Artificial intelligence judgment system for teenager robot competition

Abstract

The invention discloses an artificial intelligence judgment system for a teenager robot competition. The artificial intelligence judgment system comprises a camera, an industrial tablet computer and a cloud server, the camera is used for collecting pictures of a competition field in real time, and the industrial panel uploads the competition pictures collected by the camera to the cloud server; the cloud server performs task judgment on the uploaded site picture according to the competition name and group, the cloud server analyzes the competition video, identifies tasks completed by the robots in the competition process and scores each robot according to task requirements, and the scoring results are returned to the industrial panel and displayed on the robots; the industrial tablet is used for executing a series of operations in the whole competition process; according to the method, judgment is performed based on task rules and is not influenced by human factors, and deviation or omission possibly generated by referees is avoided.

Description

Technical Field

[0001] The present invention relates to the technical field related to intelligent refereeing in robot competitions, and in particular to an artificial intelligence refereeing system for youth robot competitions. Background Art

[0002] The China Youth Robot Competition is a youth science and technology activity actively conceived and organized by the Youth Work Department of the China Association for Science and Technology at the end of the last century. Entering the 21st century, with the rapid application and development of electronic and information technology, youth robot activities have sprung up in more than 20 provinces, autonomous regions, and municipalities in my country. Research, creativity, and hands-on robot production activities have become a new highlight and new field of youth science and technology innovation activities in primary and secondary schools in the new century. AI referee systems in sports competitions, such as VAR (video assistant referee) in football matches and Hawk-Eye (Hawkeye system) in tennis matches, are classic applications of AI technology in the field of sports. These systems use multiple high-resolution cameras to capture game images and use image processing and AI technology to achieve key judgments such as ball out of bounds and scoring.

[0003] Hawk-Eye accurately locates the position of the ball by calculating the trajectory to ensure fair judgment, while VAR assists referees through centralized video analysis to replay and analyze potentially controversial moments. The core feature of these systems is real-time, especially in sports competitions where instant decisions need to be made quickly. However, there is a lack of a refereeing system that can be applied to actual robot competitions and greatly improve the fairness, transparency and work efficiency of referees. Summary of the invention

[0004] In order to solve the defects of the prior art, the present invention provides an artificial intelligence referee system for youth robot competitions.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0006] The present invention discloses an artificial intelligence referee system for a youth robot competition, comprising a camera, an industrial tablet and a cloud server; the camera is used to collect images of a competition venue in real time, and the industrial tablet uploads the competition images collected by the camera to the cloud server; the cloud server performs task determination on the uploaded venue images according to the competition name and group, and the cloud server identifies the tasks completed by the robot during the competition by analyzing the competition video, and scores each robot according to the task requirements, and the scoring results are then returned to the industrial tablet and displayed on the robot; the industrial tablet is used to perform a series of operations in the entire competition process, including scanning a code to confirm information of a contestant, selecting a contestant's match, starting and ending recording of the competition, recording the number of retries of the contestant, displaying the competition results, and signing and confirming with the contestant;

[0007] The refereeing method of the intelligent refereeing system comprises the following steps:

[0008] Step 1: Preparation before the competition. Before the competition begins, the system will perform a series of initialization preparations. First, the contestants scan the QR code or other confirmation methods to ensure that the competition information, such as the contestant's name and group, has been entered into the system. This operation is completed through an industrial tablet. The referee scans the contestant's QR code through the industrial tablet or manually enters the contestant's information.

[0009] Step 2: The camera will record the entire competition process. From a bird's-eye view, the camera can clearly capture the activity trajectory and task execution status of each robot in the competition. The camera's video signal will be transmitted to the cloud server in real time for subsequent task determination.

[0010] Step 3: After the game, the Android industrial tablet will upload the game footage recorded by the camera to the cloud server. The cloud server will make a detailed judgment on the players' task completion in the game based on the uploaded video, combined with the name and group of the game. Through convolutional neural network image recognition technology, the system can accurately identify whether the players have completed specific tasks, such as object placement and route travel.

[0011] As a preferred technical solution of the present invention, after the competition results are generated, the industrial tablet will display the final scores of all contestants, including the scores of each task and the overall score; the referees can confirm with the contestants based on these results, and the contestants and referees can sign on the industrial tablet to confirm the final scores.

[0012] As a preferred technical solution of the present invention, the system archives the data of each game through a cloud server. All game videos, task completion status and final results will be saved in the cloud server for post-game analysis and review. Referees can retrieve historical data at any time to further analyze the players' performance.

[0013] As a preferred technical solution of the present invention, the method for the system to accurately identify whether the player has completed a specific task through convolutional neural network image recognition technology is to divide the image into S×S grids, establish a target detection model based on the convolutional neural network, and use the target detection model to perform target detection for each grid, wherein the target detection includes the number of bounding boxes B predicted for each grid; the number of object categories C; the probability p (class) that the object in the grid belongs to a certain category; the probability p (project) whether the object exists in the grid; t x ,t y ,t w ,t h: The relative coordinates of the bounding box of the object; then the action of the robot in the image is identified to determine whether the robot has completed a specific task.

[0014] As a preferred technical solution of the present invention, the target detection model outputs a vector containing a bounding box and an object category prediction for each grid, and the length of the vector is (B×5+C), where 5 are values ​​related to each bounding box: used to penalize the deviation between the predicted coordinates of the bounding box and the true bounding box; the formula is:

[0015]

[0016] in is the parameter of the ground-truth bounding box, 1 obj is an indicator function whose value is 1 when there is an object in the grid.

[0017] Confidence loss: used to calculate the confidence loss of the object's existence, measuring the model's confidence in the existence of the object.

[0018]

[0019] where λ noobj It is to adjust the loss weight of the mesh without objects.

[0020] Class loss: It is used to calculate the difference between the predicted class and the true class.

[0021]

[0022] Where p(class i ) is the predicted probability of category i, is the probability of the true class.

[0023] Combining the above three parts, the total loss function of the model is:

[0024]

[0025] As a preferred technical solution of the present invention, when multiple bounding boxes can detect the same object, a non-maximum suppression method is used to retain the optimal box and remove duplicate boxes. The calculation formula of NMS is as follows:

[0026]

[0027] If two bounding boxes B i and B j If the ratio of the overlapping area to their union is greater than the set threshold, one of the boxes is deleted.

[0028] The beneficial effects of the present invention are:

[0029] 1. This kind of youth robot competition artificial intelligence referee system automates the task judgment and scoring process. This system greatly reduces the subjective interference of human referees and ensures that the task completion of each player can be evaluated fairly and impartially. The camera records the entire process and uploads it to the cloud server. The AI ​​system makes judgments based on the task rules, which is not affected by human factors and avoids possible deviations or omissions by referees.

[0030] 2. The present invention automatically completes task determination and scoring through intelligent processing, and the referee only needs to confirm, which greatly improves the efficiency of game determination. This not only shortens the waiting time after the game, but also ensures that the game results can be accurately displayed in the shortest time. All operations are centralized through the tablet, and the referee only needs to operate through a simple interface, such as confirming player information, selecting game sessions, and recording games, etc., which reduces the complexity of manual intervention. After the game, the system automatically uploads the video to the cloud server for determination, and the referee does not need to manually review the completion of each game task, reducing the workload and the impact of human fatigue. The present invention provides detailed task completion and score reports, and all game processes and results can be clearly traced. This transparency can enhance the trust of players and spectators in the results of the game and reduce disputes and uncertainties in the game. Players can check their scores at any time through the tablet to ensure the fairness and transparency of the process and results. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0032] In the attached picture:

[0033] Figure 1 It is a structural schematic diagram of an artificial intelligence referee system for a youth robot competition according to the present invention.

[0034] In the picture: 1. Camera; 2. Industrial tablet; 3. Cloud server. DETAILED DESCRIPTION

[0035] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0036] Example: Figure 1As shown, the present invention provides an artificial intelligence referee system for a youth robot competition, comprising a camera 1, an industrial tablet 2 and a cloud server 3; the camera 1 is used to collect images of the competition venue in real time, and the industrial tablet 2 uploads the competition images collected by the camera to the cloud server 3; the cloud server 3 performs task determination on the uploaded venue images according to the competition name and group, and the cloud server 3 identifies the tasks completed by the robot during the competition by analyzing the competition video, and scores each robot according to the task requirements, and the scoring results are then returned to the industrial tablet and displayed on the robot; the industrial tablet is used to perform a series of operations in the entire competition process, including scanning a code to confirm the information of the contestants, selecting the contestants' competition sessions, starting and ending the recording of the competition, recording the number of retries of the contestants, displaying the competition results, and signing and confirming with the contestants;

[0037] The refereeing method of the intelligent refereeing system comprises the following steps:

[0038] Step 1: Preparation before the competition. Before the competition begins, the system will perform a series of initialization preparations. First, the contestants scan the QR code or other confirmation methods to ensure that the competition information, such as the contestant's name and group, has been entered into the system. This operation is completed through an industrial tablet. The referee scans the contestant's QR code through the industrial tablet or manually enters the contestant's information.

[0039] Step 2: The camera will record the entire competition process. From a bird's-eye view, the camera can clearly capture the activity trajectory and task execution status of each robot in the competition. The camera's video signal will be transmitted to the cloud server in real time for subsequent task determination.

[0040] Step 3, after the game, the Android industrial tablet will upload the game screen recorded by the camera to the cloud server. The cloud server will make a detailed judgment on the completion of the players' tasks in the game based on the uploaded video, combined with the name and group of the game. Through the convolutional neural network image recognition technology, the system can accurately identify whether the players have completed specific tasks, such as object placement and route travel. The present invention automatically completes task judgment and scoring through intelligent processing, and the referee only needs to confirm, which greatly improves the efficiency of game judgment. This not only shortens the waiting time after the game, but also ensures that the game results can be accurately displayed in the shortest time. All operations are centralized through the tablet. The referee only needs to operate through a simple interface, such as confirming player information, selecting game sessions and recording games, etc., which reduces the complexity of manual intervention. After the game, the system automatically uploads the video to the cloud server for judgment, and the referee does not need to manually review the completion of each game task, reducing the workload and the impact of human fatigue. The present invention provides detailed task completion and score reports, and all game processes and results can be clearly traced. This transparency can enhance the trust of players and spectators in the results of the game and reduce disputes and uncertainties in the game. Players can check their scores at any time through tablets to ensure fairness and transparency of the process and results.

[0041] Among them, after the competition results are generated, the industrial tablet will display the final scores of all players, including the scores of each task and the overall score; the referees can confirm with the players based on these results, and the players and referees can sign on the industrial tablet to confirm the final scores.

[0042] The system archives the data of each game through a cloud server. All game recordings, task completion status and final results will be saved in the cloud server for post-game analysis and review. Referees can retrieve historical data at any time to further analyze the players' performance.

[0043] The method by which the system accurately identifies whether a player has completed a specific task through convolutional neural network image recognition technology is to divide the image into S×S grids, establish a target detection model based on the convolutional neural network, and use the target detection model to perform target detection for each grid, wherein the target detection includes the number of bounding boxes B predicted for each grid; the number of object categories C; the probability p(class) that an object in the grid belongs to a certain category; the probability p(project) that an object exists in the grid; t x ,t y ,t w ,t h: The relative coordinates of the bounding box of the object are the ratio of the offset, width and height of the center point of the bounding box relative to the grid position; and then the action of the robot in the image is identified to determine whether the robot has completed a specific task.

[0044] The target detection model outputs a vector containing the bounding box and object category prediction for each grid. The length of the vector is (B×5+C), where 5 are the value center point coordinates associated with each bounding box. x ,t y , width and height t w ,t h and confidence p(project);

[0045] Each bounding box outputs the following information: the probability p(project) of whether there is an object in the grid; the position offset t of the center of the bounding box relative to the grid x ,t y ; The relative size of the bounding box width and height t w ,t h ;Probability distribution of categories C;

[0046] The target detection model output vector is:

[0047] Output=[p(project),t x ,t y ,t w ,t h ,p(class 1 ),p(class 2 ),...,p(class C )]

[0048] Where p(project) is the confidence of each grid predicting an object, p(class i ) is the probability that an object belongs to a certain category.

[0049] The training of the target detection model uses a loss function to measure the difference between the predicted box and the true box; the loss function of the target detection model usually contains three parts:

[0050] Location loss Localizationloss: used to penalize the predicted coordinate center coordinate t of the bounding box x ,t y and width and height w ,t h Deviation from the true bounding box; where the formula is:

[0051]

[0052] in is the parameter of the ground-truth bounding box, 1 obj is an indicator function whose value is 1 when there is an object in the grid.

[0053] Confidence loss: used to calculate the confidence loss of the existence of an object, measuring the model's confidence in the existence of the object.

[0054]

[0055] where λ noobj It is to adjust the loss weight of the mesh without objects.

[0056] Classificationloss: used to calculate the difference between the predicted category and the true category.

[0057]

[0058] Where p(class i ) is the predicted probability of category i, is the probability of the true class.

[0059] Combining the above three parts, the total loss function of the model is:

[0060]

[0061] When multiple bounding boxes can detect the same object, the non-maximum suppression NMS method is used to retain the optimal box and remove duplicate boxes. The calculation formula of NMS is as follows:

[0062]

[0063] If two bounding boxes B i and B j If the ratio of the intersection of the overlapping areas to their union is greater than the set threshold, one of the boxes is deleted.

[0064] In addition to the visual solution, other sensors such as infrared sensors, pressure sensors, force sensors, etc. can also be used to monitor the robot's movements on the field. By collecting sensor data, the system can analyze the robot's completion of tasks during the game. For example, a force sensor can determine whether the robot has completed a task, or a sensor can determine whether the robot has reached a specific target location. This method does not rely on image recognition technology, but may be limited in the accuracy and complexity of task determination, especially for tasks that require precise image analysis.

[0065] When working, in this kind of artificial intelligence referee system for youth robot competitions, camera 1 is used to collect the screen of the competition venue in real time, and the industrial tablet 2 uploads the competition screen collected by the camera to the cloud server 3; the cloud server 3 performs task judgment on the uploaded venue screen according to the competition name and group, and the cloud server 3 analyzes the competition video to identify the tasks completed by the robot during the competition, and scores each robot according to the task requirements. The scoring results are then returned to the industrial tablet and displayed on the robot; the industrial tablet is used to perform a series of operations in the entire competition process, including scanning the code to confirm the information of the contestants, selecting the contestants' competition sessions, starting and ending the recording of the competition, recording the number of retries of the contestants, displaying the competition results, and confirming with the contestants.

[0066] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An artificial intelligence referee system for youth robot competitions, characterized by: The system comprises a camera (1), an industrial tablet (2) and a cloud server (3); the camera (1) is used to collect images of the competition venue in real time, and the industrial tablet (2) uploads the competition images collected by the camera to the cloud server (3); the cloud server (3) performs task determination on the uploaded venue images according to the competition name and group, and the cloud server (3) identifies the tasks completed by the robot during the competition by analyzing the competition video, and scores each robot according to the task requirements, and the scoring results are then returned to the industrial tablet and displayed on the robot; the industrial tablet is used to perform a series of operations in the entire competition process, including scanning a code to confirm the information of the contestants, selecting the contestants' competition sessions, starting and ending the recording of the competition, recording the number of retries of the contestants, displaying the competition results, and confirming with the contestants' signatures; The refereeing method of the intelligent refereeing system comprises the following steps: Step 1: Preparation before the competition. Before the competition begins, the system will perform a series of initialization preparations. First, the contestants scan the QR code or other confirmation methods to ensure that the competition information, such as the contestant's name and group, has been entered into the system. This operation is completed through an industrial tablet. The referee scans the contestant's QR code through the industrial tablet or manually enters the contestant's information. Step 2: The camera will record the entire competition process. From a bird's-eye view, the camera can clearly capture the activity trajectory and task execution status of each robot in the competition. The camera's video signal will be transmitted to the cloud server in real time for subsequent task determination. Step 3: After the game, the Android industrial tablet will upload the game footage recorded by the camera to the cloud server. The cloud server will make a detailed judgment on the players' task completion in the game based on the uploaded video, combined with the name and group of the game. Through convolutional neural network image recognition technology, the system can accurately identify whether the players have completed specific tasks, such as object placement and route travel.

2. The artificial intelligence referee system for youth robot competition according to claim 1 is characterized in that: After the competition results are generated, the industrial tablet will display the final scores of all contestants, including the scores of each task and the overall score; the referees can confirm with the contestants based on these results, and the contestants and referees can sign on the industrial tablet to confirm the final scores.

3. The artificial intelligence referee system for youth robot competition according to claim 1 is characterized in that: The system archives the data of each game through a cloud server. All game recordings, task completion status and final results will be saved in the cloud server for post-game analysis and review. Referees can retrieve historical data at any time to further analyze the players' performance.

4. The artificial intelligence referee system for youth robot competition according to claim 1 is characterized in that: The method for the system to accurately identify whether a player has completed a specific task by using a convolutional neural network image recognition technology is to divide the image into S×S grids, establish a target detection model based on a convolutional neural network, and use the target detection model to perform target detection for each grid, wherein the target detection includes the number of bounding boxes B predicted for each grid; the number of object categories C; The probability that an object in the grid belongs to a certain class p(class); the probability that an object exists in the grid p(project); t x ,t y ,t w ,t h : The relative coordinates of the bounding box of the object (the ratio of the offset, width and height of the center point of the bounding box relative to the grid position); and then identify the action of the robot in the image, so as to determine whether the robot has completed a specific task.

5. The artificial intelligence referee system for youth robot competition according to claim 4 is characterized in that: The target detection model outputs a vector containing a bounding box and object category prediction for each grid. The length of the vector is (B×5+C), where 5 are the values ​​associated with each bounding box (center point coordinates t x ,t y , width and height t w ,t h and confidence p(project); Each bounding box outputs the following information: the probability p(project) of whether there is an object in the grid; the position offset t of the center of the bounding box relative to the grid x ,t y ; The relative size of the bounding box width and height t w ,t h ;Probability distribution of categories C; The target detection model output vector is: Output =[p(project),t x ,t y ,t w ,t h ,p(class1),p(class2),...,p(class C )] Where p(project) is the confidence of each grid predicting an object, p(class i ) is the probability that an object belongs to a certain category.

6. The artificial intelligence referee system for youth robot competition according to claim 4 is characterized in that: The training of the target detection model uses a loss function to measure the difference between the predicted box and the true box; the loss function of the target detection model usually contains three parts: Localization loss: used to penalize the predicted coordinates of the bounding box (center coordinate t x ,t y and width and height w ,t h ) is the deviation from the true bounding box; the formula is: in is the parameter of the ground-truth bounding box, 1 obj is an indicator function whose value is 1 when there is an object in the grid. Confidence loss: used to calculate the confidence loss of the existence of an object, measuring the model's confidence in the existence of the object. where λ noobj It is to adjust the loss weight of the mesh without objects. Classification loss: It is used to calculate the difference between the predicted category and the true category. Where p(class i ) is the predicted probability of category i, is the probability of the true class. Combining the above three parts, the total loss function of the model is:

7. The artificial intelligence referee system for youth robot competition according to claim 4 is characterized in that: When multiple bounding boxes can detect the same object, the non-maximum suppression (NMS) method is used to retain the optimal box and remove duplicate boxes. The calculation formula of NMS is as follows: If two bounding boxes B i and B j The ratio of the overlapping area (intersection) to their union is greater than the set threshold, and one of the boxes is deleted.

Images