CN114582020A - Pull-up counting method based on image vision technology - Google Patents

Abstract

The invention discloses a chin counting-up method based on an image vision technology, which is characterized in that a horizontal bar, a human face and a human hand are detected and positioned based on a convolutional neural network, and the correction of the bar-holding direction is judged; the method is based on the convolutional neural network to detect and position the joints of the human body, judges whether the pull-up action is standard or not through the posture, the joint angle and the action duration of the hand, can effectively judge the action number meeting the standard, and improves the accuracy and the real-time performance of pull-up counting.

Description

Pull-up counting method based on image vision technology

Technical Field

The invention relates to the field of physical ability testing, in particular to a method for automatically counting a pull-up direction based on image vision.

Background

The pull-up technical method based on the deep learning technology and the image vision technology judges whether the action is standard or not by modeling and analyzing the testee, thereby calculating the standard action quantity.

The current pull-up counting method mainly has the following defects:

(1) the cost is high;

(2) whether the action is standard or not cannot be judged;

(3) manual operation is needed for counting;

the above disadvantages will affect the accuracy and real-time performance of pull-up counting.

Disclosure of Invention

The invention aims to: the pull-up counting method based on the image vision technology is provided, a human body joint is detected and positioned based on a convolutional neural network, whether pull-up action is standard or not is judged through judgment of hand posture, joint angle and action duration, and the action quantity meeting the standard can be effectively judged.

The technical scheme of the invention is as follows:

a pull-up counting method based on image vision technology comprises the following steps:

s101, installing pressure sensors at two hands on a horizontal bar, and sensing the pressure signals to start detection;

s102, inputting a single-frame image;

s103, detecting and positioning a horizontal bar, a human face and a human hand by using the trained convolutional neural network;

s104, judging the holding bar direction according to the detection and positioning of the horizontal bar, the human face and the human hand: if the user holds the bar reversely, the detection is finished; if the user holds the bar, the step S105 is executed;

s105, detecting and positioning the human body joint points by using the trained convolutional neural network, wherein the positioned human body joint points comprise: wrist a, elbow b, shoulder c;

s106, calculating the vertical distance Y from the horizontal bar to the center of mass of the face, and putting the calculation result into a set Y;

s107, calculating a bending angle ^ abc of the elbow, and putting a calculation result into the set A;

s108, if the angle abc is larger than an angle threshold value alpha, and the continuous time t is larger than a time threshold value t_thres1Or less than a time threshold t_thres2And the detection is finished;

s109, after the detection is finished, the image is not input any more, the set A is taken, and n subsets A are obtained by dividing the set A at the maximum value_n；

S110, taking subset A_nTwo endpoint values A_n1，A_n2；

S111, enabling the set Y to correspond to n subsets A_nIs divided into n subsets Y_nSubset Y_nMaximum value of (Y)_{n max}；

S112, judging whether each pull-up action is standard or not, and inputting a distance threshold value y_thresIf A is_n1Not less than alpha, and A_n2≥α，Y_{n max} >y_thresIf the three conditions are met simultaneously, the pull-up action standard of the time is considered, counting is added by 1, otherwise, the pull-up action of the time is considered not to be standard and not counting is considered;

and S113, outputting a counting result.

Preferably, the angle threshold α is 165 ° to 175 °.

The invention has the advantages that:

the pull-up counting method based on the image vision technology detects and positions human joints based on the convolutional neural network, judges whether pull-up actions are standard or not by judging the posture, joint angle and action duration of a hand, can effectively judge the number of actions meeting the standard, and improves the accuracy and real-time performance of counting.

Drawings

The invention is further described with reference to the following figures and examples:

FIG. 1 is a flow chart of a pull-up counting method based on image vision technology according to the present invention;

fig. 2 is a schematic view of joint positioning.

Detailed Description

As shown in FIG. 1, the method for counting the pull-up based on the image vision technology comprises the following steps:

s101, installing pressure sensors at two hands on a horizontal bar, and sensing the pressure signals to start detection;

s102, inputting a single-frame image;

s103, detecting and positioning a horizontal bar, a human face and a human hand by using the trained convolutional neural network;

s104, judging the bar holding direction according to the detection and positioning of the horizontal bar, the human face and the human hand: if the user holds the bar reversely, the detection is finished; if the user holds the bar, the step S105 is executed;

s105, detecting and positioning the human body joint points by using the trained convolutional neural network, wherein the positioned human body joint points comprise: wrist a, elbow b, shoulder c, as shown in fig. 2;

s106, calculating the vertical distance Y from the horizontal bar to the center of mass of the face, and putting the calculation result into a set Y;

s107, calculating a bending angle ^ abc of the elbow, and putting a calculation result into the set A;

s108, if the angle abc is larger than 170 degrees, and the continuous time t is larger than a time threshold t_thres1Or less than a time threshold t_thres2And the detection is finished;

s109, after the detection is finished, the image is not input any more, the set A is taken, and n subsets A are obtained by dividing the set A at the maximum value_n；

S110, taking subset A_nTwo endpoint values A_n1，A_n2；

S111, corresponding the set Y to n subsets A_nIs divided into n subsets Y_nSubset Y_nMaximum value of (Y)_{n max}；

S112, judging whether each pull-up action is standard or not, and inputting a distance threshold value y_thresIf A is_n1Not less than alpha, and A_n2≥170°，Y_{n max} >y_thresIf the three conditions are met simultaneously, the pull-up action standard of the time is considered, counting is added by 1, otherwise, the pull-up action of the time is considered not to be standard and not counting is considered;

and S113, outputting a counting result.

The above embodiments are merely illustrative of the technical ideas and features of the present invention, and the purpose of the embodiments is to enable those skilled in the art to understand the contents of the present invention and implement the present invention, and not to limit the protection scope of the present invention. All modifications made according to the spirit of the main technical scheme of the invention are covered in the protection scope of the invention.

Claims (2)

1. A pull-up counting method based on image vision technology is characterized by comprising the following steps:

s101, installing pressure sensors at two hands on a horizontal bar, and sensing the pressure signals to start detection;

s102, inputting a single-frame image;

s103, detecting and positioning a horizontal bar, a human face and a human hand by using the trained convolutional neural network;

s104, judging the bar holding direction according to the detection and positioning of the horizontal bar, the human face and the human hand: if the user holds the bar reversely, the detection is finished; if the user holds the bar, the step S105 is executed;

s105, detecting and positioning the human body joint points by using the trained convolutional neural network, wherein the positioned human body joint points comprise: wrist a, elbow b, shoulder c;

s106, calculating the vertical distance Y from the horizontal bar to the center of mass of the face, and putting the calculation result into a set Y;

s107, calculating a bending angle ^ abc of the elbow, and putting a calculation result into the set A;

s108, if the angle abc is larger than an angle threshold value alpha, and the continuous time t is larger than a time threshold value t_thres1Or less than a time threshold t_thres2And the detection is finished;

s109, after the detection is finished, the image is not input any more, the set A is taken, and n subsets A are obtained by dividing the set A at the maximum value_n；

S110, taking subset A_nTwo endpoint values A_n1，A_n2；

S111, corresponding the set Y to n subsets A_nIs divided into n subsets Y_nSubset Y_nMaximum value of (Y)_{n max}；

S112, judging whether each pull-up action is standard or not, and inputting a distance threshold value y_thresIf A is_n1Not less than alpha, and A_n2≥α，Y_{n max} >y_thresIf the three conditions are met simultaneously, the pull-up action standard of the time is considered, counting is added by 1, otherwise, the pull-up action of the time is considered not to be standard and not counting is considered;

and S113, outputting a counting result.

2. The pull-up counting method based on image vision technology as claimed in claim 1, wherein the angle threshold α is 165 ° to 175 °.

Images