CN117919677A - Archery training and competition simulation system - Google Patents

Abstract

The invention discloses an archery training and competition simulation system, which includes: a target recognition module, a physiological characteristic detection module, an environment detection module and a competition simulation module; wherein: the target recognition module uses a target automatic recognition method based on a convolutional neural network to recognize the coordinates of an arrow on a target; the physiological characteristic detection module is used to obtain the user's physiological characteristics in real time; the environment detection module is used to obtain environmental information in real time; the competition simulation module is used to interact with the user to arrange the user to perform a simulated competition; during the simulated competition, the user's hit score is detected by the target recognition module, the user's physiological characteristics are detected by the physiological characteristic detection module, and the environmental information of the competition field is detected by the environment detection module. The system can perform real-time physiological sign detection, environmental information monitoring, and automatic scheduling of competition processes, solving the problems existing in the prior art.

Description

An archery training and competition simulation system

Technical Field

The invention relates to the field of sports training, and in particular to an archery training and competition simulation system.

Background technique

With the development of the times, the progress of science and technology has affected all aspects of people's lives. In major sports events, science and technology are increasingly involved in the integration of various processes of the game. During training, athletes wear smart wearable devices to monitor physiological characteristics in real time for more refined analysis. Intelligent analysis systems also further affect the game process. In the field of football, some studies on the impact of video assistant referees (VAR) on the decision-making of association football referees have been carried out. Due to the improvement of the correct prediction probability of game judgment by scientific data analysis methods, it has successfully influenced the referee's game judgment and decision-making, and has been used in large-scale international events such as the World Cup. The use of sports simulation training systems can provide athletes and coaches with more effective and scientific training methods. After comparing and analyzing the three-dimensional kinematics of judo athletes' back throwing movements in training and simulated competition environments, the researchers concluded that athletes in simulated competitions have shorter time to complete the corresponding movements and higher stability.

Archery is a sport with a long history. It was born in ancient times as a means of hunting for survival. It developed as an important tool of war in the cold weapon era. Until today, it has been on the stage of large-scale international competitions as a modern sport. The number of archery athletes is growing day by day, and archery has gradually developed from a niche sport to a more popular hobby. However, nowadays, whether it is a large international competition or a small archery venue, the combination of archery and technology is still in its initial stage. In terms of target reporting, the score broadcasting and statistics of archery events are still in the manual stage, requiring people to observe the specific situation of the target dozens of meters away through a small telescope. This method is dangerous and inefficient; the rules of archery events are confusing, and it is difficult to train athletes' mentality for competitions by only training simple bow and arrow shooting; data collection and analysis are still in the primitive stage, whether it is the hot spot analysis of arrow impact points or the analysis of physiological characteristics data, it still remains at the theoretical level. These shortcomings of the backward archery training system have hindered the improvement of the overall level of archery.

Summary of the invention

The purpose of the present invention is to provide an archery training and competition simulation system to solve the problems of process management and automated competition evolution that are currently lacking in archery training.

In order to solve the above problems, the present invention provides an archery training and competition simulation system, which includes: a target recognition module, a physiological feature detection module, an environment detection module and a competition simulation module; wherein:

The target recognition module uses a convolutional neural network-based target automatic recognition method to identify the coordinates of the arrow on the target;

The physiological characteristic detection module is used to obtain the user's physiological characteristics in real time;

The environment detection module is used to obtain environmental information in real time;

The competition simulation module is used to interact with the user to arrange the user to participate in a simulated competition; during the simulated competition, the user's hit score is detected by the target recognition module, the user's physiological characteristics are detected by the physiological characteristic detection module, and the environmental information of the competition venue is detected by the environmental detection module.

A further improvement of the present invention is that the target recognition module adopts a convolutional neural network-based target automatic recognition method comprising:

S11, obtaining an RGB video image of the target, performing affine transformation on each frame and image interception to form an original RGB image of a uniform size, thereby obtaining an RGB video frame image sequence;

S12, obtaining key frame information by comparing changes of adjacent frames in the RGB video frame image sequence;

S13, performing target detection on the key frames in the RGB video frame image sequence obtained in S2, and obtaining the coordinate position and angle information of the arrow of the video frame of the initial inspection;

S14, using a difference method to extract the difference between the key frame in S2 and the previous non-key frame, obtain a target image, and process the target image to obtain a binary image with arrow information;

S15. Determine the key point position of the arrow target by combining the arrow coordinate position and angle information obtained in S3 and the binary image with arrow information obtained in S4.

A further improvement of the present invention is that the detection process of the physiological characteristic detection module includes:

S21: collecting the athlete's photoelectric volume pulse wave signal through the smart ring, and preprocessing it to obtain a PPG signal waveform;

S22: The PPG signal collected by the ring is transmitted to the athlete's personal mobile phone APP via Bluetooth;

S23: The mobile APP forwards the signal to the remote server every 2 seconds;

S24: the server calculates the ring information and the athlete information every 2 seconds, obtains the athlete's heart rate and heart rate variability information through motion artifact removal and frequency estimation algorithms, and stores the information in the server;

S25: The archery system requests the server for data and displays the athlete's heart rate and heart rate variability data on the big screen.

A further improvement of the present invention is that in step S21:

The smart ring emits green light to the athlete's skin through LED, which reflects from the blood vessels to form a photoelectric volumetric pulse wave signal. The signal is acquired through the sensor of the ring, and then pre-processed through the circuit, and the influence of ambient light is removed by signal subtraction.

A further improvement of the present invention is that in step S22:

The PPG signal and acceleration signal are transmitted to the paired smartphone APP through the Bluetooth module of the smart ring.

A further improvement of the present invention is that in step S24:

Step S41: the server saves the 2s PPG and acceleration signal data sent by the mobile phone in a database, and reads the PPG and acceleration signal data, historical heart rate, historical peak-to-peak distance and other calculation parameters 10s before the current moment and inputs them into the algorithm function;

Step S42: The algorithm obtains the athlete's heart rate and heart rate variability data through algorithms such as signal filtering, artifact removal, frequency calculation, and peak-to-peak distance fitting, and stores them in a database.

A further improvement of the present invention is that the simulation module for the simulated game comprises:

Athlete management module. Athletes include user athletes and simulated athletes. Each athlete has a unique user_ID, which cannot be repeated. Athlete types include men's single, women's single, men's team, women's team, and mixed team. Athletes enter their personal information and bind their physiological health equipment so that the competition system can store, analyze, and manage information for individual athletes.

The equipment management module includes functions for managing the hardware devices included in the system; for selecting and binding the target cameras, athlete side cameras and corresponding environmental detection equipment required for the event; for viewing and testing the preview images and information stability of the target cameras, athlete side cameras and corresponding environmental detection equipment required for the event; for selecting the area of interest of the camera image and sending data to the background;

The plan making module includes a module for making simulated game plans for athletes included in the database, and is also used to view and manage simulated game plans for athletes included in the database; wherein, plans can be managed and viewed as needed, and plan types include optional training modes and competition modes, and competition types include optional men's singles, women's singles, men's teams, women's teams, and mixed teams; training plans can be made for specific user athletes, including selecting plan types, competition types, simulated athletes, simulated event types, competition sequence, and environmental enhancement mode selection; when making plans in competition mode, the background will generate a corresponding final schedule; past plan results can be managed, and users can conduct targeted training based on past training data.

A further improvement of the present invention is that it also includes a competition layout, which is used to display the required real-time information and prompt information during the simulated competition. The complete competition process warning sound, namely the competition preparation prompt sound, the competition start prompt sound and the competition end prompt sound; the archery warning light, when it is the user athlete's turn to compete, the warning light will light up to remind the athlete to start archery, and when the time is up or the target is judged, the warning light will go out; the archery score prompt for the current round is used to display the score and win-loss status of the user athlete and the simulated athlete in this round; the countdown display, according to the different competition types planned, different countdowns are implemented and the athlete is prompted with the remaining time; the historical results display is used to display the results between the athlete's opponents in this competition; the real-time physiological characteristics information is used to display the athlete's real-time heart rate, blood oxygen and heart rate responsiveness and other information; the real-time environmental information is used to display the wind speed, wind direction, temperature and humidity of the current archery venue.

The competition simulation module also includes an event management module, which includes a complete event process based on the underlying logic of the archery final competition system; wherein, the user can start the competition according to the plan, and after starting the competition, complete the competition according to the competition system. During the competition, the user's archery score, physiological characteristic changes and environmental information changes are displayed in real time on the event panel, and guide the user to the next step; after a group of competitions are completed, the user can manually modify the competition score to deal with the situation where the image recognition algorithm is not accurate enough; after the competition, the competition information including score situation, hot spot statistics, physiological characteristic changes and other information will be entered into the event database, which is convenient for the next step of analysis and adjustment of the training plan.

The planning module assigns a unique game ID to each game. Users can quickly index to the corresponding game based on key information of the planned game, such as keywords, plan type, game type, game status, start time and end time; users can view detailed plan information and game logs of any historical plan.

Furthermore, when selecting a training mode, the planning module will model the players' scores in the database and simulate the match score matching mechanism, plan to select the opponent's virtual scores, and present them during the game; users can also select players with different score probability models to match the competition according to their own needs.

The event progress page will display the complete competition process, including the first athlete starting to shoot arrows at the beginning of the game. If the first athlete is a user athlete, the screen will show the athlete's side camera shooting screen. When the athlete draws the bow and prepares to shoot, the screen switches to the target camera screen, showing all the pictures from the arrow hitting the target to reading the score. If the first athlete is a simulated athlete, the screen will show the simulated athlete portrait screen configured according to the plan and the target score screen generated according to different score probability models. When the first athlete finishes shooting and judging the target, it will quickly switch to the next athlete in order to continue the competition process until the winner is clear.

The device provided by the present invention has the following technical effects:

1. The video image analysis and processing module of the present invention obtains the RGB original image sequence through a high-definition camera, and obtains the score recognition result through an image detection algorithm. The algorithm has high recognition accuracy and fast speed, can meet the needs of real-time score discrimination, and realizes fully automatic target discrimination; the module has strong applicability and can be used for existing archery environments without unnecessary hardware changes, and the modification cost is low.

2. The physiological characteristic detection algorithm of the present invention adopts a smart ring composed of high-performance sensors and signal processing circuits, which has good signal collection position, high accuracy, long battery life and low wearing pressure. In outdoor sports and training scenarios, the collected PPG signals are easily interfered by ambient light and motion artifacts. Our algorithm can effectively remove these noises and improve the monitoring accuracy of athletes' heart rate and heart rate variability under the same conditions.

3. The present invention realizes an archery training and competition simulation system with composite intelligent score recognition, real-time physiological sign detection, environmental information monitoring, and automated competition process. It realizes an online system for a series of processes from athlete data management and analysis, competition plan formulation, competition equipment management, to automated execution of competition processes, making it more convenient for athletes and coaches to conduct simulated training and scientific training.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is an overall functional framework diagram of an archery training and competition simulation system provided by an embodiment of the present invention;

FIG2 is a schematic diagram of hardware deployment of an archery training and competition simulation system provided by an embodiment of the present invention;

FIG3 is a flow chart of a competition section of an archery training and competition simulation system provided by an embodiment of the present invention;

FIG4 is a flow chart of a method for automatic target recognition based on a convolutional neural network;

FIG5 is a diagram showing the architecture of a neural network used in an automatic target recognition method based on a convolutional neural network.

Detailed ways

The following describes the embodiments of the present invention by specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict.

It should be noted that the illustrations provided in the following embodiments are only used to schematically illustrate the basic concept of the present invention, and thus the illustrations only show components related to the present invention rather than being drawn according to the number, shape and size of components in actual implementation. In actual implementation, the type, quantity and proportion of each component may be changed arbitrarily, and the component layout may also be more complicated.

As shown in FIG1 , an embodiment of the present invention provides an archery training and competition simulation system, which includes: a target recognition module, a physiological characteristic detection module, an environment detection module and a competition simulation module; wherein:

The target recognition module uses a convolutional neural network-based target automatic recognition method to identify the coordinates of the arrow on the target;

The physiological characteristic detection module is used to obtain the user's physiological characteristics in real time;

The environment detection module is used to obtain environmental information in real time;

The competition simulation module is used to interact with the user to arrange the user to participate in a simulated competition; during the simulated competition, the user's hit score is detected by the target recognition module, the user's physiological characteristics are detected by the physiological characteristic detection module, and the environmental information of the competition venue is detected by the environmental detection module.

The system of this embodiment can perform composite intelligent score recognition, real-time physiological sign detection, environmental information monitoring, and automatic arrangement of competition process. Among them, the hardware equipment includes a target side camera, an athlete side camera, an environmental information collection device, a smart ring, a workstation, a router, a smart phone, a large display screen, information transmission network cables and wires, etc. The equipment deployment method is shown in Figure 2.

In this embodiment, the game simulation module includes an athlete management module, an equipment management module, a plan making module and an event management module. FIG1 shows the relationship between the modules, and each component module is described in detail below.

Athlete management module: includes functions such as athlete information identification, athlete type management, and athlete equipment binding.

Athlete management is achieved in the following ways: Athlete types include user athletes and simulated athletes, where simulated athletes are configured through background entry, and ordinary users have no editing privileges and can only add, edit, view and delete user athletes. Ordinary users generate user athletes by entering the team name (user athlete name entered), selecting the team type (including men's single, women's single, men's group, women's group, mixed group), specifying the smart ring MAC address (a 12-bit MAC address, such as 00:00:00:00:00:00, bound to the athlete), and adding notes. After each user athlete is generated, the system generates a unique user ID. Ordinary users can index and view the detailed information of the user athlete according to the user ID, team name, team type, and ring MAC address, and can re-edit the information of all the above user athletes, and can delete the created user athletes. After the user athlete enters personal information and the physiological health equipment is bound, the competition system stores information analysis and management on an individual basis.

The device management module includes the function of managing the hardware devices included in the system. Device management is achieved in the following ways: Administrator users can create new cameras by entering camera ID, camera name, camera type, and camera status, and enter them into the background database; ordinary users do not have the authority to create new cameras. Ordinary users can view the preview images taken by all cameras in the database, and operate them to move up, down, left, and right to fully capture the target surface or athlete screen; they can view and edit the camera ID, camera name, camera type (optional types are facing the athlete and facing the target) and add notes to them; they can choose to enable or disable a camera to bind the target surface camera and athlete side camera required for the event; for cameras facing the target, you can select the area of interest of the camera screen. The specific operation is to click the upper left, lower left, upper right, and lower right of the target image with the mouse in turn. The system will record the coordinates of the clicked points relative to the standard image taken by the camera and send the data to the background.

The plan making module includes making, viewing and managing simulated game plans containing athletes in the database. Plan making is achieved in the following ways: In the simulated plan column, ordinary users create a new simulated game plan through the "Add button". The plan types are divided into training and competition. Among them, when the training mode is selected, the background will generate a random number seed based on the selected matching simulated athlete, and assign different numbers of rings (10 rings, 9 rings, 8 rings, 7 rings, 6 rings) according to the athlete's historical competition results. The probability of appearing in the game; when the competition mode is selected, the background will call up the selected simulated athlete's selected past competition results and the corresponding real video. After selecting the plan type, select the required competition type (including men's individual, women's individual, men's team, women's team, mixed team), simulated event type, select the user athletes entered in the athlete management module, and the simulated athletes in the database (in competition mode, you need to select 7 different competition groups and their corresponding real competition matches, and the background will generate a top 8 matchup chart), select the first player at the start of the archery competition (including user athlete first, simulated athlete first and random mode), choose whether to play ambient sound, optionally add notes, and click the "Plan" button to complete the creation of the simulated competition plan. The background will assign a unique ID to each simulated plan.

Plan management is achieved in the following ways: users can quickly index to the corresponding game based on the key information of the planned game, such as keywords, plan type, game type, game status, start time and end time; users can view the detailed plan information and game log of any historical plan. After a simulation plan is completed (so all the game processes are completed), ordinary users can click the "Results" button to view the results of all simulated games in the plan (including the game scores and the sub-scores of each game). Ordinary users can click the "Delete" button to delete a specific simulation game plan that has not completed all games. All plan information (keywords, plan type, game type, game status, game results, plan ID, start time and end time) will be stored in the database.

The event management module includes a complete event process based on the underlying logic of the archery finals. Among them, the necessary information is displayed on the event start panel, including the current number of rounds, the current match score, the historical number of rounds, the names of the two teams in the current game, the current match score, the current number of arrows shot and the corresponding number of rings, warning lights, countdown time, physiological characteristics (including heart rate, blood oxygen, heart rate responsiveness), real-time display of motion capture module (showing the athlete's archery posture), environmental information (including temperature, humidity, wind speed, wind direction) and real-time display of the game screen video module.

The process of event operation is shown in Figure 3, which is implemented in the following way: Taking the single-player competition type as an example, in the planning module, click the "Start Competition" button of the plan to be executed to enter the competition interface. After 10 seconds of preparation time, the competition officially begins. If the first player is selected as the user player in the plan, the real-time display video module of the game screen will display the picture taken by the camera on the side of the player. When the player completes the action of drawing the bow and shooting the arrow, the real-time display video module of the game screen will switch to the picture of the camera on the side of the target. After receiving the target image analysis result sent back by the video image analysis processing algorithm, the real-time display video module of the game screen will switch to the video picture of the simulated player, and the countdown will be paused. This will be carried out alternately in sequence until the end of a game. If the first player is selected as the simulated player in the plan, when the selected plan type is training, the real-time display video module of the game screen will first display the video picture of the simulated player selected in the database. After the simulated player video is played, the real-time display video module of the game screen will switch to the target video picture in the database, wherein the selection of the target video is completed by the background plan. After the target video is played, the real-time display video module of the game screen will switch to the user player video picture, and the countdown will start. This will be carried out alternately in sequence until the end of a game. When the selected plan type is competition, the simulated player picture will be composed of a certain game picture of the athlete selected in the database. In the random mode, the background will generate the simulated player and user player first-hand competition mode with equal probability in the form of random numbers. The competition process of mixed team competitions and team competitions is similar to that of single-player competitions. After playing the screen of a single athlete (for simulated athletes, it is the continuous video screen of the simulated athlete and the target video screen; for user athletes, it is the continuous screen of the athlete's side camera screen and the target side camera screen), it will switch to the screen of the athletes on the same team until the rotation of the team's athletes' screens is completed, and then switch to the screen of another type of athlete.

The event process includes complete score acquisition and statistics functions. Among them, the score acquisition is carried out in real time. After the user athlete shoots the arrow onto the target, the video image analysis and processing algorithm analyzes and processes the video image to obtain the score corresponding to the current arrow, and transmits it back to the system background through the MQTT communication protocol, and displays it in real time on the current game score on the event start panel. The score of the simulated athlete is retrieved from the database by the system background and displayed on the current game score on the event start panel. After a group of games is over, the user can manually modify the game score to deal with the situation where the image recognition algorithm is not accurate enough. After a game is over, the historical game score will be entered into the background and displayed on the event start panel. When all planned games are completed, the game information (including all game scores, target impact hotspot data, and physiological characteristics continuous data) can be viewed on the plan management interface.

The real-time information and necessary prompts required for the layout display of the event are shown in Figure 5, including: a complete game process warning sound, namely, the game preparation prompt sound, the game start prompt sound and the game end prompt sound; an archery warning light, when it is the turn of the user athlete to compete, the warning light will light up to remind the athlete to start archery, and when the time is up or the target is judged, the warning light will go out; the archery score prompt for the current round is used to display the score and win-loss status of the user athlete and the simulated athlete in this round; when the plan chooses to turn on the ambient sound, the background calls the ambient sound of the actual competition venue in the database (including audience cheers, audience boos, noise, wind, etc.) to simulate a more realistic competition scene; countdown display, according to the different competition types planned (men's singles, women's singles, men's teams, women's teams, mixed teams), different countdowns are implemented and athletes are prompted with the remaining time; historical results display is used to display the results between the athletes' opponents in this competition; real-time physiological characteristics information is used to display the athlete's real-time heart rate, blood oxygen and heart rate allergy information; real-time environmental information is used to display the wind speed, wind direction, temperature and humidity of the current archery venue.

As shown in FIG4 , the target recognition module in this embodiment adopts a convolutional neural network-based target automatic recognition method including:

S11, obtaining an RGB video image of the target, performing affine transformation on each frame and image interception to form an original RGB image of a uniform size, thereby obtaining an RGB video frame image sequence;

S12, obtaining key frame information by comparing changes of adjacent frames in the RGB video frame image sequence;

S13, performing target detection on the key frames in the RGB video frame image sequence obtained in S2, and obtaining the coordinate position and angle information of the arrow of the video frame of the initial inspection;

S14, using a difference method to extract the difference between the key frame in S2 and the previous non-key frame, obtain a target image, and process the target image to obtain a binary image with arrow information;

S15. Determine the key point position of the arrow target by combining the arrow coordinate position and angle information obtained in S3 and the binary image with arrow information obtained in S4.

In step S11 of the present embodiment, the process of obtaining the RGB video frame image sequence includes the following steps: first, a high-definition network camera (target surface camera) is used to capture the target image, its RTSP stream is obtained through streaming technology, and an RGB image is captured using an image queue; for the obtained RGB image, the scale-invariant feature transform (SIFT) is used to obtain the matching feature points of the original RGB image and the template image, and the homography matrix thereof is obtained, and an affine transformation is performed, and finally the image is intercepted and processed into an original RGB image of uniform size.

In step S12 of this embodiment, according to the application scenario of archery target recognition, the three-frame inter-frame difference method is used to process the RGB video frame image sequence to detect key frames. Among them, the image frames required to be stored in the hardware include the front, middle and back three frames, and the front, middle and back two frames form image frame pairs respectively, and the image frames constituting these frame pairs are continuously updated according to the recognition results.

The image frame update algorithm is implemented by the following steps: For a standard original RGB video frame sequence, obtain the first, middle, and last three frames of images (hereinafter referred to as frame A, frame B, and frame C) arranged in time order. Subtract frame A from frame B, frame B from frame C, and obtain the motion parameters in these image frames. Under ideal conditions, the subtraction between non-key frames will result in zero motion parameters for two frames with no new arrows in the target surface to be detected; in actual application scenarios, due to environmental factors, the vibration factor of the arrow shooting into the target surface, the flight posture of the arrow passing through the shooting area, and the subtraction between two frames with no new arrows in the target surface area will also show that the motion parameters are not zero. The method to solve the above problem is: morphological processing is performed on the image frame difference whose motion parameter is not zero, and the target area segmentation is performed using the hysteresis threshold to eliminate the phenomenon of discontinuous line detection caused by environmental factors such as illumination; specifically, to obtain key frame B (a new arrow is added to the target area and the arrow is stable), the difference between frame A and frame B should be processed by the above method to control its motion parameter to be not 0; the difference between frame B and frame C should be processed by the above method to control its motion parameter to be zero, thereby eliminating the problem of inaccurate key frame judgment caused by the vibration factor of the arrow shooting into the target surface and the flying posture of the arrow passing through the shooting area, which affects the score judgment. The obtained frame B is the key frame.

The update logic of the video frames to be processed is implemented by the following steps: when frame B is determined to be a non-key frame and the inter-frame difference motion parameter before processing is 0, it is considered that no arrow has entered the target surface, and the newly added image frame D in the video sequence replaces the latest frame C in chronological order, and the previous frames C and B replace frames B and A in turn; if frame B is determined to be a non-key frame and the inter-frame difference motion parameter before processing is not 0, it is considered that there are interference items such as vibration or environmental noise, then the image frame is removed, and a new image frame D is obtained from the video frame sequence to replace frame C, and frame C replaces frame B. If frame B is determined to be a key frame, the frame difference between the key frame (frame B) and the non-interference non-key frame (frame A) closest to the key frame is obtained to obtain the processed key frame binary image and RGB key frame image.

In the above-mentioned archery training and competition simulation system, in step three, the data set collection is first completed. Target images composed of different weather conditions, different shooting angles, and different numbers of arrows are collected, and the arrow targets in the images are manually annotated. The annotated images are annotated using a rotating box annotation method, and are divided into training sets and test sets. Secondly, a convolutional neural network model is designed, and hyperparameters such as learning rate, batch size, and number of iterations are set. Finally, the training image data set is read, and the detection method with the previously designed convolutional neural network model as the core is used for training. The error value between the actual network output and the specified target output is reduced through training. When a certain number of iterations is reached, the training is stopped to obtain the final model. In the design of the neural network, the specific neural network composition architecture is shown in Figure 5:

The steps of using neural network for recognition are as follows: divide the image obtained by subtracting adjacent key frames into S×S grids; for each grid, predict B bounding boxes (including the confidence that each bounding box is a target, the probability that each bounding box area is in the arrow category, and the tilt angle of the detection box when the predicted box is an arrow target). The confidence includes two meanings: one is the probability that the bounding box contains the target, Pr(object). When the bounding box does not contain the target, Pr(object) = 0, otherwise if the bounding box contains the target, Pr(object) = 1; the other is the accuracy of the bounding box. The accuracy of the bounding box can be recorded by the intersection of the predicted box and the actual box, IOU, recorded as IOU. In summary, the expression of confidence can be given by Pr(object)×IOU. The size and position of the bounding box can be represented by a 184-dimensional vector: (x, y, w, h, θ ₀ , ..., θ ₁₇₉ ), where (x, y) is the center coordinate of the bounding box, and w, h are the width and height of the bounding box, respectively, and θ _n is the classification probability of the tilt angle of the bounding box. In particular, the predicted value of the bounding box is actually nc+5+180-dimensional vector (x, y, w, h, conf., θ ₀ , ..., θ ₁₇₉ , C ₀ , ..., C _nc ), where nc is the number of target categories to be detected, which is 1 in this embodiment, and conf. is the confidence level. When making a prediction, S×S×B target windows can be obtained, and then the target windows with relatively low probability are removed according to the threshold, and finally the non-maximum suppression algorithm is used to remove redundant windows. Among them, the use of the non-maximum suppression algorithm mainly solves the problem of multiple detections of a target. Find the box with the highest confidence among all the detection boxes, calculate its IOU with the remaining boxes one by one, and if its value is always greater than a certain threshold, delete the box and repeat the above process until all the detection boxes are processed.

In the key frame and the previous and next frame pairs obtained previously, the two image frame pairs are converted into grayscale images, and image difference is performed respectively, and color space conversion is performed to obtain the initial binary image. The initial binary image obtained is subjected to morphological processing. For the binary image obtained, the hysteresis threshold segmentation method is used, and all pixels with gradient values greater than the high threshold are considered to be strong edges, and all pixels with gradient values less than the low threshold are considered to be non-edges, and the pixels between the high and low thresholds are marked as weak edges. Finally, the weak edges are connected with the strong edges to obtain the processed image; finally, the Hough transform is used to perform line detection and contour detection on the obtained image, and the endpoint coordinates in the obtained contour information are extracted, and the actual coordinates are obtained according to the camera placement angle and the arrow target model. The processed binary image containing the arrow position information is output.

Finally, the processed RGB image is input into the trained convolutional neural network to obtain the arrow coordinate position and angle information of the initial inspection video frame, and the binary image containing the arrow information as described in claim 7, and the final precise coordinates of the arrow are obtained by averaging the extracted coordinates.

In this embodiment, the detection process of the physiological characteristic detection module includes:

S21: collecting the athlete's photoelectric volume pulse wave signal through the smart ring, and preprocessing it to obtain a PPG signal waveform;

S22: The PPG signal collected by the ring is transmitted to the athlete's personal mobile phone APP via Bluetooth;

S23: The mobile APP forwards the signal to the remote server every 2 seconds;

S24: the server calculates the ring information and the athlete information every 2 seconds, obtains the athlete's heart rate and heart rate variability information through motion artifact removal and frequency estimation algorithms, and stores the information in the server;

S25: The archery system requests the server for data and displays the athlete's heart rate and heart rate variability data on the big screen.

In step S21: the smart ring emits green light to the athlete's skin through the LED, which forms a photoelectric volumetric pulse wave signal after reflection through the blood vessels. The signal is acquired through the sensor of the ring, and then the signal is pre-processed through the circuit, and the influence of ambient light is removed by signal subtraction.

In step S22: the PPG signal and the acceleration signal are transmitted to the paired smartphone APP via the Bluetooth module of the smart ring.

In step S24: Step S41: the server saves the 2s PPG and acceleration signal data sent by the mobile phone in the database, and reads the PPG and acceleration signal data, historical heart rate, historical peak-to-peak distance and other calculation parameters 10s before the current moment and inputs them into the algorithm function;

Step S42: The algorithm obtains the athlete's heart rate and heart rate variability data through algorithms such as signal filtering, artifact removal, frequency calculation, and peak-to-peak distance fitting, and stores them in a database.

This embodiment implements an archery training and competition simulation system, and realizes an online system for a series of processes including athlete data management and analysis, competition plan formulation, competition equipment management, and automatic execution of competition processes. Its advantages are:

1. Improve the standardization of the competition process. The current archery simulation competition system is in an immature development stage, and basically only has a simple function of identifying the number of archery rings in the competition. This embodiment realizes an integrated process design from formulating a competition plan, completing the competition process, and reviewing the competition situation, which allows users to understand the competition rules more efficiently and improve training efficiency.

2. Complete the collection and integration of key competition data. By analyzing past competition data and training data, we extracted the key factors affecting the contestants, quantified environmental factors and physiological characteristics, and collected relevant data in the competition and training environment in real time through smart devices and sensors. This not only makes it convenient for contestants to adjust their archery posture according to the data during the competition, but also makes it convenient for coaches to analyze and adjust training methods after the competition.

The above embodiments are merely illustrative of the principles and effects of the present invention, and are not intended to limit the present invention. Anyone familiar with the art may modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by a person of ordinary skill in the art without departing from the spirit and technical ideas disclosed by the present invention shall still be covered by the claims of the present invention.

Claims (8)

1. An archery training and competition simulation system, characterized by comprising: a target recognition module, a physiological characteristic detection module, an environment detection module and a competition simulation module; wherein: The target recognition module uses a convolutional neural network-based target automatic recognition method to identify the coordinates of the arrow on the target; The physiological characteristic detection module is used to obtain the user's physiological characteristics in real time; The environment detection module is used to obtain environmental information in real time; The competition simulation module is used to interact with the user to arrange the user to participate in a simulated competition; during the simulated competition, the user's hit score is detected by the target recognition module, the user's physiological characteristics are detected by the physiological characteristic detection module, and the environmental information of the competition venue is detected by the environmental detection module. 2. An archery training and competition simulation system according to claim 1, characterized in that the target automatic recognition method based on convolutional neural network adopted by the target recognition module comprises: S11, obtaining an RGB video image of the target, performing affine transformation on each frame and image interception to form an original RGB image of a uniform size, thereby obtaining an RGB video frame image sequence; S12, obtaining key frame information by comparing changes of adjacent frames in the RGB video frame image sequence; S13, performing target detection on the key frames in the RGB video frame image sequence obtained in S2, and obtaining the coordinate position and angle information of the arrow of the video frame of the initial inspection; S14, using a difference method to extract the difference between the key frame in S2 and the previous non-key frame, obtain a target image, and process the target image to obtain a binary image with arrow information; S15. Determine the key point position of the arrow target by combining the arrow coordinate position and angle information obtained in S3 and the binary image with arrow information obtained in S4. 3. The archery training and competition simulation system according to claim 1, wherein the detection process of the physiological characteristic detection module comprises: S21: collecting the athlete's photoelectric volume pulse wave signal through the smart ring, and preprocessing it to obtain a PPG signal waveform; S22: The PPG signal collected by the ring is transmitted to the athlete's personal mobile phone APP via Bluetooth; S23: The mobile APP forwards the signal to the remote server every 2 seconds; S24: the server calculates the ring information and the athlete information every 2 seconds, obtains the athlete's heart rate and heart rate variability information through motion artifact removal and frequency estimation algorithms, and stores the information in the server; S25: The archery system requests the server for data and displays the athlete's heart rate and heart rate variability data on the big screen. 4. An archery training and competition simulation system according to claim 3, characterized in that in step S21: The smart ring emits green light to the athlete's skin through LED, which reflects from the blood vessels to form a photoelectric volumetric pulse wave signal. The signal is acquired through the sensor of the ring, and then pre-processed through the circuit, and the influence of ambient light is removed by signal subtraction. 5. The archery training and competition simulation system according to claim 3, characterized in that in step S22: The PPG signal and acceleration signal are transmitted to the paired smartphone APP through the Bluetooth module of the smart ring. 6. An archery training and competition simulation system according to claim 3, characterized in that in step S24: Step S41: the server saves the 2s PPG and acceleration signal data sent by the mobile phone in a database, and reads the PPG and acceleration signal data, historical heart rate, historical peak-to-peak distance and other calculation parameters 10s before the current moment and inputs them into the algorithm function; Step S42: The algorithm obtains the athlete's heart rate and heart rate variability data through algorithms such as signal filtering, artifact removal, frequency calculation, and peak-to-peak distance fitting, and stores them in a database. 7. An archery training and competition simulation system according to claim 1, characterized in that the competition simulation module comprises: Athlete management module. Athletes include user athletes and simulated athletes. Each athlete has a unique user_ID, which cannot be repeated. Athlete types include men's single, women's single, men's team, women's team, and mixed team. Athletes enter their personal information and bind their physiological health equipment so that the competition system can store, analyze, and manage information for individual athletes. The equipment management module includes functions for managing the hardware devices included in the system; for selecting and binding the target cameras, athlete side cameras and corresponding environmental detection equipment required for the event; for viewing and testing the preview images and information stability of the target cameras, athlete side cameras and corresponding environmental detection equipment required for the event; for selecting the area of interest of the camera image and sending data to the background; The planning module includes a method for formulating a simulated game plan for athletes included in the database. 8. An archery training and competition simulation system according to claim 1, characterized in that it also includes a competition progress page for displaying required real-time information and prompt information during the simulated competition.