CN109087328A - Shuttlecock drop point site prediction technique based on computer vision - Google Patents
Shuttlecock drop point site prediction technique based on computer vision Download PDFInfo
- Publication number
- CN109087328A CN109087328A CN201810549488.1A CN201810549488A CN109087328A CN 109087328 A CN109087328 A CN 109087328A CN 201810549488 A CN201810549488 A CN 201810549488A CN 109087328 A CN109087328 A CN 109087328A
- Authority
- CN
- China
- Prior art keywords
- shuttlecock
- place
- drop point
- image
- point site
- 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
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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/215—Motion-based segmentation
-
- 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
-
- 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30221—Sports video; Sports image
- G06T2207/30224—Ball; Puck
Abstract
The present invention relates to shuttlecock drop point site prediction techniques based on computer vision, the video frame picture obtained according to badminton game video, using place center as origin, the perspective transform relationship between model place, determines shuttlecock place model with calculating video frame picture mesoptile court；Then moving region and background parts are distinguished using time differencing method by moving region detection module；Binary conversion treatment, image segmentation are carried out to difference image, Contour extraction extracts the characteristic information of shuttlecock in turn；The flight path of shuttlecock is determined using track following algorithm；It is predicted using track drop point site of the Kalman filtering algorithm to shuttlecock；According to track drop point site, the coordinate in practical shuttlecock place is converted to using perspective transform method, to realize the prediction to shuttlecock drop point site.The prediction of the shuttlecock drop point site under complex scene when this method can be suitable for illumination gradual change, air speed influence, airflow influence or shuttlecock speed quickly.
Description
Technical field
The invention belongs to Digital Image Processing, pattern-recognition and machine learning, video object detection technique field, especially
It is related to a kind of method that shuttlecock drop point site is predicted using computer vision technique.
Background technique
Movement meeting, Olympic Games feather can be applied to using the method for computer vision technique prediction shuttlecock drop point site
In ball match, major function is the badminton game video to input, and system can quickly identify badminton game place
The drop point site of shuttlecock is predicted according to the flight path of shuttlecock with the shuttlecock of flight.Predicting feather placement position
In the method set, image preprocessing, image recognition, image segmentation, moving target are carried out to the badminton game video of input and examined
The sequence of operations such as survey, target following, motion profile prediction.Detection to badminton target is one of crucial ring
Section, extracting shuttlecock key feature is the key that carry out next step target following operation.For this target following of shuttlecock
Effect will directly affect the performance of whole system.Therefore many scholars have carried out deep grind for target following in recent years
Study carefully, and proposes many algorithms；However, the tracking for this specific objective of shuttlecock, it is easy to be illuminated by the light, the factors such as wind speed
It influences, so that being still a challenge for the prediction of shuttlecock drop point site.
Summary of the invention
In order to overcome the shortcomings of the prior art described above, a kind of based on computer vision it is an object of the invention to propose
Shuttlecock drop point site prediction technique, can be suitable for illumination gradual change, air speed influence, airflow influence or shuttlecock speed very
The prediction of the shuttlecock drop point site under complex scene when fast.
In order to achieve the above object, the technical scheme adopted by the invention is that:
Shuttlecock drop point site prediction technique based on computer vision, which comprises the steps of:
Step 1: badminton game video, and input video frame picture collection module are acquired by video acquisition module；
Step 2: the badminton game video that step 1 obtains being pre-processed by video frame picture collection module, is obtained
Video frame picture, and input picture preprocessing module；
Step 3: the video frame picture that step 2 obtains being calculated using place center as origin by image pre-processing module
Perspective transform relationship of the video frame picture mesoptile court ground between model place, determines shuttlecock place model；
Step 4: the shuttlecock place model that the video frame picture and step 3 obtained according to step 2 is established, by moving region
Detection module distinguishes moving region and background parts using time differencing method；
Step 5: the moving region distinguished according to step 4 and background parts are identified by target identification module to difference diagram
As carrying out binary conversion treatment, image segmentation, Contour extraction extracts the characteristic information of shuttlecock in turn；
Step 6: the shuttlecock characteristic information extracted according to step 5 uses track following algorithm by target tracking module
Determine the flight path of shuttlecock；
Step 7: according to the flight path for the shuttlecock that step 6 obtains, karr being used by target trajectory drop point prediction module
Graceful filtering algorithm predicts the track drop point site of shuttlecock；
Step 8: according to the track drop point site of step 7, being extracted, utilized by coordinate of the mapping block to drop point site
Mapping relations mathematically are converted to the coordinate in practical shuttlecock place using perspective transform method, fall to realize to shuttlecock
The prediction of point position.
Further, using place center as origin described in step 3, with calculating video frame picture mesoptile court and mould
Perspective transform relationship between type place, determines shuttlecock place model, and specific steps include:
Step 3.1, acquire the video camera of badminton game be it is fixed, wherein precincts of the bath is set as 13m × 6m, and net is high
Degree is set as 1.5m, and the characteristic point marked manually is respectively in four intersection points of four angle points of net and court rectangular area；
Step 3.2, according to the obtained video frame picture of step 2, wherein the available image of first frame sufficiently includes court
The video frame of feature is needed to extract movement home court, be picked since badminton place is made of net and boundary line
Except unwanted Background, so that subsequent operation extracts court using the method counted based on color space histogram
Main region；
Step 3.3, the court main region extracted according to step 3.2, using quasi- based on Hough transform and least-squares line
It closes the method combined to detect the lines in shuttlecock place, while identifying the outermost layer field of shuttlecock rectangular field ground
Four characteristic points of ground wire intersection, using based on the simple characteristic point matching method for surrounding method, i.e., using place center as origin, field
Outermost layer four feature point sets in ground sort according to certain rules as the Feature Points Matching pair for calculating initial perspective transformation matrix
After matched, so that it may the initial perspective transform relationship between first frame picture and model is calculated, to establish place mould
Type.
Further, use time differencing method described in step 4 distinguishes moving region and background parts, specific steps packet
It includes:
Step 4.1, the ball field model established according to step 3, following step is to identify moving region and background, back
Scape refers to the scene other than shuttlecock and sportsman；
Step 4.2, moving region is identified using time differencing method, time differencing method is to two in continuous image sequence
Or three consecutive frames carry out time difference pixel-based, the basic principle of two frame differences is exactly by front and back two field pictures for picture
The gray value of vegetarian refreshments subtracts each other, and in the case where ambient brightness variation is little, if corresponding pixel grey scale differs very little, can recognize
It is static for scenery herein, if the gray-value variation in image-region somewhere is very big, it is believed that due to moving object in image
Caused by body, these zone markers are got up, the pixel region marked using these, so that it may find out moving target in the picture
Position.
Further, binary conversion treatment, image segmentation are carried out to difference image described in step 5, Contour extraction mentions in turn
The characteristic information of shuttlecock is taken out, specific steps include:
Step 5.1, the difference image obtained for step 4 needs to extract the fortune of sportsman and shuttlecock from background
Dynamic region, by binary conversion treatment and image Segmentation Technology, existing Threshold sementation has fixed threshold method and maximum kind
Between variance method；
Step 5.2, it is made of since image can be regarded as foreground and background, the gray scale of sportsman and shuttlecock and background
Section differs greatly, and target and background visually have very big contrast, since maximum variance between clusters are to a part of variogram
As badminton region can only be partitioned into, Athletess region cannot be divided, so selection fixed threshold method carrys out segmentation figure
Picture；
Step 5.3, for difference image after binary conversion treatment, the target area moved in a frame image has become one
Then the region that can be connected measures the feature of each region and identifies shuttlecock and sportsman；
Step 5.4, for a frame bianry image, zone marker is carried out to existing Connected component, according to the wheel of shuttlecock
Wide feature identifies the characteristic point on shuttlecock, may recognize that shuttlecock using Contour extraction.
Further, the flight path of shuttlecock, specific steps packet are determined described in step 6 using track following algorithm
It includes:
Step 6.1, certain matching criterior is established to carry out the center of gravity of the moving region between consecutive frame in sequence image
Matching, to obtain the track of all moving objects；
Step 6.2, when not carrying out at matched with longest track, system turns to whether analysis longest track has plumage
The motion profile of ball top, if longest track is the motion profile of shuttlecock, system extracts the flight path of shuttlecock；
Step 6.3, when the motion profile that longest track is not shuttlecock, system continues to match later data.
Further, it is predicted described in step 7 using track drop point site of the Kalman filtering algorithm to shuttlecock,
Specific steps include:
Step 7.1, it first has to establish shuttlecock locus model, kth is predicted according to the state position at -1 moment of shuttlecock kth
The state position at moment；
Step 7.2, the prediction error at -1 moment of shuttlecock kth is calculated, estimates the prediction error at kth moment；
Step 7.3, kalman gain and system optimal estimated value are calculated；
Step 7.4, system kth moment prediction error is calculated, nonlinear optimization processing is carried out, reduction system prediction misses
Difference improves precision of prediction；
Step 7.5, the position at system output shuttlecock kth moment.
Further, the coordinate of drop point site is extracted described in step 8, using mapping relations mathematically, is adopted
The coordinate in practical shuttlecock place is converted to perspective transform method, specific steps include:
Step 8.1, according to step 7 Kalman filtering algorithm to the predicted position of track drop point, available shuttlecock exists
Coordinate position under two-dimentional place model；
Step 8.2, it is converted using geometrical perspective, its position is transformed under world coordinate system, can be obtained raw in reality
Coordinate position under living.
Compared with prior art, the beneficial effects of the present invention are: using computer vision technique the invention proposes a kind of
To the method for shuttlecock drop point site prediction, compared with existing algorithm, remarkable advantage is the algorithm:
(1) method based on color space Histogram statistics is proposed, court main region is extracted, is fully able to
The main region for meeting conventional sports video is extracted, and has preferable feasibility, versatility, validity.
(2) it proposes to detect moving object using time differencing method, and binaryzation is carried out to difference image, using fixed value method
Bianry image is split, the feature of shuttlecock is extracted, provides the foundation for target following.Calculus of finite differences detection object not by
The influence of light variation, and arithmetic speed is fast, can satisfy the needs applied in real time.
(3) it proposes and matched jamming is carried out to target using object of which movement characteristic, obtain track group, transported further according to shuttlecock
Dynamic rail mark extracts its track, and the approximate location of shuttlecock is predicted according to Kalman filtering algorithm, is obtained using perspective transform
The specific coordinate in shuttlecock place in real life.
Detailed description of the invention
Fig. 1 is shuttlecock place illustraton of model.
Fig. 2 is based on the simple Feature Points Matching schematic diagram for surrounding method.
Fig. 3 is shuttlecock drop point site prediction technique flow chart of the present invention.
Fig. 4 is the time differencing method algorithm flow chart being related in shuttlecock drop point site prediction technique of the present invention.
Fig. 5 is the track following algorithm flow chart being related in shuttlecock drop point site prediction technique of the present invention.
Fig. 6 is the Kalman filtering algorithm flow chart being related in shuttlecock drop point site prediction technique of the present invention.
In Fig. 1: the court 1-, 2- net, 3- camera lens, 4- digital camera.
Specific embodiment
For the ease of those of ordinary skill in the art understand and implement the present invention, below with reference to embodiment to the present invention make into
The detailed description of one step, it should be understood that implementation example described herein is merely to illustrate and explain the present invention, and is not used to limit
The fixed present invention.
As shown in figure 3, shuttlecock drop point site prediction technique based on computer vision of the invention, specific steps are such as
Under:
Step 1: the video of video acquisition device acquisition badminton game.
Specific practice is: utilizing Canon's video camera acquisition device, the video of one section of badminton game of high-definition shooting, wherein taking the photograph
The frame per second of camera is 120fps/s.
Step 2: the sports tournament video inputted to step 1 pre-processes, and obtains video frame picture.
Specific practice is: installing Free Studio software under window7 environment, converts the HD video of shooting to
Video frame (and picture), frame per second when conversion are set as 120, required video frame picture can be obtained.
Step 3: the video frame picture that step 2 obtains being calculated using place center as origin by image pre-processing module
Perspective transform relationship of the video frame picture mesoptile court ground between model place, determines shuttlecock place model.
Specific practice is: the video frame picture obtained for step 2, is used according to video frame picture straight based on color space
The method of side's figure statistics, extracts court main region.It is mutually tied using based on Hough transform with least squares line fitting simultaneously
The method of conjunction detects the lines in shuttlecock place.Shuttlecock place as shown in Figure 1, with shuttlecock rectangular field subscript
Eight labels (respectively angles of court and net quadrangle totally eight points) of note are used as characteristic point, and it is special to extract the intersection of place line
Point is levied, using based on the characteristic point matching method for simply surrounding method, using place center as origin, establishes known place model and spy
Sign point matching, the perspective transform relationship between model place, determines shuttlecock with calculating video frame picture mesoptile court
Place model.
Wherein, step 3 includes:
Step 3.1, acquire the video camera of badminton game be it is fixed, wherein precincts of the bath is set as 13m × 6m, and net is high
Degree is set as 1.5m, and the characteristic point marked manually is respectively in four angle points of four angle points of net and rectangular area.Feather
Court, the position of video camera, and the point marked manually is as shown in Figure 1.
Step 3.2, according to the obtained video frame picture of step 2, wherein the available image of first frame sufficiently includes court
The video frame of feature, since there is very big difference in badminton place and other environment, there are net and boundary in athletic ground
Line composition needs to extract movement home court, unwanted Background is rejected, so as to subsequent operation.Using based on face
The method of colour space statistics with histogram extracts court main region.
Step 3.3, the court main region extracted according to step 3.2, using quasi- based on Hough transform and least-squares line
It closes the method combined to detect the lines in shuttlecock place, while identifying the outermost layer field of shuttlecock rectangular field ground
Four marked manually the characteristic point of ground wire intersection.Using the characteristic point matching method based on simple encirclement method, i.e., in place
The heart is origin, and outermost layer four feature point sets in place are as the Feature Points Matching pair for calculating initial perspective transformation matrix, according to one
It is matched after fixed rule compositor, so that it may the initial perspective transform relationship between first frame picture and model is calculated, from
And establish place model.
Step 4: the shuttlecock place model that the video frame picture and step 3 obtained according to step 2 is established, by moving region
Detection module distinguishes moving region and background parts using time differencing method.
Specific practice is: the place model established for step 3, following step is exactly to isolate moving target and back
Scape image.Due to acquisition video video camera be it is fixed, background is relatively fixed not to change, and illumination is relatively stable, plumage
The movement of ball top is not in block.Time differencing method is that two or three adjacent interframe use in continuous image sequence
Time difference pixel-based, and thresholding extracts the moving region in image, so examined using time differencing method
Survey the moving target in sequence image.By operation, the motion parts between image are clearly showed that, can distinguish fortune
Dynamic region and background parts.Time differencing method algorithm flow chart is as shown in Figure 4.
Wherein, step 4 includes:
Step 4.1, the ball field model established according to step 3, following step are exactly to identify moving region and back
Scape.Background refers to the scene other than shuttlecock and sportsman.
Step 4.2, moving region is identified using time differencing method.Time differencing method is to two in continuous image sequence
Or three consecutive frames carry out time difference pixel-based, the basic principle of two frame differences is exactly by front and back two field pictures for picture
The gray value of vegetarian refreshments subtracts each other, and in the case where ambient brightness variation is little, if corresponding pixel grey scale differs very little, can recognize
It is static for scenery herein, if the gray-value variation in image-region somewhere is very big, it is believed that due to moving object in image
Caused by body, these zone markers are got up, the pixel region marked using these, so that it may find out moving target in the picture
Position.
Step 5: the moving region distinguished according to step 4 and background parts are identified by target identification module to difference diagram
As carrying out binary conversion treatment, image segmentation, Contour extraction extracts the characteristic information of shuttlecock in turn.
Specific practice is: the moving region distinguished according to step 4 and background parts, and connecing down is exactly identification and analysis image
In target, that is to say shuttlecock, need to separate target from image.For the difference image that step 4 obtains, need
The moving region of shuttlecock is extracted from background image, this needs to carry out binary conversion treatment to difference image, to carry out image
Segmentation extracts shuttlecock from complicated background, using fixed threshold split plot design, using Contour extraction and then extracts
The feature of shuttlecock.
Wherein, step 5 includes:
Step 5.1, the difference image obtained for step 4 needs to extract the fortune of sportsman and shuttlecock from background
Dynamic region, by binary conversion treatment, this processing technique is also image Segmentation Technology.Existing Threshold sementation has fixed threshold
Value method and maximum variance between clusters.
Step 5.2, it is made of since image can be regarded as foreground and background, the gray scale of sportsman and shuttlecock and background
Section differs greatly, and target and background visually have very big contrast.Since maximum variance between clusters are to a part of variogram
As badminton region can only be partitioned into, Athletess region cannot be divided, so the threshold value of selection fixation carrys out segmentation figure
Picture.
Step 5.3, for difference image after binary conversion treatment, the target area moved in a frame image has become one
Then the region that can be connected measures the feature of each region and identifies shuttlecock and sportsman.
Step 5.4, for a frame bianry image, zone marker is carried out to existing Connected component, according to the wheel of shuttlecock
Wide feature identifies the characteristic point on shuttlecock, may recognize that shuttlecock using Contour extraction.
Step 6: the shuttlecock characteristic information extracted according to step 5 uses track following algorithm by target tracking module
Determine the flight path of shuttlecock.
Specific practice is: the shuttlecock characteristic information extracted according to step 5, uses track for shuttlecock target following
Track algorithm.Path matching is the mistake that tracking and matching criterion is associated the movement center of gravity data detected on sequence image
Journey.It determines that matching criterior is to carry out the key point of path matching, first has to the motion profile record for establishing shuttlecock, then use
The spatial coherence of badminton is as path matching criterion, namely utilizes point on track during badminton and point
Between distance is continuous and track on angle between points keep the criterion of continuous feature as path matching, thus to plumage
Ball top carries out target following.Track following algorithm flow chart is as shown in Figure 5.
Wherein, step 6 includes:
Step 6.1, certain matching criterior is established to carry out the center of gravity of the moving region between consecutive frame in sequence image
Matching, to obtain the track of all moving objects.
Step 6.2, when not carrying out at matched with longest track, system turns to whether analysis longest track has plumage
The motion profile of ball top, if longest track is the motion profile of shuttlecock, system extracts the flight path of shuttlecock.
Step 6.3, when the motion profile that longest track is not shuttlecock, system continues to match later data.
Step 7: according to the flight path for the shuttlecock that step 6 obtains, karr being used by target trajectory drop point prediction module
Graceful filtering algorithm predicts the track drop point site of shuttlecock.Kalman filtering algorithm flow chart is as shown in Figure 6.
Wherein, step 7 includes:
Step 7.1, it first has to establish shuttlecock locus model, kth is predicted according to the state position at -1 moment of shuttlecock kth
The state position at moment.
Step 7.2, the prediction error at -1 moment of shuttlecock kth is calculated, estimates the prediction error at kth moment.
Step 7.3, kalman gain and system optimal estimated value are calculated.
Step 7.4, system kth moment prediction error is calculated, nonlinear optimization processing is carried out, reduction system prediction misses
Difference improves precision of prediction.
Step 7.5, the position at system output shuttlecock kth moment.
Step 8: according to the track drop point site of step 7, being extracted, utilized by coordinate of the mapping block to drop point site
Mapping relations mathematically are converted to the coordinate in practical shuttlecock place using perspective transform method, fall to realize to shuttlecock
The prediction of point position.
Wherein, step 8 includes:
Step 8.1, according to step 7 Kalman filtering algorithm to the predicted position of track drop point, available shuttlecock exists
Coordinate position under two-dimentional place model.
Step 8.2, it is converted using geometrical perspective, its position is transformed under world coordinate system, can be obtained raw in reality
Coordinate position under living.
Main innovation point of the present invention is embodied in:
(1) determination in shuttlecock place is marked manually at four angles of four, shuttlecock place corner and net
Characteristic point, using the method combined based on Hough transform with least squares line fitting to the lines in shuttlecock place into
Row detection, identifies eight, shuttlecock rectangle place characteristic point, establishes place model according to characteristic point.Specific place model is shown in Fig. 1.
(2) according to video frame picture, using time differencing method be in continuous image sequence two or three it is adjacent
Interframe uses time difference pixel-based, and the moving region in image is extracted using thresholding, distinguishes moving region
And background parts.For obtained difference image, the moving region of shuttlecock is extracted from background image, to difference image into
Row binary conversion treatment carries out image segmentation, and shuttlecock is extracted from complicated background, using fixed threshold split plot design,
The feature of shuttlecock is extracted using Contour extraction.Specific algorithm flow chart is shown in Fig. 4.
(3) track following algorithm is used to shuttlecock target following, path matching is tracking and matching criterion to sequence image
On the process that is associated of the movement center of gravity data that detects.It determines that matching criterior is to carry out the key point of path matching, establishes
The motion profile of shuttlecock records, and then using the spatial coherence of badminton as path matching criterion, namely utilizes
Angle of the distance during badminton on track between points continuously and on track between points keeps continuous
Criterion of the feature as path matching, to carry out target following to shuttlecock.Specific algorithm flow chart is shown in Fig. 5.
(4) it first has to establish shuttlecock locus model, the kth moment is predicted according to the state position at -1 moment of shuttlecock kth
State position, first calculate the prediction error at -1 moment of shuttlecock kth, estimate the prediction error at kth moment, calculate karr
Graceful gain and system optimal estimated value show that system predicts error at the kth moment, carry out nonlinear optimization processing, reduce system
System prediction error, improves precision of prediction.The position at final system output shuttlecock kth moment.It can be realized and filtered using Kalman
Wave algorithm realizes the prediction to shuttlecock drop point site.Specific algorithm flow chart is shown in Fig. 6.
(5) drop point site according to Kalman filtering algorithm under two-dimentional place model is turned drop point using perspective transform
It changes under world coordinate system, that is to say the drop point site under actual life.
It should be understood that the part that this specification does not elaborate belongs to the prior art.
It should be understood that the above-mentioned description for preferred embodiment is more detailed, can not therefore be considered to this
The limitation of invention patent protection range, those skilled in the art under the inspiration of the present invention, are not departing from power of the present invention
Benefit requires to make replacement or deformation under protected ambit, fall within the scope of protection of the present invention, this hair
It is bright range is claimed to be determined by the appended claims.
Claims (7)
1. shuttlecock drop point site prediction technique based on computer vision, which comprises the steps of:
Step 1: badminton game video, and input video frame picture collection module are acquired by video acquisition module；
Step 2: the badminton game video that step 1 obtains being pre-processed by video frame picture collection module, obtains video
Frame picture, and input picture preprocessing module；
Step 3: video being calculated using place center as origin to the video frame picture that step 2 obtains by image pre-processing module
Perspective transform relationship of the frame picture mesoptile court ground between model place, determines shuttlecock place model；
Step 4: the shuttlecock place model that the video frame picture and step 3 obtained according to step 2 is established is detected by moving region
Module distinguishes moving region and background parts using time differencing method；
Step 5: the moving region distinguished according to step 4 and background parts, by target identification module identification to difference image into
Row binary conversion treatment, image segmentation, Contour extraction extract the characteristic information of shuttlecock in turn；
Step 6: the shuttlecock characteristic information extracted according to step 5 is determined by target tracking module using track following algorithm
The flight path of shuttlecock；
Step 7: according to the flight path for the shuttlecock that step 6 obtains, being filtered by target trajectory drop point prediction module using Kalman
Wave algorithm predicts the track drop point site of shuttlecock；
Step 8: according to the track drop point site of step 7, being extracted by coordinate of the mapping block to drop point site, utilize mathematics
On mapping relations, the coordinate in practical shuttlecock place is converted to using perspective transform method, thus realize to feather placement position
The prediction set.
2. shuttlecock drop point site prediction technique based on computer vision according to claim 1, which is characterized in that step
Using place center as origin described in rapid 3, the perspective between model place becomes with calculating video frame picture mesoptile court
Relationship is changed, determines shuttlecock place model, specific steps include:
Step 3.1, acquire the video camera of badminton game be it is fixed, wherein precincts of the bath is set as 13m × 6m, and net height is set
It is set to 1.5m, the characteristic point marked manually is respectively in four angle points of four angle points of net and court rectangular area；
Step 3.2, according to the obtained video frame picture of step 2, wherein the available image of first frame sufficiently includes court feature
Video frame, since badminton place is made of net and boundary line, need to movement home court extract, reject not
The Background needed, so that subsequent operation extracts court primary area using the method counted based on color space histogram
Domain；
Step 3.3, the court main region extracted according to step 3.2, using based on Hough transform and least squares line fitting phase
In conjunction with method the lines in shuttlecock place are detected, while identifying the outermost layer place line of shuttlecock rectangular field ground
Four characteristic points of intersection, using based on the simple characteristic point matching method for surrounding method, i.e., using place center as origin, place is most
Four feature point sets of outer layer sort laggard according to certain rules as the Feature Points Matching pair for calculating initial perspective transformation matrix
Row matching, so that it may the initial perspective transform relationship between first frame picture and model is calculated, to establish place model.
3. shuttlecock drop point site prediction technique based on computer vision according to claim 2, which is characterized in that step
Use time differencing method described in rapid 4 distinguishes moving region and background parts, and specific steps include:
Step 4.1, the ball field model established according to step 3, following step is to identify moving region and background, and background refers to
Be scene other than shuttlecock and sportsman；
Step 4.2, moving region is identified using time differencing method, time differencing method is to two or three in continuous image sequence
A consecutive frame carries out time difference pixel-based, and the basic principle of two frame differences is exactly by front and back two field pictures for pixel
Gray value subtract each other, ambient brightness variation less in the case where, if corresponding pixel grey scale differ very little, it is believed that this
It is static for locating scenery, if the gray-value variation in image-region somewhere is very big, it is believed that since moving object is drawn in image
It rises, these zone markers is got up, the pixel region marked using these, so that it may find out the position of moving target in the picture
It sets.
4. shuttlecock drop point site prediction technique based on computer vision according to claim 1, which is characterized in that step
Binary conversion treatment, image segmentation are carried out to difference image described in rapid 5, Contour extraction extracts the feature letter of shuttlecock in turn
Breath, specific steps include:
Step 5.1, the difference image obtained for step 4 needs to extract the motor area of sportsman and shuttlecock from background
Domain, by binary conversion treatment and image Segmentation Technology, existing Threshold sementation has side between fixed threshold method and maximum kind
Poor method；
Step 5.2, it is made of since image can be regarded as foreground and background, the gray scale interval of sportsman and shuttlecock and background
It differs greatly, target and background visually have very big contrast, since maximum variance between clusters are to a part of variance image
It can be partitioned into badminton region, Athletess region cannot be divided, so selection fixed threshold method carrys out segmented image；
Step 5.3, difference image is after binary conversion treatment, and the target area moved in a frame image has become one can be with
Then the region being connected measures the feature of each region and identifies shuttlecock and sportsman；
Step 5.4, for a frame bianry image, zone marker is carried out to existing Connected component, it is special according to the profile of shuttlecock
Sign, identifies the characteristic point on shuttlecock, may recognize that shuttlecock using Contour extraction.
5. shuttlecock drop point site prediction technique based on computer vision according to claim 1, which is characterized in that step
Determine that the flight path of shuttlecock, specific steps include: described in rapid 6 using track following algorithm
Step 6.1, certain matching criterior is established to match the center of gravity of the moving region between consecutive frame in sequence image,
To obtain the track of all moving objects；
Step 6.2, when not carrying out at matched with longest track, system turns to whether analysis longest track has shuttlecock
Motion profile, if longest track is the motion profile of shuttlecock, system extracts the flight path of shuttlecock；
Step 6.3, when the motion profile that longest track is not shuttlecock, system continues to match later data.
6. shuttlecock drop point site prediction technique based on computer vision according to claim 1, which is characterized in that step
Predict that specific steps include: using track drop point site of the Kalman filtering algorithm to shuttlecock described in rapid 7
Step 7.1, it first has to establish shuttlecock locus model, the kth moment is predicted according to the state position at -1 moment of shuttlecock kth
State position；
Step 7.2, the prediction error at -1 moment of shuttlecock kth is calculated, estimates the prediction error at kth moment；
Step 7.3, kalman gain and system optimal estimated value are calculated；
Step 7.4, system kth moment prediction error is calculated, nonlinear optimization processing is carried out, reduction system is predicted error, mentioned
High precision of prediction；
Step 7.5, the position at system output shuttlecock kth moment.
7. shuttlecock drop point site prediction technique based on computer vision according to claim 1, which is characterized in that step
The coordinate of drop point site is extracted described in rapid 8, using mapping relations mathematically, reality is converted to using perspective transform method
The coordinate in border shuttlecock place, specific steps include:
Step 8.1, according to step 7 Kalman filtering algorithm to the predicted position of track drop point, available shuttlecock is in two dimension
Coordinate position under the model of place；
Step 8.2, it is converted using geometrical perspective, its position is transformed under world coordinate system, can be obtained under actual life
Coordinate position.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810549488.1A CN109087328A (en) | 2018-05-31 | 2018-05-31 | Shuttlecock drop point site prediction technique based on computer vision |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810549488.1A CN109087328A (en) | 2018-05-31 | 2018-05-31 | Shuttlecock drop point site prediction technique based on computer vision |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109087328A true CN109087328A (en) | 2018-12-25 |
Family
ID=64839230
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810549488.1A Pending CN109087328A (en) | 2018-05-31 | 2018-05-31 | Shuttlecock drop point site prediction technique based on computer vision |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109087328A (en) |
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109876417A (en) * | 2019-04-06 | 2019-06-14 | 接楚添 | Shuttlecock service linesman's system based on Computer Vision Detection Technique |
| CN110135308A (en) * | 2019-04-30 | 2019-08-16 | 天津工业大学 | A kind of direct free kick type identification method based on video analysis |
| CN110433471A (en) * | 2019-08-13 | 2019-11-12 | 宋雅伟 | A kind of badminton track monitoring analysis system and method |
| CN110796019A (en) * | 2019-10-04 | 2020-02-14 | 上海淡竹体育科技有限公司 | Method and device for identifying and tracking spherical object in motion |
| CN110796085A (en) * | 2019-10-29 | 2020-02-14 | 新华智云科技有限公司 | Method for automatically distinguishing basketball goal segment AB team based on deep learning object detection algorithm |
| CN111369629A (en) * | 2019-12-27 | 2020-07-03 | 浙江万里学院 | Ball return trajectory prediction method based on binocular visual perception of swinging, shooting and hitting actions |
| CN112734834A (en) * | 2020-12-14 | 2021-04-30 | 清华大学深圳国际研究生院 | Badminton space coordinate detection method and system |
| CN112967318A (en) * | 2021-03-23 | 2021-06-15 | 武汉大学 | Method, device and equipment for analyzing fish behaviors in electric field |
| CN113082680A (en) * | 2021-04-09 | 2021-07-09 | 谢弘毅 | Automatic judgment system for ball out of bound of ball game based on machine vision |
| CN113744305A (en) * | 2021-01-20 | 2021-12-03 | 北京京东乾石科技有限公司 | Target detection method, target detection device, electronic equipment and computer storage medium |
| CN113808167A (en) * | 2021-09-17 | 2021-12-17 | 北京赛搏体育科技股份有限公司 | Volleyball movement track extraction method based on video data |
| CN114419737A (en) * | 2022-03-29 | 2022-04-29 | 广州青莲网络科技有限公司 | Bow and arrow drop point prediction method and system, electronic device and medium |
| CN114494356A (en) * | 2022-04-02 | 2022-05-13 | 中傲数据技术(深圳)有限公司 | Badminton video clip processing method and system based on artificial intelligence |
| CN114522410A (en) * | 2022-02-14 | 2022-05-24 | 复旦大学 | Badminton net passing height detection method |
| CN115019206A (en) * | 2022-06-13 | 2022-09-06 | 北京拙河科技有限公司 | Airplane landing environment safety detection method and system |
| CN115114466A (en) * | 2022-08-30 | 2022-09-27 | 成都实时技术股份有限公司 | Method, system, medium and electronic device for searching target information image |
| CN115414656A (en) * | 2022-08-29 | 2022-12-02 | 江苏大学 | Eagle eye challenge system for badminton amateurs |
| CN116629844A (en) * | 2023-07-26 | 2023-08-22 | 山东碧汀智能科技有限公司 | Automatic dispatch method and system for operation and maintenance of drinking water equipment |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20070155502A1 (en) * | 2005-12-16 | 2007-07-05 | Pixart Imaging Inc. | Device for motion tracking and object for reflecting infrared light |
| CN201590108U (en) * | 2010-01-13 | 2010-09-22 | 北京视博数字电视科技有限公司 | Multi-target processing device and terminal |
| CN104504694A (en) * | 2014-12-16 | 2015-04-08 | 成都体育学院 | Method for acquiring three-dimensional information of moving ball |
| CN104966304A (en) * | 2015-06-08 | 2015-10-07 | 深圳市赛为智能股份有限公司 | Kalman filtering and nonparametric background model-based multi-target detection tracking method |
| CN105477845A (en) * | 2015-06-17 | 2016-04-13 | 电子科技大学 | Method and system for predicting motion trajectory of shuttlecock |
| CN107729295A (en) * | 2017-10-19 | 2018-02-23 | 广东工业大学 | Real-time drop point pre-judging method, platform and the equipment of a kind of shuttlecock |
-
2018
- 2018-05-31 CN CN201810549488.1A patent/CN109087328A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20070155502A1 (en) * | 2005-12-16 | 2007-07-05 | Pixart Imaging Inc. | Device for motion tracking and object for reflecting infrared light |
| CN201590108U (en) * | 2010-01-13 | 2010-09-22 | 北京视博数字电视科技有限公司 | Multi-target processing device and terminal |
| CN104504694A (en) * | 2014-12-16 | 2015-04-08 | 成都体育学院 | Method for acquiring three-dimensional information of moving ball |
| CN104966304A (en) * | 2015-06-08 | 2015-10-07 | 深圳市赛为智能股份有限公司 | Kalman filtering and nonparametric background model-based multi-target detection tracking method |
| CN105477845A (en) * | 2015-06-17 | 2016-04-13 | 电子科技大学 | Method and system for predicting motion trajectory of shuttlecock |
| CN107729295A (en) * | 2017-10-19 | 2018-02-23 | 广东工业大学 | Real-time drop point pre-judging method, platform and the equipment of a kind of shuttlecock |
Cited By (28)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109876417A (en) * | 2019-04-06 | 2019-06-14 | 接楚添 | Shuttlecock service linesman's system based on Computer Vision Detection Technique |
| CN110135308A (en) * | 2019-04-30 | 2019-08-16 | 天津工业大学 | A kind of direct free kick type identification method based on video analysis |
| CN110433471A (en) * | 2019-08-13 | 2019-11-12 | 宋雅伟 | A kind of badminton track monitoring analysis system and method |
| CN110796019A (en) * | 2019-10-04 | 2020-02-14 | 上海淡竹体育科技有限公司 | Method and device for identifying and tracking spherical object in motion |
| CN110796085B (en) * | 2019-10-29 | 2022-04-22 | 新华智云科技有限公司 | Method for automatically distinguishing basketball goal segment AB team based on deep learning object detection algorithm |
| CN110796085A (en) * | 2019-10-29 | 2020-02-14 | 新华智云科技有限公司 | Method for automatically distinguishing basketball goal segment AB team based on deep learning object detection algorithm |
| CN111369629A (en) * | 2019-12-27 | 2020-07-03 | 浙江万里学院 | Ball return trajectory prediction method based on binocular visual perception of swinging, shooting and hitting actions |
| CN112734834A (en) * | 2020-12-14 | 2021-04-30 | 清华大学深圳国际研究生院 | Badminton space coordinate detection method and system |
| CN112734834B (en) * | 2020-12-14 | 2023-07-07 | 清华大学深圳国际研究生院 | Badminton space coordinate detection method and system |
| CN113744305B (en) * | 2021-01-20 | 2023-12-05 | 北京京东乾石科技有限公司 | Target object detection method, target object detection device, electronic equipment and computer storage medium |
| CN113744305A (en) * | 2021-01-20 | 2021-12-03 | 北京京东乾石科技有限公司 | Target detection method, target detection device, electronic equipment and computer storage medium |
| CN112967318A (en) * | 2021-03-23 | 2021-06-15 | 武汉大学 | Method, device and equipment for analyzing fish behaviors in electric field |
| CN112967318B (en) * | 2021-03-23 | 2023-09-15 | 武汉大学 | Analysis method, device and equipment for fish behaviors in electric field |
| CN113082680A (en) * | 2021-04-09 | 2021-07-09 | 谢弘毅 | Automatic judgment system for ball out of bound of ball game based on machine vision |
| CN113808167A (en) * | 2021-09-17 | 2021-12-17 | 北京赛搏体育科技股份有限公司 | Volleyball movement track extraction method based on video data |
| CN113808167B (en) * | 2021-09-17 | 2023-11-17 | 北京赛搏体育科技股份有限公司 | Volleyball motion trail extraction method based on video data |
| CN114522410A (en) * | 2022-02-14 | 2022-05-24 | 复旦大学 | Badminton net passing height detection method |
| CN114419737B (en) * | 2022-03-29 | 2022-06-03 | 广州青莲网络科技有限公司 | Bow and arrow drop point prediction method and system, electronic device and medium |
| CN114419737A (en) * | 2022-03-29 | 2022-04-29 | 广州青莲网络科技有限公司 | Bow and arrow drop point prediction method and system, electronic device and medium |
| CN114494356B (en) * | 2022-04-02 | 2022-06-24 | 中傲数据技术(深圳)有限公司 | Badminton video clip processing method and system based on artificial intelligence |
| CN114494356A (en) * | 2022-04-02 | 2022-05-13 | 中傲数据技术(深圳)有限公司 | Badminton video clip processing method and system based on artificial intelligence |
| CN115019206A (en) * | 2022-06-13 | 2022-09-06 | 北京拙河科技有限公司 | Airplane landing environment safety detection method and system |
| CN115019206B (en) * | 2022-06-13 | 2023-02-03 | 北京拙河科技有限公司 | Airplane landing environment safety detection method and system |
| CN115414656A (en) * | 2022-08-29 | 2022-12-02 | 江苏大学 | Eagle eye challenge system for badminton amateurs |
| CN115114466A (en) * | 2022-08-30 | 2022-09-27 | 成都实时技术股份有限公司 | Method, system, medium and electronic device for searching target information image |
| CN115114466B (en) * | 2022-08-30 | 2022-12-13 | 成都实时技术股份有限公司 | Method, system, medium and electronic device for searching target practice information image |
| CN116629844A (en) * | 2023-07-26 | 2023-08-22 | 山东碧汀智能科技有限公司 | Automatic dispatch method and system for operation and maintenance of drinking water equipment |
| CN116629844B (en) * | 2023-07-26 | 2023-09-26 | 山东碧汀智能科技有限公司 | Automatic dispatch method and system for operation and maintenance of drinking water equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109087328A (en) | Shuttlecock drop point site prediction technique based on computer vision | |
| CN103413120B (en) | Tracking based on object globality and locality identification | |
| CN102819749B (en) | A kind of football offside automatic discrimination system and method based on video analysis | |
| CN110264493B (en) | Method and device for tracking multiple target objects in motion state | |
| CN107067413B (en) | A kind of moving target detecting method of time-space domain statistical match local feature | |
| CN103824070B (en) | A kind of rapid pedestrian detection method based on computer vision | |
| CN109145708B (en) | Pedestrian flow statistical method based on RGB and D information fusion | |
| CN106780620A (en) | A kind of table tennis track identification positioning and tracking system and method | |
| US20090180669A1 (en) | Device, system and method for determining compliance with a positioning instruction by a figure in an image | |
| Albiol Colomer et al. | Who is who at different cameras: people re-identification using depth cameras | |
| CN104298964B (en) | A kind of human body behavior act method for quickly identifying and device | |
| CN104794737B (en) | A kind of depth information Auxiliary Particle Filter tracking | |
| CN105046206B (en) | Based on the pedestrian detection method and device for moving prior information in video | |
| CN109345568A (en) | Sports ground intelligent implementing method and system based on computer vision algorithms make | |
| Teachabarikiti et al. | Players tracking and ball detection for an automatic tennis video annotation | |
| CN109460764A (en) | A kind of satellite video ship monitoring method of combination brightness and improvement frame differential method | |
| CN109684919B (en) | Badminton service violation distinguishing method based on machine vision | |
| CN109359549A (en) | A kind of pedestrian detection method based on mixed Gaussian and HOG_LBP | |
| CN109708658A (en) | A kind of visual odometry method based on convolutional neural networks | |
| CN110021029A (en) | A kind of real-time dynamic registration method and storage medium suitable for RGBD-SLAM | |
| Zhang et al. | Real-time golf ball detection and tracking based on convolutional neural networks | |
| Naik et al. | DeepPlayer-track: player and referee tracking with jersey color recognition in soccer | |
| JP4465150B2 (en) | System and method for measuring relative position of an object with respect to a reference point | |
| CN113435336B (en) | Running intelligent timing system and method based on artificial intelligence | |
| Sokolova et al. | Human identification by gait from event-based camera |
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 | ||
| AD01 | Patent right deemed abandoned | ||
| AD01 | Patent right deemed abandoned |
Effective date of abandoning: 20220802 |