KR101428922B1 - Golf swing classification apparatus and method using fuzzy system - Google Patents

Abstract

The present invention relates to a golf swing classification apparatus and to a method using a fuzzy system, wherein the apparatus includes: a body joint extraction module detecting body joint information from a sensor to extract a coordinate of both hands from the body joint information; a golf club detecting and head chasing module detecting a golf club and a golf club head on the basis of the coordinate of the both hands; and a golf swing motion classification module classifying golf swing motion by applying fuzzy logic so as to improve recognition ability for the golf swing motion. Thus, analysis with respect to sports motion in each section can be conducted in more detail.

Description

Technical Field [0001] The present invention relates to a golf swing classification apparatus and method using a fuzzy system,

The present invention relates to a golf swing classification technique, and more particularly, to a fuzzy system for classifying a golf swing motion using a fuzzy control method in terms of a position, a velocity, and a moving direction of a golf club head in a golf swing motion using a camera and a sensor The present invention relates to a golf swing sorting apparatus and method using the same.

Modern sports have been developed into sports science that incorporates science, and various analytical simulation systems for improving records have been developed. These analytical simulations have shown good results in sporting mega events like the Olympics. As described above, since the evaluation and analysis of the sports motion with the human eyes can be greatly changed according to the subject of the human being, the technique of analyzing the sports image using the camera is widely utilized.

However, in Korea, a high speed camera is used to shoot the players' movements at the training fields and stadiums of the athletes, and the attitude of the athletes is analyzed offline by various manual operations of the experts. In other words, in order to provide fast feedback to the athletes in the field, it is required to develop a sports motion analysis technique for analyzing the content or posture of a game in real time. However, in the case of domestic, The technology is lagging behind because of its high dependency [1].

As a practical example, SPORTSCAD [2] in Germany has designed and used software to analyze motion dynamics extracted from users by constructing a motion analysis database of pre-collected athletes. In the UK, The Quintic System [3 ] Has developed and used software that analyzes sports behavior and biological and epidemiological information using Casio's high-speed camera.

As a representative sports analysis software in Korea, it is possible to select the technique of golf zone [5] as shown in FIG. 1 (a) and Dart Fish Korea [4] and FIG. 1 (b). Dartfish Korea has been providing special video broadcast services for major sports events around the world and presented a new methodology for analyzing various sports movements using digital image analysis technology at the 2004 Athens Olympics and the 2008 Beijing Olympics. In addition, the golf zone records the user's golf swing motion and reproduces the recorded video, so that the attitude of the golfer is corrected under the coaching of the expert.

On the other hand, sporting action has become an important field to be reported so far that articles on patenting are granted to specific actions and technologies for posture correction and learning as well as baseball pitching and golf swing motion. In addition, patent applications for cameras for sports motion analysis and video systems for posture teaching are also increasingly being studied, and studies on sports motion analysis are becoming more active. Among these sports, golf is the exercise that many people want to receive posture correction while placing their most important attitude. Golf swing motion posture correction is the most important part of the first step for swing, and it is classified into seven actions as address, take-back, top of swing, down swing, impact, follow-up, and finish [ 6].

Korean Patent Publication No. 10-2009-0084576 (published on August 5, 2009)

[1] Kim. Changwook, "Study on the Needs of Data Analysis ", The Korean Journal of Sports Science, vol. 20, No. 5, pp. 1331-1340, 2011 [2] http://www.sportscad.com/ [3] http://www.quintic.com/ [4] Jung. Sunil, "The effect of video analysis on the learning side of Taekwonde side kick", Dongeui Univ. Physics, Master Thesis, pp. 1-64, 2010 [5] Hwang. Saebom, "An Analysis of the User Interface in an Interactive Game ", KDK Journal, vol. 22, pp. 97-106, June, 2012 [6] Cho. Cangho, "Optimum Swing Action of Golf ", KCDC, vol. 7, No. 3, pp. 37-45, 2005

SUMMARY OF THE INVENTION Accordingly, the present invention has been made keeping in mind the above problems occurring in the prior art, and it is an object of the present invention to provide a golf club head capable of swinging a golf swing using a camera and a sensor, And a golf swing classifying apparatus and method using a fuzzy system that classifies a swing motion using a fuzzy control method according to a moving direction.

According to another aspect of the present invention, there is provided a golf swing classifying apparatus using a purge system, comprising: a body joint extracting module for detecting body joint information from a sensor and extracting two-handed coordinates from detected body joint information; A golf club detection and head tracking module for detecting a golf club and a golf club head based on the two-handed coordinates; And a golf swing motion classification module for classifying the golf swing motion by applying fuzzy logic to improve recognition performance for each golf swing motion.

The body joint extraction module may include an RGB (Red Green Blue) camera; An IR (InfraRed) projector capable of recognizing the depth of motion; An IR camera capable of recognizing reflected IRs; A controller for controlling the RGB camera, the IR projector, and the IR camera; A skeleton image generation unit for generating a skeleton image through an image generated from the RGB camera and the IR camera; And a body coordinate extraction unit for recognizing a human joint and extracting coordinates.

The golf club detection and head tracking module may further comprise: a difference image generation unit for generating a difference image using a difference between a current frame and a previous frame; An edge detector for detecting an edge portion from the difference image using a canyon edge; A linear transform unit for transforming the pixels into a straight line using Hough transform; A golf club detecting unit for detecting a golf club in the straight line converted; And a golf club head tracker for tracking and defining a golf club head as a far end point in both hands out of the detected both ends of the straight line.

The golf swing motion classifying module may classify the rule base using three language variables such as a position of the golf club head, a velocity of the golf club head, and a direction of the golf club head, A membership function and a rule base setting unit for selecting a membership function to characterize a fuzzy set for each language variable; A fuzzy logic execution unit for performing fuzzy logic according to the membership function and the rule base; And a golf swing motion classifying unit classifying the golf swing motion according to a result of performing the fuzzy logic.

Meanwhile, the golf swing classification method using the purge system of the present invention includes the steps of (a) detecting body joint information from a sensor and extracting two-handed coordinates from the detected body joint information; (b) detecting a golf club and a golf club head based on the two-handed coordinates; And (c) classifying the golf swing operation by applying fuzzy logic to improve recognition performance for each golf swing motion.

The step (b) includes: generating a difference image using a difference between a current frame and a previous frame; Detecting an edge portion from the difference image using a canyon edge; Converting the pixels to a straight line using Hough transform; Detecting a golf club in the straight line transformed; And tracking the far end point of the detected straight line as both ends of the golf club head, which is far from the two-handed coordinates.

The step (c) creates a rule base using three language variables such as the position of the golf club head, the velocity of the golf club head, and the direction of movement of the golf club head Selecting a membership function that characterizes a fuzzy set for each language variable; Performing fuzzy logic according to the membership function and the rule base; And classifying the golf swing operation according to the result of performing the purge logic.

As described above, according to the golf swing classifying apparatus and method using the purge system according to the present invention, the analysis of each section of the sports operation can be performed in more detail based on the proposed classification method.

1 shows an application example of a sports motion analysis technique.
2 is a configuration diagram of a golf swing classifying apparatus using the purge system of the present invention.
3 is a block diagram of a body joint extraction module of the present invention.
4 is a view showing body joint tracking information.
FIG. 5 is an image showing the result of extraction of a body joint including two-handed coordinates.
6 is an image showing the result of performing the difference image and the canyon edge.
7 is a block diagram of the golf club detection and head tracking module of the present invention.
8 is an image showing a golf club detection result.
9 is an image showing a result of tracking a golf club head.
10 is a block diagram of a golf swing motion classification module according to the present invention.
FIG. 11 is a view showing the golf club heads classified by sections.
12 is a diagram showing a membership function.
13 is a view showing a rule race.
14 is a diagram showing a result of classification of the golf swing motion.
15 is an example of deletion errors and replacement errors.

Hereinafter, a golf swing sorting apparatus and method using the purge system of the present invention will be described in detail with reference to the accompanying drawings.

2 is a configuration diagram of a golf swing classifying apparatus using the purge system of the present invention.

Referring to FIG. 2, the golf swing classifying apparatus using the purge system of the present invention mainly includes a body joint extracting module 1, a golf club detecting and head tracking module 2, and a golf swing motion classifying module 3 .

The body joint extraction module 1 detects body joint information from the sensor and extracts two-handed coordinates for detection of the golf club in the detected body joint information.

The golf club detection and head tracking module 2 detects a golf club and a golf club head using a Huff linear transformation algorithm based on two-handed coordinates.

The golf swing motion classification module 3 classifies the golf swing motion by applying fuzzy logic in order to reduce recognition error and improve recognition performance per motion.

In the golf swing classifier using the thus configured fuzzy system, the golf swing motion is classified into seven sections using a camera, a sensor, and a fuzzy system. , This information is detected from the golfer's joint information and color image information acquired from the camera and the sensor. As a result of the performance evaluation of the proposed method for the real - time golf swing image, the proposed method showed an average 85.2% confidence level of golf swing motion classification.

3 is a block diagram of a body joint extraction module of the present invention.

3, the body joint extraction module 1 of the present invention includes an RGB (Red Green Blue) camera 11 capable of recognizing colors, an IR (InfraRed) projector 12, an IR camera 13 capable of recognizing the reflected IR, a controller 14 for controlling the RGB camera 11, the IR projector 12, the IR camera 13, A skeleton image generating unit 15 for generating a skeleton image through the human body, and a body coordinate extracting unit 16 for recognizing 20 major human joints and extracting coordinates.

When the infrared ray is emitted from the IR projector 12, the body joint extraction module 1 configured as described above is reflected by the golfer's body and picked up by the IR camera 13, calculates the distance per pixel corresponding to the reflected IR, Processing software recognizes the location and distance of a person in real time. As shown in FIG. 4, the body joint extraction module 1 is applied to a joint tracking software capable of recognizing 20 major joints of a person. Based on the joint tracking, a skeleton image generation and a body coordinate extraction .

FIG. 5 is an image showing the result of extraction of a body joint including two-handed coordinates.

Referring to FIG. 5, coordinates of a joint are extracted using the body joint extraction module 1 of FIG. 3, and the coordinates of each joint are expressed by a skeleton image.

Here, the two-handed coordinates included in the coordinates of the joint are used as coordinates for detecting the golf club.

6 is a block diagram of the golf club detection and head tracking module of the present invention.

Referring to FIG. 6, the golf club detection and head tracking module 2 of the present invention includes a difference image generation unit 21 for generating a difference image using a difference between a current frame and a previous frame, An edge detecting unit 22 for detecting an edge portion using the Hough transform, a straight line converting unit 23 for converting pixels into a straight line by using Hough transform, a golf club detecting unit 24 for detecting a golf club in the straight line, , And a golf club head tracker (25) that defines both the hand and the far end of the detected straight line as both ends of the golf club head.

The operation of the thus constructed golf club detection and head tracking module 2 will be described in detail.

First, in the present invention, a differential image technique is used to extract a golf club moving in real time. The difference image technique is an algorithm used to detect motion that moves in real time using the difference between the current frame and the previous frame.

,

,

라고 정의할 때, 두 프레임의 차이

는 (식 1)과 같다.The present invention performs operations on three frames to detect moving golf clubs and to reduce motion blurring that occurs during a golf swing operation. The current frame and the two previous frames

,

,

, The difference between the two frames

Is the same as (Equation 1).

--- (식 1)

--- (1)

과

를 구하고, 두 영상에서 발생하는 잡음에 의한 원치 않은 분할을 최소한으로 줄이고자 닫힘 연산을 수행한다. 닫힘 연산을 수행한 결과

는 (식 2)와 같다. 아래의 식에서

는 3*3 사각형 커널,

는 침식연산,

는 팽창연산을 의미한다.Using equation (1) above,

and

And the closure operation is performed by minimizing the undesired division due to the noise generated in the two images. The result of performing a closed operation

Is expressed by (Equation 2). In the equation below

A 3 * 3 rectangular kernel,

Erosion operation,

Is an expansion operation.

--- (식 2)

--- (Equation 2)

과

를 구하고, 두 영상의 공통부분을 추출하여 모션 블러를 최소화하기 위한 교집합(intersection) 연산을 수행한다. 교집합 연산을 수행한 결과는 (식 3)과 같다.Using Equation 2 above,

and

And extract intersection of two images to perform intersection operation to minimize motion blur. The result of the intersection operation is shown in (Equation 3).

--- (식 3)

--- (Equation 3)

FIG. 7 shows an image obtained by using the above equations and an edge portion extracted using a Canny edge.

7A is an image showing a result of performing a difference operation, FIG. 7B is an image showing a result of performing a morphology operation, FIG. 7C is an image showing an intersection operation result, (D) is the image showing the canyon edge detection result.

좌표 평면 영상에서 모두 직선으로 표현하고, 직선들이 지나가는 위치에 있는 픽셀 값을 모두 축척하면,

좌표 평면에서의 직선은

좌표 평면에서는 국지적 최댓값을 갖는 형태로 나타난다. 내부 파라미터로는 근접한 이웃 픽셀까지의 직선 가정 파라미터와 직선 길이의 임계값을 적용하여 직선의 길이를 제한할 수 있는 파라미터, 그리고 허프 직선 변환의 종류를 구분하는 파라미터 등으로 구성되어 있다. 허프 직선 변환 과정을 거쳐 골프 클럽을 검출한 결과는 도 8에서 빨간색으로 표현하고 있다.
Next, a golf club is a swinging tool represented in a straight line in an image. In order to detect such a swing tool, a Huff linear transformation algorithm which shows the best performance in linear detection is used. The Hough linear transformation algorithm transforms non-zero pixels

If all of the pixel values at the positions where the straight lines pass are scaled,

The straight line in the coordinate plane

In the coordinate plane, the form has the local maximum value. The internal parameters consist of a hypothetical assumption of a straight line to a neighboring neighboring pixel and a parameter capable of limiting the length of the straight line by applying a threshold of a straight line length, and a parameter for classifying the type of the Hough linear transformation. The result of detecting the golf club through the Hough linear transformation process is shown in red in FIG.

축에 가까운 점을

,

축에 먼 점을

으로 반환하게 되는데, 본 발명에서 제안하는 방법은 두 점을 신체 관절 추출 모듈(1)로부터 얻은 양손 좌표와 비교하여 더 멀리 있는 점을 골프 클럽 헤드로 정의한다. 골프 클럽 헤드를 추적한 결과는 도 9에서 노란색으로 표현하고 있다.
On the other hand, in order to track the golf club head, information on both end points in the straight line detected by the Hough linear transformation is used. One straight line detected by the Hough linear transformation

The point near the axis

,

Distant points on the axis

In the method proposed by the present invention, the two points are compared with the two-hand coordinates obtained from the body joint extraction module 1 to define a farther point as a golf club head. The result of tracking the golf club head is shown in yellow in FIG.

10 is a block diagram of a golf swing motion classification module according to the present invention.

Referring to FIG. 10, the golf swing motion classification module 3 of the present invention includes three types of golf swing motion classification modules such as a position of a golf club head, a velocity of a golf club head, and a direction of a golf club head A membership function and a rule base setting unit 31 for selecting a membership function that characterizes a fuzzy set for each language variable, a rule base setting unit 31 for creating a rule base by using a language variable, a fuzzy logic performing unit And a golf swing motion classifying unit 33 for classifying the golf swing motion according to the result of performing the fuzzy logic.

The operation of the golf swing motion classifying module 3 configured as described above will be described in detail.

The goal of the golf swing motion is to use a golf club to send a golf ball in a stopped state to a desired point. In golf swing, there are kinds of chipping shot and bunker shot in addition to full swing, and golf swing basically means full swing. A full swing refers to a swing motion when a golf club is used to achieve a maximum distance to the golf ball. In golf, a full swing requires the right grip in the swing motion section and the grip in order to exert the best capability. In Table 1, the operation of golf is largely divided into 7 categories, and the definition of each segment is as follows.

[Table 1] is a classification of golf swing motion and its definition.

The fuzzy system is a knowledge-based or rule-based system whose core is the rule base called the fuzzy IF-THEN rule. A rule base is expressed using a linguistic variable similar to a human language, and a language variable is represented by a fuzzy set having a respective membership function. A fuzzy system allows us to deal with ambiguity by using a fuzzy set on a quantity that is difficult to distinguish clearly. The golf swing motion to be dealt with in the present invention can reduce the errors and uncertainties that may occur in the classification process by applying the fuzzy system because the motion swing motion is not constant and the boundaries for classifying motion are ambiguous.

The fuzzy system applied in the present invention has created rule bases by using three language variables such as a position of a golf club head, a velocity of a golf club head, and a direction of a golf club head, We select a membership function that characterizes the fuzzy set of variables.

The fuzzy set for the position of the golf club head, which is the first variable of the fuzzy system input, is divided into 8 directions in 360 占 direction as shown in Fig. 11 based on the center point of the body, and 8 And a membership function for each fuzzy set is selected as shown in FIG. 12 (a).

The second variable, golf club head speed, defines four fuzzy sets such as fast, moderately fast, moderate, and slow, and selects a membership function as shown in FIG. 12 (b). In this case, the crisp value of the club head speed at time t is calculated by using the following equation, the club head position change between t-1 frame and t frame.

--- (식 4)

--- (Equation 4)

는 클럽 헤드의 속도,

와

은

프레임과

프레임에서 골프 클럽 헤드의 좌표를 의미한다.(Equation 4)

The speed of the club head,

Wow

silver

Frame and

The coordinates of the golf club head in the frame.

The third variable, the direction of movement of the golf club head, also defines three fuzzy sets such as clockwise, counterclockwise, and no motion by sensing head position changes between t-1 and t frames, Is selected as shown in FIG. 12 (c).

Finally, the output of the fuzzy system is a language variable called a pose, and a fuzzy set for the seven operations in [Table 1] is defined, and each membership function is selected as shown in FIG. 12 (d).

The rule base of a fuzzy system consisting of three input variables and one output variable can define 8 * 4 * 3 = 96 IF-THEN rules. In the present invention, except for cases where the golf swing motion can not physically occur, such as "IF Position is", "and Velocity is", "and Direction is", and "THEN Pose is" . A typical rule base is as follows.

[Table 2] is a representative rule base.

13 is a view showing a rule race.

Referring to FIG. 13, as an example of the input to the defined rule base, the cusp value of the fuzzy system output is calculated as 0.501 when the values of the position, velocity, and direction are 0.723, 39.2, and 5, respectively.

Example

In order to test the methodology proposed in the present invention, data directly shot using a KINECT sensor was used. The image size is 640x480 and acquired 30 frames per second. For the golf swing data recording, the position of the Kinect sensor was placed on a desk about 70cm high and the golf swing position was taken from the front in a distance of about 2m in consideration of the sensor measuring distance range of the Kinect. We used MATLAB's Fuzzy toolbox and implemented it in the Microsoft Visual Studio 2010 environment using the Kinect SDK version 1.0.

The interface for swing motion classification is visualized by using the upper part of the image in a way that collects processed data in real time and informs the user of the current operation section. The swing motion interface is basically a yellow rectangle, and is filled with a red rectangle for the current golf swing motion.

14 is a diagram showing a result of classification of the golf swing motion.

Fig. 14A shows the address operation, Fig. 14B shows the take-back operation, Fig. 14C shows the top of swing operation, Fig. 14D shows the downward swing operation, (E) of FIG. 14 is a follow-through operation, FIG. 14 (f) is an impact operation, and FIG. 14 (g) is a finishing operation.

That is, FIG. 14 shows the results of the golf swing motion classification defined in [Table 1]. The experimental results track the golf club and the golf club head detected in the color image and display the current swing motion.

은 입력된 영상의 프레임 수,

는 올바르게 분류된 개수,

는 삭제 오류의 개수,

는 대체 오류의 개수,

는 삽입 오류의 개수,

은 분류 신뢰도를 의미한다.Three types of error and reliability were used to evaluate the golf swing motion classification performance of the proposed method. That is, three types of errors are deletion errors, replacement errors, and insertion errors. An erroneous error is an error that occurs when an existing golf swing motion in the input image can not be classified. Substitution errors are errors that occur when a motion classification is classified as an incorrect motion. An insertion error is an error that occurs when a nonexistent operation is detected to exist. Using the above three types of errors, the classification performance was evaluated as shown in (Equation 5), and the results were as shown in [Table 3]. Table 3

The number of frames of the input image,

Is a correctly categorized number,

The number of erase errors,

The number of replacement errors,

The number of insertion errors,

Means the classification reliability.

--- (식 5)

--- (Equation 5)

The results of the experiment using the fuzzy logic were calculated as (Equation 5) above, and the reliability of the golf swing motion was shown to be between 75% and 93%. The reliability was the highest in the beginning of the golf swing, Address, and the golf swing was the slowest in the down swing.

Table 3 shows the golf swing motion classification performance.

An error generated in the golf swing motion classification process is shown in FIG. FIG. 15 (a) shows an example of erroneous erasure when the current swing motion can not be classified, and this phenomenon can not be detected in the image blur generated during the rapid golf swing motion process and in the overlapping region of the background ≪ / RTI > FIG. 15 (b) is an example of a replacement error when the current golf swing motion is classified as another class. To eliminate this error, it is necessary to have an environment in which a Kinect camera and a high-speed camera of 120 fps or higher are interlocked.

In the present invention, a method of classifying real-time sports operation in an image by classifying the golf swing motion into 7 types using the color image and sensor data of the Kinect input in real time, Respectively.

Domestic sports analysis software analyzes the input image of the camera in most areas, but the system which analyzes it directly by the computer is not popularized. The present sports analysis is performed by a professional analyzing the camera image directly and feeding back the athletes to correct the wrong part. The software which can analyze the each section more in detail based on the classification method proposed in the present invention Is developed, it can be applied as a self-diagnostic program for semi-professional players.

The method of classifying the golf swing motion proposed in the present invention further includes a step of extracting the joint information and the kinematic data of the golf club that can be obtained from the swing of the golf professional to construct a database, It will be compared with the swing appearance of a professional player and the swing appearance of his swing, and it will be aimed at developing a learning type entertainment that can learn by oneself without a professional coaching process.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the invention is not limited to the disclosed exemplary embodiments, but, on the contrary, is intended to cover various modifications and equivalent arrangements included within the spirit and scope of the invention.

1: Body joint extraction module
2: Golf club detection and head tracking module
3: Golf Swing Motion Classification Module

Claims (7)

delete A body joint extraction module for detecting body joint information from the sensor and extracting two-hand coordinates from the detected body joint information;
A golf club detection and head tracking module for detecting a golf club and a golf club head based on the two-handed coordinates; And
And a golf swing motion classification module for classifying the golf swing motion by applying fuzzy logic to improve recognition performance of each golf swing motion,
Wherein the body joint extraction module comprises:
RGB (Red Green Blue) camera;
An IR (InfraRed) projector capable of recognizing the depth of motion;
An IR camera capable of recognizing reflected IRs;
A controller for controlling the RGB camera, the IR projector, and the IR camera;
A skeleton image generation unit for generating a skeleton image through an image generated from the RGB camera and the IR camera; And
And a body coordinate extracting unit for recognizing a human joint and extracting coordinates.
A body joint extraction module for detecting body joint information from the sensor and extracting two-hand coordinates from the detected body joint information;
A golf club detection and head tracking module for detecting a golf club and a golf club head based on the two-handed coordinates; And
And a golf swing motion classification module for classifying the golf swing motion by applying fuzzy logic to improve recognition performance of each golf swing motion,
Wherein the golf club detection and head tracking module comprises:
A difference image generation unit for generating a difference image using a difference between a current frame and a previous frame;
An edge detector for detecting an edge portion from the difference image using a canyon edge;
A linear transform unit for transforming the pixels into a straight line using Hough transform;
A golf club detecting unit for detecting a golf club in the straight line converted; And
And a golf club head tracing unit for defining and tracking the end points far from the two-handed coordinates as the golf club head among the both end points of the detected straight line.
A body joint extraction module for detecting body joint information from the sensor and extracting two-hand coordinates from the detected body joint information;
A golf club detection and head tracking module for detecting a golf club and a golf club head based on the two-handed coordinates; And
And a golf swing motion classification module for classifying the golf swing motion by applying fuzzy logic to improve recognition performance of each golf swing motion,
Wherein the golf swing motion classification module comprises:
A rule base is created using three language variables such as the position of the golf club head, the velocity of the golf club head, and the direction of movement of the golf club head, A membership function and a rule base setting unit for selecting a membership function characterizing the fuzzy set;
A fuzzy logic execution unit for performing fuzzy logic according to the membership function and the rule base; And
And a golf swing motion classifying unit classifying the golf swing motion according to a result of performing the fuzzy logic.
delete (a) detecting body joint information from the sensor and extracting two-handed coordinates from the detected body joint information;
(b) detecting a golf club and a golf club head based on the two-handed coordinates; And
(c) classifying the golf swing motion by applying fuzzy logic to improve recognition performance of each golf swing motion,
The step (b)
Generating a difference image using a difference between a current frame and a previous frame;
Detecting an edge portion from the difference image using a canyon edge;
Converting the pixels to a straight line using Hough transform;
Detecting a golf club in the straight line transformed; And
And tracing a far end of the detected straight line from both ends of the straight line to a golf club head.
(a) detecting body joint information from the sensor and extracting two-handed coordinates from the detected body joint information;
(b) detecting a golf club and a golf club head based on the two-handed coordinates; And
(c) classifying the golf swing motion by applying fuzzy logic to improve recognition performance of each golf swing motion,
The step (c)
A rule base is created using three language variables such as the position of the golf club head, the velocity of the golf club head, and the direction of movement of the golf club head, Selecting a membership function that characterizes the fuzzy set;
Performing fuzzy logic according to the membership function and the rule base; And
And classifying the golf swing operation according to a result of performing the fuzzy logic.

Images