KR102369359B1 - Image-based intelligent push-up discrimination method and system - Google Patents

Abstract

An image-based intelligent push-up discrimination method and push-up discrimination that can make the subject recognize abnormal push-ups by precisely measuring the push-up posture of the subject and discriminating the correct posture and incorrect posture It relates to a system, an image collection module that collects image data by photographing a push-up state of a subject, and 18 joints of the subject with an Openpose-based algorithm for the image data collected in the image collection module A pose estimation module for estimating position information on When the posture characteristic evaluation module evaluates to a normal push-up state, a configuration including a push-up counting module that counts the number of push-ups is provided, so that the push-up is accurately performed regardless of the body condition such as the height of the subject It can be measured and counted.

Description

Image-based intelligent push-up discrimination method and system

The present invention relates to an image-based intelligent push-up determination method and a push-up determination system, in particular, by precisely measuring the push-up posture of a subject and discriminating an upright posture and an incorrect posture to determine abnormal push-up It relates to an image-based intelligent push-up discrimination method and a push-up discrimination system that can allow a subject to recognize .

In general, push-ups are widely used as a test method for strengthening upper body muscle strength and evaluating physical fitness levels in fields such as schools, military, police, and firefighting.

Various instruments have been developed to measure these push-ups and have been used for entrance exams for sports-related colleges, military academies, police, firefighters, and military personnel for new recruitment tests or promotions.

Conventionally, the measurer directly performed the push-up measurement posture test and counting, but it was inconvenient for the measurer to check the measuring posture and count the number of times for each subject, and there was a problem in that the counting standard was not clear. That is, during the push-up test, a large number of measurers must be present to count and record the push-up score, and each measurer's counting criteria can be subjective, affecting the measurement results and fairness. Another problem was that it was difficult for the measurer to accurately determine the scores of multiple subjects at the same time.

To solve this problem, a device capable of automatically measuring the number of push-ups using a sensor such as an optical sensor, a pressure (contact) sensor, and an ultrasonic sensor has been developed.

An example of such a technique is disclosed in Documents 1 to 3 and the like.

For example, in Patent Document 1 below, a height adjustment rod provided on a pedestal to be inclined toward the user, a sensor rod that is inserted into the height adjustment rod to move up and down elastically, is provided on the upper end of the sensor rod so that the chest of the user is It includes a flat body in contact, but the height adjustment bar is provided with a detection sensor, and the sensor bar is provided with a sensing target sensor, so that the sensor bar moves downward and detects as the user does push-ups with the chest against the flat body Disclosed is a push-up measuring device capable of measuring the number of push-ups by crossing a target sensor and a detection sensor.

In addition, Patent Document 2 includes a hand support for supporting both hands for push-ups, and a measurement module for measuring the number of push-ups is installed at a position opposite to a user supporting both hands on the hand support. , the measuring module is a push-up measuring device comprising an ultrasonic measuring unit for measuring a distance to the user using ultrasound and a number measuring unit for performing counting the number of times for the push-ups according to the distance to the user has been disclosed for

On the other hand, in Patent Document 3 below, a case for accommodating an image recognition system that compares pixel values of a plurality of digital image data with each other and recognizes the difference therein, a guide rod coupled to the front of the case, and upper and lower portions of the guide rod A digital camera that is movably coupled in the direction, converts a measurement motion of a measurer into digital image data and provides the image data to the image recognition system, and is installed on one side of the upper surface of the case A measuring mat having a main body including an operation panel for controlling an operation, a measuring mat having a side facing the digital camera and being spaced apart from the front of the measuring device according to a photographing angle of the digital camera, both sides of the upper surface of the measuring mat Disclosed is an apparatus for measuring push-ups on the stomach using image data comparison including first and second fixing bars respectively coupled to and a gripping rod having both sides respectively coupled to the first and second fixing bars to be gripped by a measurer has been

Republic of Korea Patent Publication No. 10-2031878 (registered on October 7, 2019) Republic of Korea Patent Publication No. 10-1269561 (Registered on May 24, 2013) Republic of Korea Registered Utility Model Publication No. 20-0346479 (Registered on March 19, 2004)

In the technology disclosed in Patent Document 1 as described above, the user does push-ups in a state where the chest is pressed against the flat body, and the number of push-ups is measured by the user pressing the flat body, and in the technology disclosed in Patent Document 2, ultrasound Because the sensor cannot measure the position of the hip and leg by measuring the distance from the chest of the user who is doing push-ups with both hands supported on the hand support, it is impossible to measure the correct push-ups. There was a problem.

Moreover, in the said patent documents 1 and 2, when the subject performing a push-up performs an abnormal push-up, there also existed a problem that this could not be recognized.

On the other hand, in the technique disclosed in Patent Document 3, the contrast (luminance) value for the boundary line of the image is extracted from the image data about the posture of the person being measured by the grid method, and this value is extracted with the contrast value of the standard posture data and the As a method of generating difference data by comparing, push-ups according to the y-coordinate pixel value of the boundary value where the gripping rod of the measuring mat is located and the y-coordinate pixel value of the boundary value where the chest of the measurer is located on the image data captured by the digital camera However, there was a problem in that push-ups could not be accurately measured because, for example, a case in which the angle of one shoulder and an elbow was misaligned or a case in which the angle to the knee-hip-neck was abnormal could not be recognized. In addition, even in Patent Document 3, when the subject performing push-ups performs abnormal push-ups, there is also a problem that this cannot be recognized.

That is, in the prior art as described above, as a push-up measuring device using a sensor or an image, there is an advantage that the measurer does not need to count the number of measurements directly, but abnormal push-ups of the subject, for example, both shoulders and both elbows. When measuring push-ups by obtaining the angle, it is impossible to recognize when the angle of one shoulder and the elbow is different, the angle to the knee-hip-neck is abnormal, the distance between both arms and both wrists is unbalanced. There was a problem.

An object of the present invention was made to solve the above problems, and it is possible to accurately count the number of measurements by recognizing the posture of push-up (push-up) as 18 joints and 17 links. It is to provide an image-based intelligent push-up discrimination method and a push-up discrimination system that can

Another object of the present invention is to provide an image-based intelligent push-up determination method and a push-up determination system capable of guiding the subject to perform a normal push-up by notifying an abnormal push-up state of the subject.

Another object of the present invention is to provide an image-based intelligent push-up discrimination method and a push-up discrimination system capable of measuring the precise push-up state regardless of the physical condition of the subject.

In order to achieve the above object, an image-based intelligent push-up determination system according to the present invention is an image collection module that captures the push-up state of a subject to collect image data, and the image data collected by the image collection module is open A pose estimation module for estimating position information on 18 joints of a subject with a pose-based algorithm, a normal push-up state in real time or and a posture characteristic evaluation module for evaluating the abnormal push-up state, and a push-up counting module for counting the number of push-ups when the posture characteristic evaluation module evaluates to the normal push-up state.

In addition, in the image-based intelligent push-up determination system according to the present invention, when the posture characteristic evaluation module evaluates to a normal push-up state or an abnormal push-up state, an alarm sound that outputs different alarm sounds in real time It characterized in that it further comprises an output module.

In addition, in the image-based intelligent push-up determination system according to the present invention, the pose estimation module includes nose, neck, right shoulder, right elbow, right wrist, left shoulder, left elbow, left wrist, right hip, right knee, right ankle , left hip, left knee, left ankle, right eye, left eye, right ear, and 18 joints of the left ear are connected with 17 lines to estimate the subject's pose.

In addition, in order to achieve the above object, the image-based intelligent push-up determination method according to the present invention includes (a) a collection step of collecting image data by photographing the push-up state of the subject from the front, side, and front side, ( b) estimating step of estimating position information about the subject's joints with an Openpose-based algorithm with respect to the image data collected in step (a), (c) the subject estimated in step (b) Postural feature evaluation step of evaluating the normal push-up state or the abnormal push-up state in real time according to the position information on the joint of A push-up counting step of counting the number of times, (e) an alarm sound output step of outputting different alarm sounds in real time for each of the normal push-up state or the abnormal push-up state when evaluated as the normal push-up state or the abnormal push-up state in step (c) It is characterized in that it includes.

In addition, in the image-based intelligent push-up determination method according to the present invention, the estimation of the location information in step (b) is the nose, neck, right shoulder, right elbow, right wrist, left shoulder, left elbow, left wrist, right hip , right knee, right ankle, left hip, left knee, left ankle, right eye, left eye, right ear, and left ear 18 joints and the position in the image of 17 lines connecting these 18 joints (x, y It is characterized in that the pose of the subject is estimated based on the coordinates) information.

In addition, in the image-based intelligent push-up determination method according to the present invention, the posture characteristic evaluation in step (c) is performed by (1) maintaining both elbow angles of 150 to 180 degrees during the high posture of the push-up, and -The angle of both elbows is less than 90 degrees in the low position of the push-up, (2) the distance between the wrists and the shoulders in the low position of the push-up is less than half the distance between the wrists and the shoulders in the high position, (3) ) Between push-ups, the difference between the angles of the elbows is maintained below 30 degrees so that the angles of the elbows are balanced in the process of moving from high to low and from low to high, and (4) to the side during push-ups. When the angle of the waist is maintained at 150 degrees or more, and (5) when the angle of both knees is maintained at 150 degrees or more during push-up, it is evaluated as a normal push-up state, and (1) to (5) above It is characterized in that when any one of the criteria is not satisfied, it is evaluated as an abnormal push-up state.

In addition, in the image-based intelligent push-up determination method according to the present invention, the estimation of the location information in step (b) is characterized in that it is estimated by merging image data taken from the front, side, and front side.

As described above, according to the image-based intelligent push-up determination method and push-up determination system according to the present invention, push-up measurement is performed by providing a pose estimation module for estimating position information on the joint of the subject. In the process, there is an effect that the push-up can be accurately measured and counted without contact with the subject's body and irrespective of physical conditions such as height of the subject.

In addition, according to the image-based intelligent push-up determination method and push-up determination system according to the present invention, by providing an alarm sound output module that outputs an alarm sound in real time when an abnormal push-up state is evaluated, the measurement target An effect is obtained that can guide the self to execute a normal push-up.

1 is a block diagram of an image-based intelligent push-up discrimination system according to the present invention;
Figure 2 is a frontal photograph of the subject taken in the image collection module shown in Figure 1,
3 is a photograph for explaining a state in which joints are connected with lines in the pose estimation module for the subject shown in FIG. 2;
4 is a picture showing a state for determining the push-up of the subject in the image-based intelligent push-up discrimination system according to the present invention;
5 is a diagram illustrating an example of evaluation of a normal push-up state and an abnormal push-up state;
6 is a view for explaining the coordinates and angles of the main part of the subject;
7 is a graph showing the high and low thresholds for determining normal and abnormal in push-ups of the subject to which the criteria (1) to (5) are applied;
8 is a process diagram for explaining the image-based intelligent push-up determination process according to the present invention;
9 is a photograph that captures the state of the frontal measurement in the pose estimation module according to the present invention to measure the push-up of the subject;
10 is a photograph of capturing the state of the frontal measurement in the pose estimation module according to the present invention to measure the push-up of the subject;
11 is a photograph in which the state of side measurement is captured in the pose estimation module according to the present invention to measure the push-up of the subject.

The above and other objects and novel features of the present invention will become more apparent from the description of the present specification and accompanying drawings.

As used herein, the term “push-up” refers to push-ups, and “subject” refers to a person who is attempting push-ups.

The image-based intelligent push-up determination method and posture characteristic evaluation in the push-up determination system according to the present invention (1) maintain the angle of both elbows at 150-180 degrees during the high posture of the push-up, and The angle of both elbows is less than 90 degrees in the low position of the push-up, (2) the distance between the wrists and the shoulders in the low position of the push-up is less than half the distance between the wrists and the shoulders in the high position, and (3) In the process of moving from high to low and from low to high between push-ups, the difference between the angles of the elbows is maintained below 30 degrees so that the angles of the elbows are balanced, and (4) the lateral waist during push-ups. When the angle of is maintained at 150 degrees or more, and (5) when both knees are maintained at 150 degrees or more during push-up, it is evaluated as a normal push-up state, and the criteria of (1) to (5) above If even one of the values is not satisfied, it is judged as an abnormal push-up state.

Hereinafter, an embodiment according to the present invention will be described with reference to the drawings.

1 is a block diagram of an image-based intelligent push-up discrimination system according to the present invention.

As shown in FIG. 1, the image-based intelligent push-up determination system according to the present invention is an image collection module 100 that collects image data by photographing a push-up state of a subject, the image collection module 100 Pose estimation module 200 for estimating position information on 18 joints of the subject with an Openpose-based algorithm with respect to the image data collected in Posture feature evaluation module 300 that evaluates a normal push-up state or an abnormal push-up state in real time according to location information on Push-up counting module 400 for counting the number of ups, outputting different alarm sounds in real time for each when evaluated as a normal push-up state or an abnormal push-up state in the posture characteristic evaluation module 300 and an alarm sound output module 500 .

As shown in FIG. 2 , the image collection module 100 uses an image captured by at least one camera. FIG. 2 is a frontal photograph of a subject taken by the image collection module shown in FIG. 1 . The image collection module 100 captures and shoots various camera angles such as the front, side (90 degrees), and frontal (45 degrees) of the subject who is a person who is trying to do push-ups, and is not limited to a specific camera . For example, to collect images of the subject's push-up movements, the images could be acquired by placing two cameras to capture the subject's head, chest, shoulders, elbows, wrists, waist, knees, and feet. there is.

The pose estimation module 200 detects 18 joints of a person with an open pose-based algorithm, for example, as shown in FIGS. 3 and 4 .

3 is a photograph for explaining a state in which joints are connected with lines in the pose estimation module for the subject shown in FIG. 2, and FIG. 4 is a push-up of the subject in the image-based intelligent push-up determination system according to the present invention. - It is a photograph that shows the state for determining the up.

In addition, Open Pose is an artificial intelligence algorithm for joint detection that uses multi-GPU-based deep learning to detect the postures of multiple subjects in real time and analyze the movements of the human body. The detected joints are nose, neck, right shoulder, right elbow, Right wrist, left shoulder, left elbow, left wrist, right hip, right knee, right ankle, left hip, left knee, left ankle, right eye, left eye, right ear, left ear total 18 pieces. When a joint is detected in this way, coordinate information for each joint can be known.

In the present invention, as shown in FIG. 4 , the pose of the subject is estimated by photographing the push-up state of the subject from the front, side, and front side, and connecting the 18 joints with 17 lines. That is, location (x, y coordinates) information in the images of 18 joints and 17 lines is extracted.

The posture feature evaluation module 300 evaluates whether a push-up is being performed when a specific motion is performed using coordinate information on 18 joints detected by the pose estimation module 200 .

The posture characteristic evaluation module 300 uses the coordinate information in the image for each joint extracted from the pose estimation module 200 as deep learning input data. Accordingly, it is determined whether push-up (push-up) is performed in real time for every frame captured by the camera of the image collection module 100 .

For example, the posture characteristic evaluation module 300 of the present invention applies the measurement criteria of the normal push-up state or the abnormal push-up state as follows based on the push-up criterion of the army physical examination.

That is, in the case of the normal push-up state in the posture characteristic evaluation module 300, (1) both elbow angles are maintained at 150-180 degrees in the high posture of the push-up, and both elbows in the low posture of the push-up All elbow angles are less than 90 degrees, (2) the distance between the wrists and shoulders in the low posture of the push-up is less than half the distance between the wrists and the shoulders in the high posture, and (3) the height between the push-ups In the process of moving from low position to high position, the difference between the angles of the elbows is maintained below 30 degrees so that the angles of the elbows are balanced, and (4) the angle of the lateral waist is 150 degrees or more during push-up. (5) is evaluated as a normal push-up state only when the angle of both knees is maintained at 150 degrees or more during the push-up, and any one of the above-mentioned criteria (1) to (5) Even if it is not satisfied, it is judged as an abnormal push-up state.

The posture characteristic evaluation module 300 evaluates, for example, a normal push-up state and an abnormal push-up state according to a reference posture as shown in FIG. 5 . 5 is a diagram illustrating an example of evaluation of a normal push-up state and an abnormal push-up state. Fig. 5 (a) is the highest level posture from the front, Fig. 5 (b) is the lowest level posture from the front, Fig. 5 (c) is the highest level posture from the side, Fig. 5 (d) ) represents the lowest level posture in the side, and the posture characteristic evaluation module 300 determines the normal push-up state of the subject with respect to FIGS. 5 (a) to (d).

On the other hand, Fig. 5 (e) is an incorrect posture that cannot satisfy the angle of both elbows, Fig. 5 (f) is an incorrect posture in which both shoulders are out of balance, Fig. 5 (g) is a waist up and knees bent Incorrect posture bent to the side, Fig. 5 (h) shows an incorrect posture in which the lower back is lowered, and the posture characteristic evaluation module 300 in Figs. 5 (e) to (h) is an abnormal push-up state of the subject judge

To apply these criteria, a parameter classification model is prepared.

As the reference (1) above, the elbow angle θ _ei can be measured by applying the cosine law described in (1) below.

.....(1)

.....(One)

The wrist-elbow distance d _wei , the elbow-shoulder distance d _esi , and the shoulder-wrist distance d _swi are the Cartesian coordinates of the wrist (x _wi , y _wi ), the orthogonal angle of the elbow, as shown in Equations (2)-(4) below. It is calculated using the coordinates (x _ei , y _ei ) and the Cartesian coordinates of the shoulder (x _si , y _si ).

.....(2)

.....(2)

.....(3)

.....(3)

......(4)

......(4)

는 개별 신체 구조의 차이를 최소화하기 위해 고려된다. 가장 높은 수준의 거리와 가장 낮은 자세의 거리를 측정하여 두 번째 기준이 충족되는지 확인하기 위해 식 (5)의

의 비율을 계산한다.As the criterion (2), the distance between the midpoint of both wrists and the midpoint of both shoulders

are considered to minimize differences in individual body structures. To ensure that the second criterion is met by measuring the distance of the highest level and the distance of the lowest posture,

Calculate the ratio of

.....(5)

.....(5)

As the criterion (3) above,

θ _e0 and θ _e1 are measured, and the distance between θ _e0 and θ _e1 measures the balance using the following formula (6).

.....(6)

.....(6)

For the above criteria (4) and (5), the angles θ _t and θ _k of the waist and the knee are calculated in a similar manner to the elbow angle θ _e described in (1) to (4) above, respectively. Coordinates and angles of major body parts of the subject may be described in two different ways, as shown in FIG. 6 . 6 is a view for explaining the coordinates and angles of the main part of the subject. FIG. 6 (a) is a front view, and FIG. 6 (b) is a side view.

By applying the criteria (1) to (5) as described above, the joint coordinate information of the subject is used for deep learning learning, and through this, it is finally determined whether the push-ups of the subject are normal or abnormal.

7 is a graph showing high and low thresholds for determining normal and abnormal in push-ups of a subject to which criteria (1) to (5) are applied.

The push-up coefficient module 400 performs push-ups normally through real-time deep learning-based inference based on the detected joint coordinate data when it is determined as a normal push-up state by the posture characteristic evaluation module 300 . The number of times is counted, and such a count may be output as a voice.

That is, the alarm sound output module 500 may notify the subject by notifying the accumulated count of the number of push-ups in the normal case of push-up, and in an abnormal case, notify the subject by outputting an alarm sound Thus, it is possible to notify the subject to perform a normal push-up.

Next, an image-based intelligent push-up determination method according to the present invention will be described with reference to FIGS. 8 to 11 .

8 is a process chart for explaining the image-based intelligent push-up determination process according to the present invention, and FIG. 9 is a state of frontal measurement captured in the pose estimation module according to the present invention to measure the push-up of the subject. 10 is a photograph captured by the pose estimation module according to the present invention to measure the push-up of the subject, and FIG. 11 is a photograph of the present invention for measuring the push-up of the subject. It is a picture that captured the state of the side measurement in the pose estimation module according to

In the image-based intelligent push-up determination method according to the present invention, first, as shown in FIG. 4 , the push-up state of the subject is photographed from the front, front, and side to collect image data (S10).

Next, with respect to the image data collected in step S10, the joint of the subject, that is, the subject's nose, neck, right shoulder, right elbow, right wrist, left shoulder, left elbow, left The 18 joints of the wrist, right hip, right knee, right ankle, left hip, left knee, left ankle, right eye, left eye, right ear, and left ear, and the position in the image of 17 lines connecting these 18 joints The pose of the subject is estimated based on (x, y coordinates) information (S20).

Then, as shown in FIGS. 9 to 11 according to the position information on the joint of the subject estimated in step S20, the normal push-up state or the abnormal push-up state in the front, the side, and the front side of the subject in real time Evaluate the up state (S30).

In step S30, (1) both elbow angles are maintained at 150 to 180 degrees in the high posture of the push-up, and the angles of both elbows are less than or equal to 90 degrees in the low posture of the push-up, (2) push-up -The distance between the wrists and shoulders in the low posture of the up is less than half the distance between the wrists and the shoulders in the high posture, and (3) in the process of moving from a high position to a low position and from a low position to a high position between push-ups. The angle difference between the elbows is maintained below 30 degrees so that the angles of the elbows are balanced, (4) the angle of the lateral waist is maintained at 150 degrees or more during push-up, and (5) the angle of the knees in the push-up is maintained. Only in the case of a criterion in which all angles are maintained at 150 degrees or more, a normal push-up state is evaluated, and the number of push-ups is counted (S40).

On the other hand, if the abnormal push-up state does not satisfy any one of the criteria (1) to (5) described above in step S30, an alarm sound is output in real time (S50), so that the correct push-up state is provided to the subject can induce

As described above, in the image-based intelligent push-up determination method according to the present invention, the push-up state can be precisely measured without contact with the subject's body during the push-up measurement process, and the height of the subject, etc. You can count push-ups accurately regardless of your physical condition.

Although the invention made by the present inventors has been described in detail according to the above embodiments, the present invention is not limited to the above embodiments and various modifications can be made without departing from the gist of the present invention.

By using the image-based intelligent push-up determination method and the push-up determination system according to the present invention, the push-up can be accurately measured and counted regardless of the body condition such as the height of the subject.

100: image acquisition module
200: pose estimation module
300: Posture feature evaluation module
400: push-up counting module
500: alarm sound output module

Claims (7)

an image acquisition module that collects image data by photographing the push-up state of the subject;
Using GPU-based deep learning for the image data collected from the image collection module, an algorithm that detects the postures of a plurality of subjects in real time and analyzes the movements of the human body selects 18 joints and the 18 joints of the subject A pose estimation module that detects position information about 17 lines connected by
A posture to evaluate the normal push-up state or the abnormal push-up state in real time by using the coordinate information in the image for the 18 joints and 17 lines of the subject detected by the pose estimation module as the input data of the deep learning feature evaluation module,
and a push-up counting module that counts the number of push-ups when the posture characteristic evaluation module evaluates to a normal push-up state,
The image collection module collects images of the subject's head, chest, shoulder, elbow, wrist, waist, knee and foot in real time by arranging a camera that captures the front and side surfaces of the subject,
The posture characteristic evaluation module maintains both elbow angles of 150 to 180 degrees during the high posture of the push-up in real time for every frame of the front and side images taken by the camera of the image collection module, and push-up If the angle of both elbows is less than 90 degrees in the low posture of
The elbow angle is an image-based intelligent push-up discrimination system, characterized in that measured by applying the wrist-elbow distance, the elbow-shoulder distance, and the shoulder-wrist distance. In claim 1,
When the posture characteristic evaluation module evaluates to a normal push-up state or an abnormal push-up state, it further comprises an alarm sound output module for outputting different alarm sounds in real time for each image-based intelligent push- up discrimination system. In claim 1,
The pose estimation module includes nose, neck, right shoulder, right elbow, right wrist, left shoulder, left elbow, left wrist, right hip, right knee, right ankle, left hip, left knee, left ankle, right eye, left eye , an image-based intelligent push-up discrimination system, characterized in that the pose of the subject is estimated by connecting 18 joints of the right and left ears with 17 lines. (a) a collection step of collecting image data by photographing the push-up state of the subject in the image acquisition module from the front and the side;
(b) Using GPU-based deep learning in the pose estimation module for the image data collected in step (a) above, it is an algorithm that detects the postures of multiple subjects in real time and analyzes the movements of 18 subjects. Estimation step of estimating position information on the joint and 17 lines connecting the 18 joints with lines;
(c) Using the coordinate information in the image for the 18 joints and 17 lines of the subject estimated in step (b) as the input data of the deep learning, a normal push-up state or Postural characteristic evaluation step to evaluate the abnormal push-up state;
(d) a push-up counting step of counting the number of push-ups in the push-up counting module if the normal push-up state is evaluated in step (c);
(e) an alarm sound output step of outputting different alarm sounds in real time from an alarm sound output module for each of the cases evaluated as a normal push-up state or an abnormal push-up state in step (c);
The image collection module collects images of the subject's head, chest, shoulder, elbow, wrist, waist, knee and foot in real time by arranging a camera that captures the front and side surfaces of the subject,
The posture characteristic evaluation module maintains both elbow angles of 150 to 180 degrees during the high posture of the push-up in real time for every frame of the front and side images taken by the camera of the image collection module, and push-up If the angle of both elbows is less than 90 degrees in the low posture of
The elbow angle is an image-based intelligent push-up determination method, characterized in that the wrist-elbow distance, elbow-shoulder distance, and shoulder-wrist distance are applied and measured. In claim 4,
Estimation of location information in step (b) is for nose, neck, right shoulder, right elbow, right wrist, left shoulder, left elbow, left wrist, right hip, right knee, right ankle, left hip, left knee, left ankle , characterized by estimating the pose of the subject by position (x, y coordinates) information in the image of 18 joints of the right eye, left eye, right ear, and left ear and 17 lines connecting these 18 joints. An image-based intelligent push-up identification method. In claim 5,
In the evaluation of the posture characteristics in step (c), (1) the angles of both elbows are maintained at 150 to 180 degrees in the high posture of the push-up, and the angles of both elbows are less than 90 degrees in the low posture of the push-up. , (2) the distance between the wrists and shoulders in the low posture of the push-up is less than half the distance between the wrists and the shoulders in the high posture, and (3) in the low posture and from the high position to the low position between the push-ups. In the process of moving in a high posture, the difference between the angles of the elbows is maintained below 30 degrees so that the angles of the elbows are balanced, (4) the angle of the lateral waist is maintained above 150 degrees during push-up, and (5) the push-up -A normal push-up condition is evaluated only when the angle of both knees is maintained at 150 degrees or higher during the up-up, and when any one of the criteria of (1) to (5) above is not satisfied, abnormal push-up condition is evaluated. An image-based intelligent push-up discrimination method, characterized in that it is evaluated as a state. delete

Images