KR102466989B1 - Expert matching method and system based on similarity between gold swing data - Google Patents

Abstract

An expert matching method based on golf swing data according to an embodiment of the present invention includes acquiring golf swing images of a user group and an expert group by an image input module; calculating golf swing data based on motion of an object in the golf swing image by a golf swing data calculation module; and matching, by an expert matching module, the user and the expert based on a similarity between the user's golf swing data and the expert's golf swing data.

Description

Expert matching method and system based on similarity of golf swing data {EXPERT MATCHING METHOD AND SYSTEM BASED ON SIMILARITY BETWEEN GOLD SWING DATA}

The present invention relates to an expert matching method and system based on similarity of golf swing data.

With the recent increase in screen golf courses, the number of general golf players is also increasing, and accordingly, the demand for analyzing one's own golf swing posture and receiving training based on it is also gradually increasing.

Event-based motion capture equipment was used for conventional golf swing analysis, but the equipment was very expensive and there were very few people using it in Korea, so it was difficult for the general public to easily access and use such equipment.

In addition, even if the user's golf swing posture is analyzed with the corresponding equipment, the analyzed data is simply presented to the golf instructor, and the golf instructor simply performs training by referring to the data.

In general, from the point of view of a golf learner, when choosing a golf instructor or mentor, you must make a judgment based on only fragmentary information such as the instructor's experience and lesson price. Optimal lessons for each individual were not possible. Accordingly, even if a person wants to receive training by selecting a golf instructor having a swing posture similar to the golf swing posture of the user, it is difficult to find such an instructor.

Republic of Korea Patent Registration No. 10-1399655 Republic of Korea Patent Publication No. 10-2012-0132281

An embodiment of the present invention records an image of a user's golf swing with a video recording device that can be easily accessed in the surroundings, analyzes the user's golf swing posture based on this, and can be matched with a golf instructor similar to the user's golf swing posture. An object of the present invention is to provide an expert matching method and system based on the similarity of golf swing data.

On the other hand, the technical problems to be achieved in the present invention are not limited to the above-mentioned technical problems, and other technical problems not mentioned will be clear to those skilled in the art from the description below. You will be able to understand.

An expert matching method based on golf swing data according to an embodiment of the present invention includes acquiring golf swing images of a user group and an expert group by an image input module; calculating golf swing data based on motion of an object in the golf swing image by a golf swing data calculation module; and matching, by an expert matching module, the user and the expert based on a similarity between the user's golf swing data and the expert's golf swing data.

Step a) may include: a-1) acquiring golf swing images of the user group and the expert group by means of an image input unit; and a-2) generating a plurality of golf swing image frames from the golf swing image by an image frame acquiring unit.

Step b) may include: b-1) extracting a moving object from the golf swing image by a motion object extraction unit; b-2) calculating motion data including motion paths and motion strengths of each of the moving objects by a motion data calculation unit; b-3) calculating, by a first golf swing data calculation unit, a first golf swing vector that is golf swing data of the user based on motion data extracted from a golf swing image of the user; and b-4) calculating, by a second golf swing data calculation unit, a second golf swing vector, which is golf swing data of the expert, based on motion data extracted from the expert's golf swing image.

Step c) includes: c-1) generating user cluster data by clustering a plurality of first golf swing vectors and generating expert cluster data by clustering a plurality of second golf swing vectors by a cluster data generator; step; c-2) calculating, by a first normalizer, a first normalization vector based on the degree of dominance of the first golf swing vector in the user cluster data; c-3) calculating a second normalization vector based on the degree of dominance of the second golf swing vector in the expert cluster data by a second normalizer; c-4) comparing similarities between each of the first normalization vectors and each of the second normalization vectors by a similarity calculating unit; and c-5) matching the user and the expert based on the degree of similarity between the first normalization vector and the second normalization vector by an expert matching unit.

Step b-1) includes: b-1-1) generating successive difference image frames by calculating a pixel value difference between each golf swing image frame and a background image by a difference image frame generation unit; b-1-2) calculating, by a difference image generation unit, a pixel value difference between the successive difference image frames to generate a golf swing difference image; b-1-3) calculating, by an average pixel size calculation unit, an average pixel size of the golf swing difference image based on pixel values of the golf swing difference image; b-1-4) determining whether a moving object exists by comparing an average pixel size of the golf swing difference image with a set threshold value by a moving object presence determination unit; and b-1-5) extracting, by a motion object extraction unit, a moving object from the golf swing image based on a convolutional neural network.

Step b-2) includes: b-2-1) extracting, by a pixel value extraction unit, pixel values within a set area based on the previous position of each object among the pixel values of each golf swing image frame; ; b-2-2) moving, by an object position moving unit, the position of each object from the previous position to the coordinates of the center of gravity of the pixel values in the setting area; b-2-3) determining, by a convergence determination unit, whether the positional movement of each object converges based on the difference between the coordinates of the center of gravity and the previous position; b-2-4) repeating steps b-2-1) to b-2-3) if it is determined by the convergence determination unit that the movement of each object is not converged; b-2-5) tracking the motion path of each object according to the converged position of each object when it is determined that the position movement of each object is converged by the motion path tracking unit; and b-2-6) calculating, by a motion strength calculating unit, the motion strength of the object through a distance that each object has moved along the motion path for a set period of time.

A computer-readable recording medium according to an embodiment of the present invention is a recording medium on which a program for executing the golf swing data-based expert matching method by a computer is recorded.

An expert matching system based on golf swing data according to an embodiment of the present invention includes an image input module configured to acquire golf swing images of a user group and an expert group; a golf swing data calculation module configured to calculate golf swing data based on motion of an object in the golf swing image; and an expert matching module configured to match the user with the expert based on a similarity between the user's golf swing data and the expert's golf swing data.

The image input module may include: an image input unit configured to acquire golf swing images of a user group and an expert group; and an image frame acquisition unit configured to generate a plurality of golf swing image frames from the golf swing image.

The golf swing data calculation module may include: a motion object extractor configured to extract a moving object from the golf swing image; a motion data calculator configured to calculate motion data including motion paths and motion strengths of each of the moving objects; a first golf swing data calculation unit configured to calculate a first golf swing vector that is golf swing data of the user based on motion data extracted from a golf swing image of the user; and a second golf swing data calculation unit configured to calculate a second golf swing vector, which is golf swing data of the expert, based on motion data extracted from the expert's golf swing image.

The expert matching module includes: a cluster data generation unit configured to generate user cluster data by clustering a plurality of first golf swing vectors and to generate expert cluster data by clustering a plurality of second golf swing vectors; a first normalization unit configured to calculate a first normalization vector based on a degree of dominance of the first golf swing vector in the user cluster data; a second normalization unit configured to calculate a second normalization vector based on a degree of dominance of the second golf swing vector in the expert cluster data; a similarity calculator configured to compare similarities between each of the first normalization vectors and each of the second normalization vectors; and an expert matching unit configured to match a user and an expert based on a similarity between the first normalization vector and the second normalization vector.

An expert matching method and system based on the similarity of golf swing data according to an embodiment of the present invention records an image of a user's golf swing with an image recording device that is easily accessible and analyzes the posture of the user's golf swing based on this. and can be matched with a golf instructor similar to the user's golf swing posture.

On the other hand, the effects obtainable in the present invention are not limited to the effects mentioned above, and other effects not mentioned will be clearly understood by those skilled in the art from the description below. You will be able to.

1 is a block diagram of an expert matching system based on golf swing data according to an embodiment of the present invention.
2 is a flowchart of an expert matching method based on golf swing data according to an embodiment of the present invention.
3 is a configuration diagram of an image input module according to an embodiment of the present invention.
4 is a flowchart illustrating step S100 in more detail.
5 is a configuration diagram of a golf swing data calculation module according to an embodiment of the present invention.
6 is a flowchart illustrating step S200 in more detail.
7 is a configuration diagram of a motion object extraction unit according to an embodiment of the present invention.
8 is a flowchart illustrating step S210 in more detail.
9 is a configuration diagram of a motion data calculation unit according to an embodiment of the present invention.
10 to 12 are conceptual diagrams for explaining the operation of the motion data calculation unit.
13 is a flowchart illustrating step S220 in more detail.
14 is a configuration diagram of an expert matching module according to an embodiment of the present invention.
15 is a flowchart illustrating step S300 in more detail.

Other advantages and features of the present invention, and methods for achieving them, will become clear with reference to the embodiments described below in detail in conjunction with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below and may be implemented in various different forms, but only the present embodiments make the disclosure of the present invention complete, and the common knowledge in the art to which the present invention belongs It is provided to fully inform the holder of the scope of the invention, and the present invention is only defined by the scope of the claims.

Even if not defined, all terms (including technical or scientific terms) used herein have the same meaning as generally accepted by common technology in the prior art to which this invention belongs. Terms defined by general dictionaries may be interpreted to have the same meaning as they have in the related art and/or the text of the present application, and are not conceptualized or overly formalized, even if not expressly defined herein. won't

Terms used in this specification are for describing embodiments and are not intended to limit the present invention. In this specification, singular forms also include plural forms unless specifically stated otherwise in a phrase. As used in the specification, 'comprise' and/or various conjugations of this verb, such as 'comprise', 'comprising', 'comprising', 'comprising', etc., refer to a mentioned composition, ingredient, component, Steps, acts and/or elements do not preclude the presence or addition of one or more other compositions, ingredients, components, steps, acts and/or elements. In this specification, the term 'and/or' refers to each of the listed elements or various combinations thereof.

Meanwhile, terms such as '~unit', '~group', '~block', and '~module' used throughout this specification may mean a unit that processes at least one function or operation. For example, it can mean software, hardware components such as FPGAs or ASICs. However, '~ unit', '~ group', '~ block', '~ module', etc. are not meant to be limited to software or hardware. '~unit', '~group', '~block', '~module' may be configured to be in an addressable storage medium or configured to reproduce one or more processors.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings of this specification.

1 is a block diagram of an expert matching system 10 based on golf swing data according to an embodiment of the present invention. Referring to FIG. 1 , an expert matching system 10 based on golf swing data according to an embodiment of the present invention includes an image input module 100, a golf swing data calculation module 200, and an expert matching module 300. .

2 is a flowchart of an expert matching method S10 based on golf swing data according to an embodiment of the present invention. Referring to FIG. 2 , the golf swing data-based expert matching method (S10) according to an embodiment of the present invention includes acquiring golf swing images of a user group and an expert group by an image input module (S100), golf swing data Calculating golf swing data based on the motion of the object in the golf swing image by the calculation module (S200) and by the expert matching module, based on the similarity between the user's golf swing data and the expert's golf swing data Matching the user and the expert with each other (S300).

3 is a configuration diagram of the image input module 100 according to an embodiment of the present invention. Referring to FIG. 2 , the image input module 100 may include an image input unit 110 and an image frame acquisition unit 120 .

The image input module 100 may be configured to acquire golf swing images of a user group and an expert group. The user may be an amateur player or a general golf player, and the expert may be a professional golf player. Hereinafter, golf player is used as a term referring to amateur players, general golf players, and professional golf players.

The image input unit 110 may be configured to obtain an image of a user's or expert's golf swing. When obtaining an image of a user's or expert's golf swing, the image of the user's golf swing may be obtained through a recording function of a general image capture device including a digital camera. In addition, the golf swing image may be obtained by transmitting or downloading the golf swing image through an information communication network such as a video streaming site, a blog, or the Internet. The image frame acquisition unit 120 may be configured to generate a plurality of golf swing image frames based on the golf swing image acquired by the image input unit 110 .

4 is a flowchart illustrating step S100 in more detail. Referring to FIG. 4 , in step S100, an image input unit acquires golf swing images of a user group and an expert group (S110), and a plurality of golf swing image frames are acquired from the golf swing images by an image frame acquisition unit. A generating step (S120) may be included.

5 is a configuration diagram of a golf swing data calculation module 200 according to an embodiment of the present invention. Referring to FIG. 5 , the golf swing data calculation module 200 includes a motion object extraction unit 210, a motion data calculation unit 220, a first golf swing data calculation unit 230, and a second golf swing data calculation unit ( 240) may be included.

6 is a flowchart illustrating step S200 in more detail. Referring to FIG. 6 , step S200 includes a step of extracting a moving object from the golf swing image by a motion object extractor (S210), and a motion path and motion intensity of each moving object by a motion data calculator. Calculating motion data to be performed (S220); Calculating, by a first golf swing data calculator, a first golf swing vector that is the user's golf swing data based on the motion data extracted from the user's golf swing image. (S230) and calculating, by a second golf swing data calculation unit, a second golf swing vector that is golf swing data of the expert based on the motion data extracted from the golf swing image of the expert (S240). have.

7 is a configuration diagram of a motion object extractor 210 according to an embodiment of the present invention. Referring to FIG. 7 , the motion object extractor 210 may be configured to extract a moving object from a golf swing image. The motion object extractor 210 includes a difference image frame generation unit 211, a difference image generation unit 212, an average pixel size calculation unit 213, a motion object presence/absence determination unit 214, and a motion object extraction unit 215. ) may be included.

The difference image frame generating unit 211 may be configured to generate successive difference image frames by calculating a pixel value difference between each golf swing image frame and the background image. The difference image frame generating unit 211 may generate a difference image frame according to [Equation 1] below, for example.

[Formula 1]

P[F(t)] = P[I(t)] - P[B]

In [Equation 1], P[F(t)] is a difference image frame, P[I(t)] is a golf swing image frame, and P[B] is a background image frame.

The difference image generation unit 212 may be configured to generate a golf swing difference image by calculating a pixel value difference between successive difference image frames. The difference image generation unit 212 may generate a golf swing difference image according to, for example, the following [Equation 2].

[Formula 2]

ΔP[F(t,t+1)] = P[F(t)] - P[F(t+1)]

In [Equation 2], ΔP[F(t,t+1)] is a golf swing difference image, P[F(t)] is a difference image frame at time t, and P[F(t+1)] is is the difference image frame at time (t+1).

The average pixel size calculation unit 213 may be configured to calculate an average pixel size of the golf swing difference image based on pixel values of the golf swing difference image.

The average pixel size calculation unit 213 may calculate the average pixel size of the golf swing difference image according to, for example, the following [Equation 3].

[Formula 3]

│ΔP[F(t,t+1)]│ = │P[F(t)] - P[F(t+1)]│

In [Equation 3], │ΔP[F(t,t+1)]│ is the average pixel size of the golf swing difference image, P[F(t)] is the difference image frame at time t, P[F( t+1)] is a difference image frame at time (t+1).

The moving object existence determination unit 214 may be configured to compare an average pixel size of the golf swing difference image with a set threshold value to determine whether a moving object exists.

The motion object presence/absence determination unit 214 may determine whether a motion object exists according to [Equation 4] below, for example.

[Formula 4]

│ΔP[F(t,t+1)]│ = │P[F(t)] - P[F(t+1)]│ > Threshold

In [Equation 4], │ΔP[F(t,t+1)]│ is the average pixel size of the golf swing difference image, P[F(t)] is the difference image frame at time t, P[F (t+1)] is a difference image frame at time (t+1), and Threshold is a set threshold.

The moving object existence determination unit 214 determines that a moving object exists when the average pixel size of the golf swing difference image is greater than a set threshold, and determines that a moving object exists when the average pixel size of the golf swing difference image is less than the set threshold. It can be determined that it does not exist.

The moving object extracting unit 215 may be configured to classify a motion area, which is an area occupied by a moving object, in a golf swing image based on a convolutional neural network, and extract the moving object. The convolutional neural network may be configured to distinguish motion areas of each object including a golf player, a golf ball, and a golf club.

In more detail, the motion object extracting unit 215 determines whether each object (eg, a golf ball, golf club, etc.) or golf player determined from the golf swing image frames is flexibly connected for each golf swing image frame, and determines whether each golf ball and golf club, A region in which an object such as a golf player moves may be distinguished.

For example, a convolutional neural network performs object detection and classification by means of a YOLOv3 neural network, and can track an object with high accuracy and high speed without a separate hand-made filter. YOLOv3 neural network is a neural network that implements integrated recognition that performs object position and range calculation and object classification at once. It is a neural network capable of high-speed inference, which finds out where is, collects them, and finally finds out where the object exists in the entire image.

8 is a flowchart illustrating step S210 in more detail. Referring to FIG. 8 , step S210 is a step of generating successive difference image frames by calculating a pixel value difference between each golf swing image frame and a background image by a difference image frame generation unit (S211), to a difference image generation unit generating a golf swing difference image by calculating pixel value differences between successive difference image frames (S212), by an average pixel size calculating unit, the golf swing based on the pixel values of the golf swing difference image Calculating an average pixel size of the difference image (S213), comparing the average pixel size of the golf swing difference image with a set threshold value by a moving object presence determination unit to determine whether a moving object exists (S214) and extracting, by a moving object extracting unit, a moving object from the golf swing image based on a convolutional neural network (S215).

9 is a configuration diagram of the motion data calculation unit 220 according to an embodiment of the present invention. Referring to FIG. 9 , the motion data calculation unit 220 may be configured to calculate motion data including a motion path and motion intensity of each moving object. For example, motion data includes club pass (path), face angle, dynamic loft, attack angle, launch angle, club speed, carry, smash factor, golf ball path, golf ball speed, and golf ball spin for a golf player. can do.

The motion data calculation unit 220 may include a pixel value extraction unit 221, an object position movement unit 222, a convergence determination unit 223, a motion path tracking unit 224, and a motion intensity calculation unit 225. have.

10 to 12 are conceptual diagrams for explaining operations by the motion data calculation unit 220 . Referring to FIGS. 9 to 12 , the pixel value extraction unit 221 may be configured to extract pixel values in a set area based on the previous position of each object among pixel values of each golf swing image frame. In more detail, the pixel value extraction unit 221 extracts pixel values (x1,y1), (x2,y2), ..., (xn,yn) ( n is the number of pixels within a set radius) may be extracted.

The object position moving unit 222 may be configured to move the position of each object from the previous position to the coordinates of the center of gravity of the pixel values in the setting area. More specifically, the object position moving unit 222 moves the position of each object from the previous position 10 to the coordinates (Σxi/n, Σyi/n) of the center of gravity 20 of the pixel values in the setting area 15 (n is the number of pixels within the set radius, i is an integer from 1 to n in the order of pixels within the set area).

The convergence determination unit 223 may be configured to determine whether the positional movement of each object converges based on the difference value between the coordinates of the center of gravity and the previous position. In addition, the convergence determination unit 223 may be configured to repeat operations by the pixel value extraction unit 221 and the object position movement unit 222 when it is determined that the positional movement of each object has not converged.

More specifically, the convergence determination unit 223 determines whether the positional movement of each object converges based on the difference between the coordinates of the center of gravity 20 of the pixel values in the setting area 15 and the previous position 10 of each object. can judge In addition, the convergence determination unit 223 performs the process by the pixel value extraction unit 221 and the object position movement unit 222 until the position of each object converges when it is determined that the position movement of each object has not converged. can be repeated When the position of each object is moved to the center of gravity (reference numeral 20 in FIG. 10) of the pixel values in the setting area 15, the moved position (coordinates) returns to the previous position of each object (reference numeral 20' in FIG. 11). ) becomes

Therefore, the pixel value extracting unit 221 determines the pixel values in the setting area 25 (e.g., at the previous position) based on the previous position 20' of each object among the pixel values of each golf swing image frame. Data coming within a set radius) is extracted.

Next, the object position moving unit 222 moves the position of each object from the previous position 20' to the coordinates of the center of gravity 30 of the pixel values in the setting area 25.

In addition, the convergence determination unit 223 determines whether the positional movement of each object converges based on the difference between the coordinates of the center of gravity 30 of the pixel values in the setting area 25 and the previous position 20' of each object. do.

If the position of each object is moved to the center of gravity (reference numeral 30 in FIG. 11) of the pixel values in the setting area 25 again, the moved position (coordinates) is the previous position of each object (reference numeral 30' in FIG. 8). ) becomes

The convergence determination unit 223 determines, for example, the coordinates of the center of gravity 40 of the pixel values in the setting area 35 based on the previous position 30' of each object and the previous position 30' of each object. If the difference value (distance) (D) between the objects is less than the reference distance, it may be determined that the positional movement of each object converges.

The motion path tracking unit 224 may be configured to track the motion path of each object according to the converged position of each object when it is determined that the positional movement of each object converges.

The motion intensity calculating unit 225 may be configured to calculate the motion intensity of the object through a distance that each object moves along the motion path for a set period of time. For example, when an image is acquired by a digital camera, a set time interval is provided between each frame according to the type of the digital camera. Therefore, by calculating the movement distance of the golf club head over the set number of frames, the speed of the golf club head, that is, the movement intensity, can be known.

For example, a first golf swing vector for a specific user A may be calculated based on motion data by analyzing a golf swing image through the golf swing data calculation module 200 . At this time, the first golf swing vector of user A is the club path (path), face angle, dynamic loft, attack angle, launch angle, club speed, carry, smash factor, golf ball path, golf ball speed, and golf ball spin amount of the vector. components may be included.

13 is a flowchart of step S220. Referring to FIG. 13 , in step S220 , a step of extracting, by a pixel value extraction unit, pixel values within a set area based on the previous position of each object among the pixel values of each golf swing image frame (S221), the object moving the position of each object from the previous position to the coordinates of the center of gravity of the pixel values in the setting area by a position moving unit (S222); by a convergence determination unit, the coordinates of the center of gravity and the previous position (S223) determining whether the positional movement of each object converges based on the difference value between the above steps. Step (S224), if it is determined that the position movement of each object is converged by the motion path tracking unit, tracking the motion path of each object according to the converged position of each object (S225) and the motion intensity Calculating, by a calculation unit, the motion intensity of each object through a distance that each object has moved along the motion path for a set time period ( S226 ).

14 is a configuration diagram of an expert matching module 300 according to an embodiment of the present invention. Referring to FIG. 14 , the expert matching module 300 may be configured to match a user and an expert based on a similarity between the user's golf swing data and the expert's golf swing data. The expert matching module 300 may include a cluster data generator 310, a first normalizer 320, a second normalizer 330, a similarity calculator 340, and an expert matcher 350.

The cluster data generator 310 may be configured to generate user cluster data by clustering a plurality of first golf swing vectors and expert cluster data by clustering a plurality of second golf swing vectors. That is, the cluster data generator 310 may generate user cluster data including all first golf swing vectors calculated from each of a plurality of users. In addition, expert cluster data including all second golf swing vectors calculated from each of a plurality of experts may be generated.

The first normalization unit 320 may be configured to calculate a first normalization vector based on the degree of dominance of the first golf swing vector in the user cluster data.

For example, club speed and golf ball speed information may be included as components of the first golf swing vector, and in this case, user cluster data may include club speed and golf ball speed information of a plurality of users. Since the club speeds and golf ball speeds of a plurality of users will have a normal distribution, the first normalization unit uses a normalization vector based on the superiority of club speeds and golf ball speeds of a specific user in the normal distribution for club speeds and golf ball speeds. can be calculated. At this time, the superiority can be calculated based on the upper percentage of the club speed and golf ball speed of a specific user in the normal distribution, and it will be obvious that the higher the superiority is calculated, the higher the normal distribution.

In other words, if the club speed and golf ball speed are components of the first golf swing vector, the club speed of a particular user is in the top 20%, and the golf ball speed is in the top 30%, the first golf swing vector is (club speed, golf ball velocity), and the first normalization vector can be calculated as (0.8, 0.7).

The second normalization unit 330 may be configured to calculate a second normalization vector based on the degree of dominance of the second golf swing vector in expert cluster data. As described in the first normalization unit 320, the second normalization unit 330 may calculate a second normalization vector for a specific expert.

For example, club speed may be included as a component of the first golf swing vector and the second golf swing vector. Assume that user A's club speed is 100 km/h and expert B's club speed is 140 km/h. The club speed of user A corresponds to the top 10% in the user group, so 0.9 can be calculated when converted into the first normalization vector. In addition, even though expert B's club speed is faster than user A's club speed, it corresponds to the top 20% in the expert cluster, so 0.8 can be calculated when converted into the second normalization vector.

The similarity calculator 340 may be configured to compare similarities between each of the first normalization vectors and each of the second normalization vectors. More specifically, the similarity calculation unit 340 may compare the similarity by calculating cosine similarity between the first normalization vector and the second normalization vector. Cosine similarity can be calculated by [Equation 5] below.

[Formula 5]

In this case, A and B are vector components included in the first normalization vector and the second normalization vector.

For example, club speed and flight distance may be included as components of the first normalization vector and the second normalization vector, and the similarity between the respective normalization vectors may be compared by calculating the cosine similarity between them.

For example, if the first normalization vector of user A is (0.8, 0.4), the second normalization vector of expert B is (0.5, 0.7), and the second normalization vector of expert C is (0.7, 0.5). , When calculating by substituting each vector component into [Equation 5], it can be seen that the similarity between user A and expert C is higher than that between user A and expert B.

The expert matching unit 350 may be configured to match a user and an expert based on a similarity between the first normalization vector and the second normalization vector. For example, in the above case, user A may be matched with expert C instead of expert B.

15 is a flowchart of step S300. Referring to FIG. 15 , in step S300, a plurality of first golf swing vectors are clustered to generate user cluster golf swing vectors, and a plurality of second golf swing vectors are clustered by the cluster data generator to generate expert cluster golf swing vectors. generating (S310), calculating, by a first normalizer, a first normalized vector based on the degree of dominance of the first golf swing vector within the user cluster golf swing vector (S320); Calculating, by a normalization unit, a second normalization vector based on the degree of dominance of the second golf swing vector within the expert group golf swing vector (S330); Comparing the degree of similarity between the vector and each second normalization vector (S340) and matching the user and the expert based on the degree of similarity between the first normalization vector and the second normalization vector by an expert matching unit (S350). ) may be included.

An expert matching method and system based on the similarity of golf swing data according to an embodiment of the present invention records an image of a user's golf swing with an image recording device that can be easily accessed in the vicinity, and analyzes the posture of the user's golf swing based on this. and can be matched with a golf instructor similar to the user's golf swing posture.

Although the present invention has been described through examples above, the above examples are only for explaining the idea of the present invention and are not limited thereto. A person skilled in the art will understand that various modifications can be made to the above-described embodiments. The scope of the present invention is defined only through the interpretation of the appended claims.

10: Expert matching system based on golf swing data
100: video input module 110: video input unit
120: image frame acquisition unit 200: golf swing data calculation module
210: motion object extraction unit 211: difference image frame generation unit
212 difference image generation unit 213 average pixel size calculation unit
214: moving object existence determination unit
215: motion object extraction unit 220: motion data calculation unit
221: pixel value extraction unit 222: object position movement unit
223 convergence determination unit 224 movement path tracking unit
225: motion intensity calculation unit 230: first golf swing data calculation unit
240: second golf swing data calculator
300: expert matching module 310: cluster data generator
320: first normalization unit 330: second normalization unit
340: similarity calculation unit 350: expert matching unit

Claims (11)

a) acquiring golf swing images of a user group and an expert group by an image input module;
b) calculating golf swing data based on motion of an object in the golf swing image by a golf swing data calculation module; and
c) matching, by an expert matching module, the user and the expert based on a similarity between the user's golf swing data and the expert's golf swing data;
Step b) is:
b-1) extracting a moving object from the golf swing image by a motion object extractor; and
b-2) calculating motion data including motion paths and motion strengths of each of the moving objects by a motion data calculation unit;
Step b-1) is:
b-1-1) calculating, by a difference image frame generating unit, a pixel value difference between each golf swing image frame and a background image to generate successive difference image frames;
b-1-2) calculating, by a difference image generation unit, a pixel value difference between the successive difference image frames to generate a golf swing difference image;
b-1-3) calculating, by an average pixel size calculation unit, an average pixel size of the golf swing difference image based on pixel values of the golf swing difference image;
b-1-4) determining whether a moving object exists by comparing an average pixel size of the golf swing difference image with a set threshold value by a moving object presence determination unit; and
b-1-5) extracting, by a motion object extraction unit, a moving object from the golf swing image based on a convolutional neural network;
Step b-2) is:
b-2-1) extracting, by a pixel value extraction unit, pixel values within a set area based on the previous position of each object among the pixel values of each golf swing image frame;
b-2-2) moving, by an object position moving unit, the position of each object from the previous position to the coordinates of the center of gravity of the pixel values in the setting area;
b-2-3) determining, by a convergence determination unit, whether the positional movement of each object converges based on the difference between the coordinates of the center of gravity and the previous position;
b-2-4) repeating steps b-2-1) to b-2-3) if it is determined by the convergence determination unit that the movement of each object is not converged;
b-2-5) tracking the motion path of each object according to the converged position of each object when it is determined that the position movement of each object is converged by the motion path tracking unit; and
b-2-6) calculating, by a motion strength calculation unit, the motion strength of each object through a distance that each object has moved along the motion path during a set period of time. According to claim 1,
Step a) is:
a-1) acquiring golf swing images of a user group and an expert group by means of an image input unit; and
a-2) Golf swing data-based expert matching method comprising generating a plurality of golf swing image frames from the golf swing image by an image frame acquisition unit. According to claim 2,
Step b) is:
b-3) calculating, by a first golf swing data calculation unit, a first golf swing vector that is golf swing data of the user based on motion data extracted from a golf swing image of the user; and
b-4) calculating, by a second golf swing data calculation unit, a second golf swing vector, which is golf swing data of the expert, based on motion data extracted from the golf swing image of the expert; Expert Matching Method. According to claim 3,
Step c) is:
c-1) generating user cluster data by clustering a plurality of first golf swing vectors, and generating expert data by clustering a plurality of second golf swing vectors, by a cluster data generator;
c-2) calculating, by a first normalizer, a first normalization vector based on the degree of dominance of the first golf swing vector in the user cluster data;
c-3) calculating a second normalization vector based on the degree of dominance of the second golf swing vector in the expert cluster data by a second normalizer;
c-4) comparing similarities between each of the first normalization vectors and each of the second normalization vectors by a similarity calculating unit; and
c-5) matching a user and an expert based on a degree of similarity between the first normalization vector and the second normalization vector by an expert matching unit; and golf swing data-based expert matching. delete delete In the computer-readable recording medium,
A recording medium on which a program for executing the golf swing data-based expert matching method according to any one of claims 1 to 4 by a computer is recorded. an image input module configured to acquire golf swing images of a user group and an expert group;
a golf swing data calculation module configured to calculate golf swing data based on motion of an object in the golf swing image; and
an expert matching module configured to match the user and the expert with each other based on a similarity between the user's golf swing data and the expert's golf swing data;
The golf swing data calculation module:
a motion object extraction unit configured to extract a moving object from the golf swing image; and
a motion data calculation unit configured to calculate motion data including motion paths and motion strengths of each of the moving objects;
The motion object extraction unit:
a difference image frame generating unit, configured to generate successive difference image frames by calculating a pixel value difference between each golf swing image frame and a background image;
a difference image generation unit configured to generate a golf swing difference image by calculating a pixel value difference between the successive difference image frames;
an average pixel size calculating unit, configured to calculate an average pixel size of the golf swing difference image based on pixel values of the golf swing difference image;
a moving object existence determination unit configured to compare an average pixel size of the golf swing difference image with a set threshold value to determine whether a moving object exists; and
a motion object extraction unit configured to extract a moving object from the golf swing image based on a convolutional neural network;
The motion data calculator:
a pixel value extracting unit configured to extract pixel values in a set area based on a previous position of each object from among pixel values of each golf swing image frame;
an object position shifting unit configured to move the position of each object from the previous position to the coordinates of the center of gravity of the pixel values in the setting area;
a convergence determination unit configured to determine whether the positional movement of each object converges based on a difference value between the coordinates of the center of gravity and the previous position;
a motion path tracking unit configured to track a motion path of each object according to the converged position of each object when it is determined that the position movement of each object is converged by the convergence determination unit; and
and a motion strength calculation unit configured to calculate a motion strength of each object through a distance that each object has moved along the motion path for a set period of time. According to claim 8,
The video input module:
an image input unit configured to acquire golf swing images of a user group and an expert group; and
A golf swing data-based expert matching system comprising an image frame acquisition unit configured to generate a plurality of golf swing image frames from the golf swing image. According to claim 9,
The golf swing data calculation module:
a first golf swing data calculation unit configured to calculate a first golf swing vector that is golf swing data of the user based on motion data extracted from a golf swing image of the user; and
A golf swing data-based expert matching system comprising a second golf swing data calculation unit configured to calculate a second golf swing vector, which is golf swing data of the expert, based on motion data extracted from a golf swing image of the expert. According to claim 10,
The expert matching module:
a cluster data generation unit configured to generate user cluster data by clustering a plurality of first golf swing vectors and expert cluster data by clustering a plurality of second golf swing vectors;
a first normalization unit configured to calculate a first normalization vector based on a degree of dominance of the first golf swing vector within the user cluster golf swing vectors;
a second normalization unit configured to calculate a second normalization vector based on a degree of dominance of the second golf swing vector within the expert cluster golf swing vectors;
a similarity calculator configured to compare similarities between each of the first normalization vectors and each of the second normalization vectors; and
and an expert matching unit configured to match a user and an expert based on a similarity between the first normalization vector and the second normalization vector.

Images