracted according to step SS51 to step SS54
The rule that sign, side static nature combination golf professional knowledge are formulated compares the action that swings of analysis sportsman, and output is static
With dynamic comparison analysis result.
15. a kind of intelligence of golf video according to claim 14 compares analysis method, which is characterized in that institute
It states and renders analysis result step, further include：
Step SS61：The dynamic comparison analysis result that the step SS55 is obtained is rendered into the front of sportsman, side swings
On video, the place of player motion mistake is outlined with red block can be seen when sportsman sees video with this at a glance
The place of the stroke defect of oneself, and indicate stroke defect reason on the right of red block and correct the opinion of mistake and build
View；
Step SS62：The static comparison analysis result that the step SS55 is obtained, is rendered into the corresponding key operations of sportsman
On frame, the place of player motion mistake is outlined with red block come can be when sportsman sees key frame picture with this at a glance
See the place of the stroke defect of oneself, and indicate on the right of red block stroke defect reason and correct mistake opinion and
It is recommended that.
16. a kind of intelligence of golf video according to claim 9 compares analysis method, which is characterized in that institute
The possible golf swing sequence of search stated further includes：
The crown frame time point for choosing coach is matched with the peak point time point of sportsman, then from this point to the upper right corner and
Inferior horn carries out traversal search to the left, and the path of search is indicated with W, and k-th of element definition of W is wk=(i, j), definition fortune
Mobilize the key point information Y of rough golf sequenceu, (1≤u≤m) and the complete golf sequence of coach
Key point information Jv, the mapping of (1≤v≤n)；
Wherein, W=w1,w2,w3,......wk,....wK1≤K<m+n-1；
Searching route must satisfy following four constraint：
Initial conditions：Searching route must be from wk=(iStudent's peak point time point,jTrain crown frame time point) this point to the upper right corner and to the left inferior horn into
Row traversal search；
Continuity：If wk-1=(a ', b '), then for next point w in pathk=(a, b) needs to meet (a-a ')≤1 He
(b-b ')≤1, that is, can not possibly go to match across some point, point alignment that can only be adjacent with oneself；
Monotonicity：If wk-1=(a ', b '), then for next point w in pathk=(a, b) needs to meet (a-a ') >=0 He
(b-b′)≥0；
End condition：When inferior horn is searched for the left from initial conditions, as u=1 or v=1, terminates inferior horn to the left and search for, at this time
W1=(istart,jstart)；When being searched for from initial conditions to the upper right corner, as u=m or v=n, terminates and is searched for the upper right corner,
W at this timeK=(iend,jend)；When inferior horn traversal search all terminates from initial conditions to the upper right corner and to the left, i at this timestartI.e.
For sportsman more accurate golf sequence at the beginning of point, iendThe preliminary golf of as sportsman is waved
The end time point of bar sequence.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810170508.4A CN108564596B (en) | 2018-03-01 | 2018-03-01 | Intelligent comparison analysis system and method for golf swing video |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810170508.4A CN108564596B (en) | 2018-03-01 | 2018-03-01 | Intelligent comparison analysis system and method for golf swing video |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108564596A true CN108564596A (en) | 2018-09-21 |
| CN108564596B CN108564596B (en) | 2022-01-25 |
Family
ID=63531402
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810170508.4A Active CN108564596B (en) | 2018-03-01 | 2018-03-01 | Intelligent comparison analysis system and method for golf swing video |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108564596B (en) |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109522793A (en) * | 2018-10-10 | 2019-03-26 | 华南理工大学 | More people's unusual checkings and recognition methods based on machine vision |
| CN109685013A (en) * | 2018-12-25 | 2019-04-26 | 上海智臻智能网络科技股份有限公司 | The detection method and device of header key point in human body attitude identification |
| CN110189364A (en) * | 2019-06-04 | 2019-08-30 | 北京字节跳动网络技术有限公司 | For generating the method and apparatus and method for tracking target and device of information |
| CN110354480A (en) * | 2019-07-26 | 2019-10-22 | 南京邮电大学 | A kind of golf swing scoring estimation method compared based on posture |
| CN111437583A (en) * | 2020-04-10 | 2020-07-24 | 哈尔滨工业大学 | Badminton basic action auxiliary training system based on Kinect |
| CN111898551A (en) * | 2020-07-31 | 2020-11-06 | 北京融链科技有限公司 | Motion analysis method and device in hydrogenation video and electronic equipment |
| CN113255622A (en) * | 2021-07-14 | 2021-08-13 | 北京壹体科技有限公司 | System and method for intelligently identifying sit-up action posture completion condition |
| CN113312947A (en) * | 2020-02-27 | 2021-08-27 | 北京沃东天骏信息技术有限公司 | Method and device for determining behavior object |
| CN113822202A (en) * | 2021-09-24 | 2021-12-21 | 河南理工大学 | Taijiquan attitude detection system based on OpenPose and PyQt |
| CN114534224A (en) * | 2022-01-13 | 2022-05-27 | 上海凯视力成科技有限公司 | Intelligent mirror for golf swing |
| TWI775636B (en) * | 2020-12-28 | 2022-08-21 | 日商樂天集團股份有限公司 | Golf swing analysis system, golf swing analysis method and information memory medium |
| CN116095363A (en) * | 2023-02-09 | 2023-05-09 | 西安电子科技大学 | Mobile terminal short video highlight moment editing method based on key behavior recognition |
| CN116152708A (en) * | 2023-01-06 | 2023-05-23 | 北京体提科技有限公司 | Method and device for extracting effective actions of golf item, storage medium and equipment |
| EP4224758A4 (en) * | 2021-05-25 | 2024-05-29 | Tencent Technology (Shenzhen) Company Limited | Data retransmission processing method and apparatus, computer device, and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101918090A (en) * | 2007-11-05 | 2010-12-15 | 布莱恩·弗朗西斯·穆尼 | Apparatus and method for analysing a golf swing |
| CN1887386B (en) * | 2006-07-24 | 2012-05-23 | 北方工业大学 | Index dial analysis method and device of golf trace |
| CN103203096A (en) * | 2012-01-11 | 2013-07-17 | 幻音科技(深圳)有限公司 | Golf swing information analysis method, related device and analysis system |
| CN103759739A (en) * | 2014-01-21 | 2014-04-30 | 北京诺亦腾科技有限公司 | Multimode motion measurement and analysis system |
-
2018
- 2018-03-01 CN CN201810170508.4A patent/CN108564596B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1887386B (en) * | 2006-07-24 | 2012-05-23 | 北方工业大学 | Index dial analysis method and device of golf trace |
| CN101918090A (en) * | 2007-11-05 | 2010-12-15 | 布莱恩·弗朗西斯·穆尼 | Apparatus and method for analysing a golf swing |
| CN103203096A (en) * | 2012-01-11 | 2013-07-17 | 幻音科技(深圳)有限公司 | Golf swing information analysis method, related device and analysis system |
| CN103203096B (en) * | 2012-01-11 | 2015-04-22 | 幻音科技(深圳)有限公司 | Golf swing information analysis method, related device and analysis system |
| CN103759739A (en) * | 2014-01-21 | 2014-04-30 | 北京诺亦腾科技有限公司 | Multimode motion measurement and analysis system |
Cited By (21)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109522793A (en) * | 2018-10-10 | 2019-03-26 | 华南理工大学 | More people's unusual checkings and recognition methods based on machine vision |
| CN109522793B (en) * | 2018-10-10 | 2021-07-23 | 华南理工大学 | Method for detecting and identifying abnormal behaviors of multiple persons based on machine vision |
| CN109685013A (en) * | 2018-12-25 | 2019-04-26 | 上海智臻智能网络科技股份有限公司 | The detection method and device of header key point in human body attitude identification |
| CN110189364A (en) * | 2019-06-04 | 2019-08-30 | 北京字节跳动网络技术有限公司 | For generating the method and apparatus and method for tracking target and device of information |
| CN110189364B (en) * | 2019-06-04 | 2022-04-01 | 北京字节跳动网络技术有限公司 | Method and device for generating information, and target tracking method and device |
| CN110354480A (en) * | 2019-07-26 | 2019-10-22 | 南京邮电大学 | A kind of golf swing scoring estimation method compared based on posture |
| CN110354480B (en) * | 2019-07-26 | 2021-04-16 | 南京邮电大学 | Golf swing action score estimation method based on posture comparison |
| CN113312947A (en) * | 2020-02-27 | 2021-08-27 | 北京沃东天骏信息技术有限公司 | Method and device for determining behavior object |
| CN111437583A (en) * | 2020-04-10 | 2020-07-24 | 哈尔滨工业大学 | Badminton basic action auxiliary training system based on Kinect |
| CN111437583B (en) * | 2020-04-10 | 2021-06-29 | 哈尔滨工业大学 | Badminton basic action auxiliary training system based on Kinect |
| CN111898551A (en) * | 2020-07-31 | 2020-11-06 | 北京融链科技有限公司 | Motion analysis method and device in hydrogenation video and electronic equipment |
| CN111898551B (en) * | 2020-07-31 | 2024-08-27 | 北京融链科技有限公司 | Action analysis method and device in hydrogenated video and electronic equipment |
| TWI775636B (en) * | 2020-12-28 | 2022-08-21 | 日商樂天集團股份有限公司 | Golf swing analysis system, golf swing analysis method and information memory medium |
| EP4224758A4 (en) * | 2021-05-25 | 2024-05-29 | Tencent Technology (Shenzhen) Company Limited | Data retransmission processing method and apparatus, computer device, and storage medium |
| CN113255622B (en) * | 2021-07-14 | 2021-09-21 | 北京壹体科技有限公司 | System and method for intelligently identifying sit-up action posture completion condition |
| CN113255622A (en) * | 2021-07-14 | 2021-08-13 | 北京壹体科技有限公司 | System and method for intelligently identifying sit-up action posture completion condition |
| CN113822202A (en) * | 2021-09-24 | 2021-12-21 | 河南理工大学 | Taijiquan attitude detection system based on OpenPose and PyQt |
| CN114534224A (en) * | 2022-01-13 | 2022-05-27 | 上海凯视力成科技有限公司 | Intelligent mirror for golf swing |
| CN116152708A (en) * | 2023-01-06 | 2023-05-23 | 北京体提科技有限公司 | Method and device for extracting effective actions of golf item, storage medium and equipment |
| CN116095363A (en) * | 2023-02-09 | 2023-05-09 | 西安电子科技大学 | Mobile terminal short video highlight moment editing method based on key behavior recognition |
| CN116095363B (en) * | 2023-02-09 | 2024-05-14 | 西安电子科技大学 | Mobile terminal short video highlight moment editing method based on key behavior recognition |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108564596B (en) | 2022-01-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108564596A (en) | A kind of the intelligence comparison analysis system and method for golf video | |
| WO2021051579A1 (en) | Body pose recognition method, system, and apparatus, and storage medium | |
| Bloom et al. | G3D: A gaming action dataset and real time action recognition evaluation framework | |
| US10186041B2 (en) | Apparatus and method for analyzing golf motion | |
| Jhuang et al. | Towards understanding action recognition | |
| US11798318B2 (en) | Detection of kinetic events and mechanical variables from uncalibrated video | |
| CN112487965B (en) | Intelligent fitness action guiding method based on 3D reconstruction | |
| CN112422946B (en) | Intelligent yoga action guidance system based on 3D reconstruction | |
| Pei et al. | An embedded 6-axis sensor based recognition for tennis stroke | |
| US20210001174A1 (en) | Method of real time monitoring of a person during an event and event dynamics system thereof | |
| US20230377336A1 (en) | Method of operating server providing sports video-based platform service | |
| JP4256289B2 (en) | Golf swing diagnostic system | |
| WO2018122956A1 (en) | Sport motion analysis support system, method and program | |
| CN110227243A (en) | Table tennis practice intelligent correcting system and its working method | |
| WO2023108842A1 (en) | Motion evaluation method and system based on fitness teaching training | |
| JP7198661B2 (en) | Object tracking device and its program | |
| Kelly et al. | Visualisation of tennis swings for coaching | |
| Liu et al. | Research on action recognition of player in broadcast sports video | |
| Neher et al. | Hyperstacknet: A hyper stacked hourglass deep convolutional neural network architecture for joint player and stick pose estimation in hockey | |
| Li et al. | Automatic video-based analysis of athlete action | |
| CN116246347A (en) | Standard judging method and scoring device for basic motor skills of human body | |
| Karunaratne et al. | Objectively measure player performance on Olympic weightlifting | |
| Škvarlová | Labeled Dataset of Speed Climbing Performances | |
| Malawski et al. | Automatic analysis of techniques and body motion patterns in sport | |
| Yu et al. | A Machine Learning Based Method to Establish Normative Action Model for Table Tennis |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |