CN112802051B - Fitting method and system of basketball shooting curve based on neural network - Google Patents
Fitting method and system of basketball shooting curve based on neural network Download PDFInfo
- Publication number
- CN112802051B CN112802051B CN202110142678.3A CN202110142678A CN112802051B CN 112802051 B CN112802051 B CN 112802051B CN 202110142678 A CN202110142678 A CN 202110142678A CN 112802051 B CN112802051 B CN 112802051B
- Authority
- CN
- China
- Prior art keywords
- fitting
- basketball
- identified
- shooting
- curve
- 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.)
- Active
Links
Images
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
- G06T7/70—Determining position or orientation of objects or cameras
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- 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
-
- 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
- G06T2207/30241—Trajectory
Abstract
The invention provides a fitting method and a system of a basketball shooting curve based on a neural network, wherein the method comprises the following steps: training the neural network model according to the training shooting video to obtain a basketball recognition model; inputting the shooting video to be identified into a basketball identification model, and identifying to obtain all basketball positions; acquiring a first preset number of fitting curves of the shooting video to be identified, calculating a fitting value of each fitting curve, and taking the fitting curve corresponding to the maximum fitting value as a basketball shooting curve; obtaining the fitted curve and the corresponding fitted value includes: randomly acquiring a second preset number of video frames to be identified from all video frames to be identified in the shooting video to be identified, and calculating according to the positions of basketballs in the acquired video frames to be identified and a parabolic calculation formula to obtain a fitting curve; and calculating the contact ratio between the basketball positions of all the video frames to be identified which are not acquired and the fitting curve, and taking the contact ratio as a fitting value. Has the beneficial effects that: and calculating the basketball shooting curve with higher accuracy.
Description
Technical Field
The invention relates to the technical field of exercise training assistance, in particular to a fitting method and system of a basketball shooting curve based on a neural network.
Background
Basketball training often needs a large amount of time and energy to put into, through the constant try and the repeated groping of sportsman form own shooting mode and shooting habit gradually, if the centre does not have standardized coach to guide, the sportsman often forms wrong motion habit to influence shooting accuracy and stability.
The current prior art adopts the track of tracking the basketball to compare the track of the basketball with the standard fitting curve, however, the above prior art acquires the track of the basketball through tracking, but the tracking algorithm for tracking the basketball cannot accurately detect all basketballs in the track of the basketball by 100%, so that the basketball track acquired in the prior art has the basketball which is easy to miss detection or false detection.
Disclosure of Invention
In view of the above problems in the prior art, a fitting method and system for a basketball shooting curve based on a neural network are provided to calculate a basketball shooting curve with high accuracy.
The specific technical scheme is as follows:
a fitting method of a basketball shooting curve based on a neural network comprises the following steps:
step S1, acquiring and extracting video frames of a plurality of training shooting videos, preprocessing the video frames, marking basketball positions, and inputting the video frames into a neural network model to train the neural network model to obtain a basketball recognition model;
step S2, acquiring shooting videos to be identified, and inputting each video frame to be identified of the shooting videos to be identified into a basketball identification model so as to identify and obtain basketball positions in all the video frames to be identified;
step S3, acquiring a first preset number of fitting curves of the shooting video to be identified, calculating a fitting value corresponding to each fitting curve, and taking the fitting curve corresponding to the maximum fitting value as a basketball shooting curve;
the specific steps of obtaining the fitting curve and the fitting value corresponding to the fitting curve each time comprise:
randomly acquiring a second preset number of video frames to be identified from all video frames to be identified in the shooting video to be identified, setting the acquired video frames to be identified as first video frames to be identified, and setting the video frames to be identified which are not acquired in the shooting video to be identified as second video frames to be identified;
calculating according to the basketball positions in all the first video frames to be identified and according to a parabolic calculation formula to obtain corresponding fitting curves;
and calculating the contact ratio between the basketball positions of all the second video frames to be recognized and the fitting curve, and taking the contact ratio as the fitting value of the fitting curve.
Preferably, the fitting method of the basketball shooting curve based on the neural network, wherein the step S1 specifically includes the following steps:
step S11, obtaining a plurality of training shooting videos, and saving at least one video frame of each training shooting video as a training picture to obtain a first training picture set comprising all training pictures;
step S12, marking the basketball position of each training picture in the first training picture set to obtain a second training picture set;
and step S13, inputting the second training picture set into the neural network model to train the neural network model so as to obtain the basketball recognition model.
Preferably, the fitting method of the basketball shooting curve based on the neural network specifically comprises the following steps of:
the method comprises the steps of acquiring a training shooting video at a fixed visual angle and a fixed position by adopting at least one image acquisition sensor, and enabling the overlapping rate of all video frames of the training shooting video to be within a preset overlapping rate.
Preferably, the fitting method of the basketball shooting curve based on the neural network, wherein the step of obtaining the fitting curve and the fitting value corresponding to the fitting curve each time specifically comprises the following steps:
step S31, a coordinate system corresponding to the shooting video to be identified is created;
step S32, randomly acquiring a second preset number of video frames to be identified from all video frames to be identified in the shooting video to be identified, setting the acquired video frames to be identified as first video frames to be identified, and setting the video frames to be identified which are not acquired in the shooting video to be identified as second video frames to be identified;
step S33, acquiring a first coordinate of the basketball position of each first video frame to be identified in the coordinate system;
step S34, inputting each first coordinate into a parabolic calculation formula to obtain a fitting curve;
step S35, acquiring a second coordinate of the basketball position of each second video frame to be identified in the coordinate system;
step S36, calculating a coordinate difference between each second coordinate and the fitting curve, taking the number of second coordinates corresponding to the coordinate difference within a preset difference range as a contact ratio, and taking the contact ratio as a fitting value of the fitting curve.
Preferably, the fitting method of the basketball shooting curve based on the neural network is characterized in that the parabola calculation formula is shown as the following formula:
y＝ax2+bx+c；
wherein y is used for representing the ordinate of the basketball in the coordinate system;
x is used for representing the abscissa of the basketball position in the coordinate system;
a. b, c are used to represent parabolic parameters.
Preferably, the method for fitting the basketball shooting curve based on the neural network is used, wherein the training shooting video comprises video frames from the beginning of shooting to the landing of the basketball.
Preferably, the fitting method of the basketball shooting curve based on the neural network is implemented, wherein the first preset number is greater than or equal to 2.
Preferably, the method for fitting a basketball shooting curve based on a neural network, wherein the first preset number is 20.
Preferably, the fitting method of the basketball shooting curve based on the neural network is implemented, wherein the second preset number is greater than or equal to 3.
Still provide a basketball shooting curve's fitting system based on neural network, wherein, include:
the model creating module is used for acquiring and inputting a plurality of training shooting videos into the neural network model after video frames are extracted from the training shooting videos and preprocessed and basketball position marks are input, so that the neural network model is trained to obtain a basketball recognition model;
the recognition module is connected with the model creation module and used for acquiring the shooting videos to be recognized and inputting each video frame to be recognized of the shooting videos to be recognized into the basketball recognition model so as to recognize and obtain the basketball position in all the video frames to be recognized;
the curve fitting module is connected with the identification module and comprises:
the shooting device comprises a fitting curve acquisition unit, a shooting unit and a recognition unit, wherein the fitting curve acquisition unit is used for randomly acquiring a second preset number of video frames to be recognized from all video frames to be recognized in the shooting video to be recognized, setting the acquired video frames to be recognized as first video frames to be recognized, and setting the video frames to be recognized which are not acquired in the shooting video to be recognized as second video frames to be recognized; calculating according to the basketball positions in all the first video frames to be identified and according to a parabolic calculation formula to obtain corresponding fitting curves; calculating the contact ratio between the basketball positions of all the second video frames to be recognized and the fitting curve, and taking the contact ratio as the fitting value of the fitting curve
And the basketball shooting curve acquisition unit is connected with the fitting curve acquisition unit and used for executing the fitting curve acquisition unit so as to acquire the fitting curves of the first preset number of the shooting videos to be identified and the fitting values corresponding to each fitting curve, and the fitting curve corresponding to the maximum fitting value is used as the basketball shooting curve.
The technical scheme has the following advantages or beneficial effects: taking the fitting curve corresponding to the maximum fitting value as a basketball shooting curve, so as to improve the precision of the basketball shooting curve, realize the subsequent fitting comparison of the basketball shooting curve with higher precision and a standard shooting curve, and further play a reference role in correcting and teaching actions of a shooting person; and the method can be used for acquiring the standard shooting curve with high precision, so that more accurate reference is provided for action correction teaching of a shooting player, the formation of wrong motion habits and the time and energy cost for correction are avoided, and the training efficiency is greatly improved.
Drawings
Embodiments of the present invention will now be described more fully hereinafter with reference to the accompanying drawings. The drawings are, however, to be regarded as illustrative and explanatory only and are not restrictive of the scope of the invention.
FIG. 1 is a parabolic plot of an embodiment of the fitting method of the present invention;
FIG. 2 is a flow chart of an embodiment of the fitting method of the present invention;
FIG. 3 is a schematic block diagram of an embodiment of a fitting system of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be obtained by a person skilled in the art without inventive efforts based on the embodiments of the present invention, shall fall within the scope of protection of the present invention.
It should be noted that the embodiments and features of the embodiments may be combined with each other without conflict.
The invention is further described with reference to the following drawings and specific examples, which are not intended to be limiting.
The invention comprises a fitting method of a basketball shooting curve based on a neural network, as shown in figure 2, comprising the following steps:
step S1, acquiring and extracting video frames of a plurality of training shooting videos, preprocessing the video frames, marking basketball positions, and inputting the video frames into a neural network model to train the neural network model to obtain a basketball recognition model;
step S2, acquiring shooting videos to be identified, and inputting each video frame to be identified of the shooting videos to be identified into a basketball identification model so as to identify and obtain basketball positions in all the video frames to be identified;
step S3, acquiring a first preset number of fitting curves of the shooting video to be identified, calculating a fitting value corresponding to each fitting curve, and taking the fitting curve corresponding to the maximum fitting value as a basketball shooting curve;
the specific steps of obtaining the fitting curve and the fitting value corresponding to the fitting curve each time comprise:
randomly acquiring a second preset number of video frames to be identified from all video frames to be identified in the shooting video to be identified, setting the acquired video frames to be identified as first video frames to be identified, and setting the video frames to be identified which are not acquired in the shooting video to be identified as second video frames to be identified;
calculating according to the basketball positions in all the first video frames to be identified and according to a parabolic calculation formula to obtain corresponding fitting curves;
and calculating the contact ratio between the basketball positions of all the second video frames to be recognized and the fitting curve, and taking the contact ratio as the fitting value of the fitting curve.
In the above embodiment, the basketball recognition model based on the neural network first detects the basketball position associated with the basketball in most video frames of the shooting video to be recognized, then fits the basketball shooting curve by using an automatic fitting method to obtain a plurality of fitting curves, calculates a fitting value corresponding to each fitting curve, and uses the fitting curve corresponding to the largest fitting value as the basketball shooting curve, thereby improving the accuracy of the basketball shooting curve, further reducing omission or false detection of the basketball position associated with the basketball in the basketball shooting curve, and comparing and analyzing the basketball shooting curve obtained during training with higher accuracy in real time with the standard basketball shooting curve with higher accuracy to feed back, timely corrects the shooting method of the shooter, and further improves the normalized shooting training effect.
In the above embodiment, the neural network model is trained through the step S1 to obtain a basketball recognition model, so that the step S2 is facilitated to input the shooting video to be recognized into the basketball recognition model, and a basketball position in the video frame to be recognized corresponding to each video frame of the shooting video to be recognized is recognized, that is, the tracking of a basketball track is realized;
it should be noted that the shooting curve of the basketball is a parabola, so that a fitting curve of the basketball when shooting can be calculated through the positions of the basketball with the second preset number;
the second preset number of first to-be-identified video frames are randomly acquired through the step S3, and the corresponding fitting curve is calculated according to the positions of basketballs in all the first to-be-identified video frames and the parabolic calculation formula, and because the fitting curve at this time is only calculated according to the positions of basketballs in the first to-be-identified video frames corresponding to any second preset number of video frames in the to-be-identified shooting video, a basketball which is missed or mistakenly detected exists in the fitting curve at this time, that is, the fitting curve at this time is not necessarily the best fitting curve;
in order to calculate the accuracy of the fitting curve, the coincidence degree of the fitting curve at the moment can be calculated according to the basketball position of the second to-be-recognized video frame which is not obtained in the shooting video to be recognized, and the coincidence degree is used as the fitting value of the fitting curve at the moment;
repeatedly executing the specific steps of obtaining the fitting curves and the fitting values corresponding to the fitting curves to obtain the fitting curves with the first preset number, taking the fitting curve corresponding to the maximum fitting value as a basketball shooting curve, and further improving the precision of the basketball shooting curve to realize the subsequent fitting comparison of the basketball shooting curve with higher precision and a standard shooting curve, so that the action correction teaching of a shooter is referred; and the method can be used for acquiring the standard shooting curve with high precision, so that more accurate reference is provided for action correction teaching of a shooting player, the formation of wrong motion habits and the time and energy cost for correction are avoided, and the training efficiency is greatly improved.
Further, in the above embodiment, step S1 specifically includes the following steps:
step S11, obtaining a plurality of training shooting videos, and saving at least one video frame of each training shooting video as a training picture to obtain a first training picture set comprising all training pictures;
step S12, marking the basketball position of each training picture in the first training picture set to obtain a second training picture set;
and step S13, inputting the second training picture set into the neural network model to train the neural network model so as to obtain the basketball recognition model.
In the above embodiment, the video shot during the basketball practice is collected as the training shooting video, wherein the shot video should cover the complete half field used during the practice;
where the full half field is used to represent:
the first video picture comprises a basketball shooting person and a basket at the same time, namely the basketball shooting person and the basket are in the same video picture;
secondly, the training shooting video comprises a video frame for shooting the basketball to start shooting and a video frame for finishing shooting, wherein the video frame for finishing shooting can be the moment when the basketball falls to the ground, and specifically can be defined by a user in a self-defining mode.
In step S12, the basketball position of each training picture may be manually labeled to obtain a second training picture set;
in step S13, the second training picture set is input into the neural network model to train the neural network model to obtain a basketball recognition model.
In a preferred embodiment, the neural network model is a convolutional neural network model, and the convolutional neural network model may be a convolutional neural network model such as YOLO (young only look once), EfficientDet, Mask RCNN, and the like.
In the above embodiment, the identification method in step S2 may specifically be: when the confidence coefficient of the basketball detected in the video frame to be identified is larger than a set threshold value, the video frame to be identified is determined to detect the basketball, and if the confidence coefficient of the detected basketball is smaller than the set threshold value, the video frame to be identified does not detect the basketball.
Further, in the above embodiment, the method for acquiring a training shooting video specifically includes the following steps:
the method comprises the steps of acquiring a training shooting video at a fixed visual angle and a fixed position by adopting at least one image acquisition sensor, and enabling the overlapping rate of all video frames of the training shooting video to be within a preset overlapping rate.
In the embodiment, the training shooting video is obtained through the fixed visual angle and the fixed position, so that other images except the basketball in all video frames of the training shooting video are almost consistent, for example, the background images can be consistent, the same coordinate system can be established for all video frames of the whole training shooting video, and then all positions of the basketball can be filled in the same coordinate system, so that subsequent curve fitting is facilitated.
In the above embodiment, in step S2, each to-be-recognized video frame of the to-be-recognized shooting video may be input into the basketball recognition model to recognize and obtain the basketball position in all to-be-recognized pictures.
Further, in the above embodiment, each time the fitting curve and the fitting value corresponding to the fitting curve are obtained, the method specifically includes the following steps:
step S31, a coordinate system corresponding to the shooting video to be identified is created;
step S32, randomly acquiring a second preset number of video frames to be identified from all video frames to be identified in the shooting video to be identified, setting the acquired video frames to be identified as first video frames to be identified, and setting the video frames to be identified which are not acquired in the shooting video to be identified as second video frames to be identified;
step S33, acquiring a first coordinate of the basketball position of each first video frame to be identified in the coordinate system;
step S34, inputting each first coordinate into a parabolic calculation formula to obtain a fitting curve;
step S35, acquiring a second coordinate of the basketball position of each second video frame to be identified in the coordinate system;
step S36, calculating a coordinate difference between each second coordinate and the fitting curve, taking the number of second coordinates corresponding to the coordinate difference within a preset difference range as a contact ratio, and taking the contact ratio as a fitting value of the fitting curve.
In the above embodiments, the steps of the corresponding methods are not necessarily performed in the order shown and described in the present specification in other embodiments. In some other embodiments, the methods may include more or fewer steps than those described herein. Moreover, a single step described in this specification may be broken down into multiple steps for description in other embodiments; multiple steps described in this specification may be combined into a single step in other embodiments. For example, step S35 in the above embodiment may be performed simultaneously with step S33.
Further, in the above embodiment, the parabolic calculation formula is as follows:
y＝ax2+bx+c；
wherein y is used for representing the ordinate of the basketball in the coordinate system;
x is used for representing the abscissa of the basketball position in the coordinate system;
a. b, c are used to represent parabolic parameters.
Further, in the above embodiment, the training shooting video includes video frames from the start of shooting to the landing of a basketball.
Further, in the above embodiment, the first preset number is greater than or equal to 2.
Further, in the above embodiment, the first preset number is 20.
Further, in the above embodiment, the second preset number is greater than or equal to 3.
In the above embodiment, since the shooting curve is a standard parabola, the parabola corresponding to the shooting curve can be determined by only 3 points.
As a preferred embodiment, as shown in fig. 1, a rectangular coordinate system may be established with the upper left corner of the video frame to be identified as the origin of coordinates, the horizontal direction of the video frame to be identified as the abscissa axis, and the vertical direction of the video frame to be identified as the ordinate axis, where the black point in fig. 1 is a corresponding coordinate point in the coordinate system for the basketball positions in all the video frames to be identified.
The positions of the basketballs in all the video frames to be identified have a corresponding coordinate point in the coordinate system, and any three points can determine a parabola. Randomly selecting 3 video frames to be identified (namely the first video frame to be identified) from all video frames to be identified to obtain coordinates of the position of the basketball, using the coordinates as three points on a parabola, and calculating three coefficients of the parabola to obtain a fitting curve corresponding to the parabola;
substituting the real abscissa corresponding to the basketball positions of all the second video frames to be recognized except the selected three first video frames to be recognized into the fitting curve to obtain the fitting ordinate corresponding to the real abscissa;
calculating a coordinate difference value between a real ordinate corresponding to the basketball position and the fitting ordinate, wherein if the coordinate difference value is within a preset difference value range, the contact ratio of the second video frame to be recognized and the fitting curve is 1; calculating to obtain the contact ratio of all the second video frames to be identified and the fitting curve, and taking the contact ratio as the fitting value of the fitting curve;
repeating the above steps, for example, 20 times, to obtain 20 fitting curves and corresponding fitting values, and selecting the fitting value with the largest fitting value from all the fitting curves as the basketball shooting curve.
There is also provided a fitting system for a basketball shooting curve based on a neural network, as shown in fig. 3, including:
the model creating module 1 is used for acquiring and extracting video frames of a plurality of training shooting videos, preprocessing the video frames, marking basketball positions and inputting the video frames into the neural network model so as to train the neural network model to obtain a basketball recognition model;
the recognition module 2 is connected with the model creation module 1 and used for acquiring shooting videos to be recognized and inputting each video frame to be recognized of the shooting videos to be recognized into the basketball recognition model so as to recognize and obtain basketball positions in all the video frames to be recognized;
the curve fitting module 3 is connected with the identification module 2 and comprises:
the fitting curve acquiring unit 31 is configured to randomly acquire a second preset number of to-be-identified video frames from all to-be-identified video frames in the to-be-identified shooting video, set the acquired to-be-identified video frames as first to-be-identified video frames, and set the to-be-identified video frames which are not acquired in the to-be-identified shooting video as second to-be-identified video frames; calculating according to the basketball positions in all the first video frames to be identified and according to a parabolic calculation formula to obtain corresponding fitting curves; calculating the contact ratio between the basketball positions of all the second video frames to be recognized and the fitting curve, and taking the contact ratio as the fitting value of the fitting curve
And the basketball shooting curve obtaining unit 32 is connected with the fitting curve obtaining unit 31 and used for executing the fitting curve obtaining unit so as to obtain the fitting curves of the first preset number of the shooting videos to be identified and the fitting value corresponding to each fitting curve, and taking the fitting curve corresponding to the maximum fitting value as the basketball shooting curve.
The specific implementation of the fitting system of the basketball shooting curve based on the neural network is basically the same as that of the fitting method of the basketball shooting curve based on the neural network, and the details are not repeated herein.
While the invention has been described with reference to a preferred embodiment, it will be understood by those skilled in the art that various changes in form and detail may be made therein without departing from the spirit and scope of the invention.
Claims (10)
1. A fitting method of a basketball shooting curve based on a neural network is characterized by comprising the following steps:
step S1, acquiring and extracting video frames of a plurality of training shooting videos, preprocessing the video frames, marking basketball positions, and inputting the video frames into a neural network model to train the neural network model to obtain a basketball recognition model;
step S2, acquiring shooting videos to be recognized, and inputting each video frame to be recognized of the shooting videos to be recognized into the basketball recognition model so as to recognize and obtain basketball positions in all the video frames to be recognized;
step S3, obtaining a first preset number of fitting curves of the shooting video to be identified, calculating a fitting value corresponding to each fitting curve, and taking the fitting curve corresponding to the largest fitting value as a basketball shooting curve;
the specific steps of obtaining the fitting curve and the fitting value corresponding to the fitting curve each time comprise:
randomly acquiring a second preset number of the video frames to be identified from all the video frames to be identified in the shooting video to be identified, setting the acquired video frames to be identified as first video frames to be identified, and setting the video frames to be identified which are not acquired in the shooting video to be identified as second video frames to be identified;
calculating according to the basketball positions in all the first video frames to be identified and according to a parabolic calculation formula to obtain corresponding fitting curves;
and calculating the contact ratio between the basketball positions of all the second video frames to be recognized and the fitting curve, and taking the contact ratio as the fitting value of the fitting curve.
2. The method for fitting a basketball shooting curve based on a neural network as claimed in claim 1, wherein said step S1 specifically comprises the following steps:
step S11, obtaining a plurality of training shooting videos, and saving at least one video frame of each training shooting video as a training picture to obtain a first training picture set comprising all the training pictures;
step S12, marking the basketball position of each training picture in the first training picture set to obtain a second training picture set;
step S13, inputting the second training picture set into the neural network model to train the neural network model, so as to obtain the basketball recognition model.
3. The method for fitting a basketball shooting curve based on a neural network as claimed in claim 1 or 2, wherein the method for obtaining the training shooting video specifically comprises the following steps:
acquiring the training shooting video at a fixed visual angle and a fixed position by adopting at least one image acquisition sensor, so that the overlapping rate of all video frames of the training shooting video is within a preset overlapping rate.
4. The method for fitting a basketball shooting curve based on a neural network as claimed in claim 1, wherein the step of obtaining the fitted curve and the fitted value corresponding to the fitted curve each time specifically comprises the following steps:
step S31, a coordinate system corresponding to the shooting video to be identified is created;
step S32, randomly acquiring a second preset number of the video frames to be identified in all the video frames to be identified in the shooting video to be identified, setting the acquired video frames to be identified as first video frames to be identified, and setting the video frames to be identified which are not acquired in the shooting video to be identified as second video frames to be identified;
step S33, acquiring a first coordinate of the basketball position of each first to-be-identified video frame in the coordinate system;
step S34, inputting each first coordinate into the parabolic calculation formula to obtain the fitting curve;
step S35, obtaining a second coordinate of the basketball position of each second video frame to be identified in the coordinate system;
step S36, calculating a coordinate difference between each second coordinate and the fitting curve, taking the number of the second coordinates corresponding to the coordinate difference within a preset difference range as the contact ratio, and taking the contact ratio as a fitting value of the fitting curve.
5. The method of claim 1 or 4, wherein the parabolic calculation formula is as follows:
y＝ax2+bx+c；
wherein y is used for representing the ordinate of the basketball in the coordinate system;
x is used for representing the abscissa of the basketball position in the coordinate system;
a. b, c are used to represent parabolic parameters.
6. The method of claim 1 or 2, wherein the training shooting video comprises video frames from the start of shooting to the landing of a basketball.
7. The method of fitting a neural network-based basketball shot curve as claimed in claim 1, wherein said first predetermined number is greater than or equal to 2.
8. The method of claim 7, wherein the first predetermined number is 20.
9. The method of claim 1, wherein the second predetermined number is greater than or equal to 3.
10. A fitting system of basketball shooting curves based on a neural network is characterized by comprising:
the model creating module is used for acquiring and inputting a plurality of training shooting videos into the neural network model after video frames are extracted from the training shooting videos and preprocessed and basketball position marks are input, so that the neural network model is trained to obtain a basketball recognition model;
the recognition module is connected with the model creation module and used for acquiring shooting videos to be recognized and inputting each video frame to be recognized of the shooting videos to be recognized into the basketball recognition model so as to recognize and obtain basketball positions in all the video frames to be recognized;
a curve fitting module connected with the identification module, comprising:
a fitting curve obtaining unit, configured to randomly obtain a second preset number of video frames to be identified from all the video frames to be identified in the shooting video to be identified, set the obtained video frames to be identified as first video frames to be identified, and set the video frames to be identified, which are not obtained in the shooting video to be identified, as second video frames to be identified; calculating according to the basketball positions in all the first video frames to be identified and according to a parabolic calculation formula to obtain corresponding fitting curves; calculating the contact ratio between the basketball positions of all the second video frames to be recognized and the fitting curve, and taking the contact ratio as the fitting value of the fitting curve;
and the basketball shooting curve obtaining unit is connected with the fitting curve obtaining unit and used for executing the fitting curve obtaining unit so as to obtain the fitting curves of the first preset number of the shooting videos to be identified and the fitting values corresponding to each fitting curve, and the fitting curve corresponding to the largest fitting value is used as the basketball shooting curve.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110142678.3A CN112802051B (en) | 2021-02-02 | 2021-02-02 | Fitting method and system of basketball shooting curve based on neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110142678.3A CN112802051B (en) | 2021-02-02 | 2021-02-02 | Fitting method and system of basketball shooting curve based on neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112802051A CN112802051A (en) | 2021-05-14 |
| CN112802051B true CN112802051B (en) | 2022-05-17 |
Family
ID=75813724
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110142678.3A Active CN112802051B (en) | 2021-02-02 | 2021-02-02 | Fitting method and system of basketball shooting curve based on neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112802051B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113295915A (en) * | 2021-07-27 | 2021-08-24 | 武汉中原电子信息有限公司 | Current calibration method and system for power grid collector and storage medium |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017143341A1 (en) * | 2016-02-19 | 2017-08-24 | Pillar Vision, Inc. | Systems and methods for monitoring objects at sporting events |
| WO2018085894A1 (en) * | 2016-11-10 | 2018-05-17 | Formalytics Holdings Pty Ltd | Measuring a property of a trajectory of a ball |
| CN109862949A (en) * | 2016-08-23 | 2019-06-07 | 沛勒尔维珍公司 | System and method for tracking rebounder's performance |
| CN110929596A (en) * | 2019-11-07 | 2020-03-27 | 河海大学 | Shooting training system and method based on smart phone and artificial intelligence |
| CN110942022A (en) * | 2019-11-25 | 2020-03-31 | 维沃移动通信有限公司 | Shooting data output method and electronic equipment |
| CN111626155A (en) * | 2020-05-14 | 2020-09-04 | 新华智云科技有限公司 | Basketball position point generation method and equipment |
| CN111866575A (en) * | 2019-04-24 | 2020-10-30 | 北京力动科技有限公司 | Real-time motion video intelligent capturing and feedback method and system |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8617008B2 (en) * | 2001-09-12 | 2013-12-31 | Pillar Vision, Inc. | Training devices for trajectory-based sports |
| WO2013025477A1 (en) * | 2011-08-12 | 2013-02-21 | Edh Holdings (South Africa) (Pty) Ltd. | Ball trajectory and bounce position detection |
| US11638854B2 (en) * | 2018-06-01 | 2023-05-02 | NEX Team, Inc. | Methods and systems for generating sports analytics with a mobile device |
-
2021
- 2021-02-02 CN CN202110142678.3A patent/CN112802051B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017143341A1 (en) * | 2016-02-19 | 2017-08-24 | Pillar Vision, Inc. | Systems and methods for monitoring objects at sporting events |
| CN109862949A (en) * | 2016-08-23 | 2019-06-07 | 沛勒尔维珍公司 | System and method for tracking rebounder's performance |
| WO2018085894A1 (en) * | 2016-11-10 | 2018-05-17 | Formalytics Holdings Pty Ltd | Measuring a property of a trajectory of a ball |
| CN111866575A (en) * | 2019-04-24 | 2020-10-30 | 北京力动科技有限公司 | Real-time motion video intelligent capturing and feedback method and system |
| CN110929596A (en) * | 2019-11-07 | 2020-03-27 | 河海大学 | Shooting training system and method based on smart phone and artificial intelligence |
| CN110942022A (en) * | 2019-11-25 | 2020-03-31 | 维沃移动通信有限公司 | Shooting data output method and electronic equipment |
| CN111626155A (en) * | 2020-05-14 | 2020-09-04 | 新华智云科技有限公司 | Basketball position point generation method and equipment |
Non-Patent Citations (1)
| Title |
|---|
| "基于神经网络的弹道曲线快速拟合方法";董雄风等;《弹箭与制导学报》;20140430;第34卷(第2期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112802051A (en) | 2021-05-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103413120B (en) | Tracking based on object globality and locality identification | |
| CN110728225B (en) | High-speed face searching method for attendance checking | |
| CN109919977B (en) | Video motion person tracking and identity recognition method based on time characteristics | |
| CN110674785A (en) | Multi-person posture analysis method based on human body key point tracking | |
| CN107909600A (en) | The unmanned plane real time kinematics target classification and detection method of a kind of view-based access control model | |
| CN112183355B (en) | Effluent height detection system and method based on binocular vision and deep learning | |
| CN107506693A (en) | Distort face image correcting method, device, computer equipment and storage medium | |
| CN109274883A (en) | Posture antidote, device, terminal and storage medium | |
| CN110298231A (en) | A kind of method and system determined for the goal of Basketball Match video | |
| CN109117753A (en) | Position recognition methods, device, terminal and storage medium | |
| CN106951826B (en) | Method for detecting human face and device | |
| CN110287907B (en) | Object detection method and device | |
| CN110021034A (en) | A kind of tracking recording broadcasting method and system based on head and shoulder detection | |
| CN107862240A (en) | A kind of face tracking methods of multi-cam collaboration | |
| CN110765814A (en) | Blackboard writing behavior recognition method and device and camera | |
| CN102831408A (en) | Human face recognition method | |
| CN112802051B (en) | Fitting method and system of basketball shooting curve based on neural network | |
| CN106447695A (en) | Same object determining method and device in multi-object tracking | |
| CN108960142A (en) | Pedestrian based on global characteristics loss function recognition methods again | |
| CN111709365A (en) | Automatic human motion posture detection method based on convolutional neural network | |
| CN114120188B (en) | Multi-row person tracking method based on joint global and local features | |
| CN104598914A (en) | Skin color detecting method and device | |
| CN114618142A (en) | Auxiliary training system and method for table tennis sports | |
| CN114639168B (en) | Method and system for recognizing running gesture | |
| CN114187664A (en) | Rope skipping counting system based on artificial intelligence |
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 |