Curling Motion State Evaluation Method and System Based on 3D Image Dynamic Analysis

Abstract

The present invention discloses a method and system for evaluating the motion state of a curling stone based on three-dimensional image dynamic analysis, belonging to the technical field of image analysis. The present invention collects data on the curling stone throwing point; obtains the data on the curling stone throwing point, and predicts the static point of the curling stone according to the curling stone prediction model; obtains the actual static point of the curling stone, and analyzes the deviation between the actual static point and the predicted static point; obtains the throwing point data and the deviation data, and obtains the positioning of the curling stone abnormality according to the curling stone abnormality positioning model. The present invention first predicts the static point of the curling stone by training a neural network, and then analyzes the positioning of the curling stone abnormal point according to the deviation of the curling stone, which is beneficial to timely and accurately discover the curling stones with abnormalities, so as to reduce the impact of abnormal curling stones on the competition or training.

Description

Technical Field

[0001] The present invention belongs to the technical field of image analysis, and particularly relates to a method and system for evaluating the state of a curling stone based on three-dimensional image dynamic analysis. Background Art

[0002] Curling, also known as throwing curling stones or sliding stones on ice, is a throwing competition event on ice in teams. Using curling stones with abnormalities affects the competition level, competition fairness, and the training efficiency of athletes. Therefore, it is extremely important to detect whether there are abnormalities in curling stones;

[0003] Before a competition or training, curling stones are inspected, usually by inspecting the appearance of the curling stones or using a scanner to conduct a comprehensive inspection of the curling stones. The former is prone to missing or overlooking abnormalities, and the latter has low detection efficiency. Therefore, there is a lack of predicting the stationary point of the curling stone first and then analyzing the abnormal points of the curling stone based on the deviation of the curling stone, resulting in the inability to accurately locate the abnormal coordinates of the curling stone and low detection efficiency;

[0004] To solve the above problems, the present invention proposes a method and system for evaluating the state of a curling stone based on three-dimensional image dynamic analysis. Summary of the Invention

[0005] Aiming at the deficiencies of the prior art, the present invention proposes a method and system for evaluating the state of a curling stone based on three-dimensional image dynamic analysis. The present invention collects data on the throwing points of curling stones; obtains data on the throwing points of curling stones and predicts the stationary point of the curling stone according to the curling stone prediction model; obtains the actual stationary point of the curling stone and analyzes the deviation between the actual stationary point and the predicted stationary point; obtains the throwing point data and deviation data, and obtains the positioning of the abnormality of the curling stone according to the curling stone abnormality positioning model. The present invention predicts the stationary point of the curling stone by training a neural network and then analyzes the positioning of the abnormal points of the curling stone based on the deviation of the curling stone, which is beneficial to timely and quickly discovering curling stones with abnormalities to reduce the impact of abnormal curling stones on competitions or training.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A method for evaluating the state of a curling stone based on three-dimensional image dynamic analysis, comprising the following specific steps:

[0008] Step 1: Collect data on the throwing points of curling stones;

[0009] Step 2: Obtain data on the throwing points of curling stones and predict the stationary point of the curling stone according to the curling stone prediction model;

[0010] Step 3: Obtain the actual stationary point of the curling stone and analyze the deviation between the actual stationary point and the predicted stationary point;

[0011] Step 4: Obtain the throwing point data and deviation data, and obtain the positioning of the curling anomaly according to the curling anomaly positioning model.

[0012] Preferably, the said Step 1 includes the following specific steps:

[0013] Step 1-1: Collect the curling throwing point data, and the said throwing point data includes rotational speed and initial velocity;

[0014] Step 1-2: Use a camera to record the curling throwing process, automatically identify the feature points on the curling through an image recognition algorithm, and establish the three-dimensional motion trajectory of the curling according to the tracking of the feature points;

[0015] Step 1-3: According to the release coordinate of the curling before throwing, set it as marking point 1, mark the moment when the curling is released, obtain the coordinate of the curling on the three-dimensional motion trajectory at the next set moment, set it as marking point 2, and calculate the linear velocity of marking point 1 according to the distance and time interval between marking point 1 and marking point 2. Among them, the calculation formula of the linear velocity is: where d is the distance between marking point 1 and marking point 2, and t is the time interval between the release moment and the next set moment; obtain the distance from marking point 1 to the center point of the curling to calculate the angular velocity of marking point 1. Among them, the calculation formula of the angular velocity is: where r is the distance from marking point 1 to the center point of the curling; obtain the rotational speed of the curling at the release moment according to the angular velocity;

[0016] Step 1-4: Calculate the initial velocity of the curling according to the displacement and time interval from marking point 1 to marking point 2. Among them, the calculation formula of the initial velocity is: where Δx is the displacement from marking point 1 to marking point 2.

[0017] Preferably, the said Step 2 includes the following specific steps:

[0018] Step 2-1: Obtain several groups of historical throwing data of the curling. The said historical throwing data includes historical release coordinates, historical rotational speed, historical initial velocity and historical actual stationary points. Construct a curling motion neural network model with the release coordinates, rotational speed and initial velocity of the curling as the input and the stationary point of the curling as the output. Divide the historical throwing data of the curling into an 80% parameter training set and a 20% parameter test set. Input the 80% parameter training set into the curling motion neural network model for training to obtain an initial curling motion neural network model. Input the 20% parameter test set into the initial curling motion neural network model for testing, and output the optimal initial curling motion neural network model with the highest judgment accuracy of the stationary point of the curling as the curling motion neural network model;

[0019] Step 2-2: The output strategy formula of specific neurons in the said curling motion neural network model is:

[0020]

[0021] Among them, is the abscissa output of the stationary point of the N-item neuron in the n+1 layer, is the ordinate output of the stationary point of the N-item neuron in the n+1 layer, and σ() is the Sigmoid activation function, is the connection weight between the k-th neuron in the n-th layer and the N-item neuron in the n+1 layer, and K is the number of neurons in the n-th layer, is the off-hand abscissa input of the k-th neuron in the n-th layer, is the off-hand ordinate input of the k-th neuron in the n-th layer, is the rotational speed input of the k-th neuron in the n-th layer, is the initial velocity input of the k-th neuron in the n-th layer, is the bias of the linear relationship between the k-th neuron in the n-th layer and the N-item neuron in the n+1 layer;

[0022] Step 3: Input the off-hand coordinates, rotational speed, and initial velocity of the curling stone into the curling stone motion neural network model, and output the predicted stationary point coordinates.

[0023] Preferably, the step 3 includes the following specific steps:

[0024] Step 3-1: Obtain the coordinates of the actual stationary point and the predicted stationary point of the curling stone;

[0025] Step 3-2: Calculate the deviation displacement of the curling stone according to the deviation displacement calculation formula, and the deviation displacement calculation formula is: where x sj is the abscissa of the actual stationary point, x yc is the abscissa of the predicted stationary point, y sj is the ordinate of the actual stationary point, y yc is the ordinate of the predicted stationary point;

[0026] Step 3-3: Calculate the deviation angle of the curling stone according to the deviation angle calculation formula, and the deviation angle calculation formula is: where arctan is the arctangent function.

[0027] Preferably, the step 4 includes the following specific steps:

[0028] Step 4.1: Obtain several groups of historical offset data, historical rotation speed, and historical initial speed of the curling stone. The historical offset data includes historical offset displacement and historical offset angle. Construct a curling stone anomaly location neural network model with the rotation speed, initial speed, offset displacement, and offset angle of the curling stone as the input and the anomaly location point of the curling stone as the output. Divide the historical offset data, historical rotation speed, and historical initial speed of the curling stone into an 80% parameter training set and a 20% parameter test set. Input the 80% parameter training set into the curling stone anomaly location neural network model for training to obtain an initial curling stone anomaly location neural network model. Input the 20% parameter test set into the initial curling stone anomaly location neural network model for testing, and output the optimal initial curling stone anomaly location neural network model with the highest accuracy in judging the anomaly point location of the curling stone as the curling stone anomaly location neural network model;

[0029] Step 4.2: The output strategy formula of specific neurons in the curling stone anomaly location neural network model is:

[0030]

[0031] where, is the abscissa output of the anomaly point of the curling stone of the I-th neuron in the i + 1 layer, is the ordinate output of the anomaly point of the curling stone of the I-th neuron in the i + 1 layer, is the connection weight between the p-th neuron in the i-th layer and the I-th neuron in the i + 1 layer, P is the number of neurons in the i-th layer, is the offset displacement input of the p-th neuron in the i-th layer, is the offset angle input of the p-th neuron in the i-th layer, is the rotation speed input of the p-th neuron in the i-th layer, V i p is the initial speed input of the p-th neuron in the i-th layer, is the bias of the linear relationship between the p-th neuron in the i-th layer and the I-th neuron in the i + 1 layer;

[0032] Step 4.3: Input the offset displacement, offset angle, rotation speed, and initial speed of the curling stone into the curling stone anomaly location neural network model, and output the coordinates of the anomaly point of the curling stone.

[0033] A curling stone motion state evaluation system based on three-dimensional image dynamic analysis, used to implement a curling stone motion state evaluation method based on three-dimensional image dynamic analysis, including a data acquisition module, an image analysis module, a coordinate prediction module, a model construction module, a deviation analysis module, an anomaly location module, and a control module;

[0034] Preferably, the data acquisition module is used to collect curling stone throwing point data, the actual static point of the curling stone, and deviation data;

[0035] The image analysis module is used to analyze the curling throwing process recorded by the camera, preprocess the images, identify the feature points of the curling stone, and obtain the movement trajectory of the curling stone.

[0036] The model construction module is used to obtain a number of groups of historical data of the curling stone, divide the historical data of the curling stone into an 80% parameter training set and a 20% parameter test set, input the 80% parameter training set into the neural network model for training, and input the 20% parameter test set into the neural network model for testing, so as to construct the neural network model.

[0037] The coordinate prediction module is used to obtain the curling throwing point data and import it into the curling movement neural network model to predict the stationary point of the curling stone.

[0038] The deviation analysis module is used to obtain the actual stationary point of the curling stone and analyze the deviation displacement and deviation angle between the actual stationary point and the predicted stationary point.

[0039] The anomaly location module is used to obtain the throwing point data and deviation data and import them into the curling anomaly location neural network model to obtain the location of the curling anomaly.

[0040] Preferably, the control module is used to control the operation of the data acquisition module, the image analysis module, the coordinate prediction module, the model construction module, the deviation analysis module and the anomaly location module.

[0041] An electronic device includes: a processor and a memory. Among them, a computer program that can be called by the processor is stored in the memory, and the processor executes the above-mentioned curling movement state evaluation method based on three-dimensional image dynamic analysis by calling the computer program stored in the memory.

[0042] A computer-readable storage medium stores instructions, and when the instructions run on a computer, the computer executes the above-mentioned curling movement state evaluation method based on three-dimensional image dynamic analysis.

[0043] The beneficial effects of the present invention are as follows: The present invention collects the curling throwing point data; obtains the curling throwing point data, and predicts the stationary point of the curling stone according to the curling prediction model; obtains the actual stationary point of the curling stone, and analyzes the deviation between the actual stationary point and the predicted stationary point; obtains the throwing point data and deviation data, and obtains the location of the curling anomaly according to the curling anomaly location model. The present invention predicts the stationary point of the curling stone by training a neural network, and then analyzes the location of the curling anomaly point according to the deviation of the curling stone, which is beneficial to timely and quickly discover the curling stones with anomalies, so as to reduce the impact of abnormal curling stones on the competition or training. Description of the Drawings

[0044] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0045] Figure 1 Schematic diagram of the flow of the curling motion state evaluation method based on three-dimensional image dynamic analysis of the present invention;

[0046] Figure 2 Schematic diagram of the flow of step one of the curling motion state evaluation method based on three-dimensional image dynamic analysis of the present invention;

[0047] Figure 3 Simulation diagram of the curling competition venue of the present invention;

[0048] Figure 4 Schematic diagram of the marking of the center point of the curling of the present invention;

[0049] Figure 5 Schematic diagram of the overall framework of the curling motion state evaluation system based on three-dimensional image dynamic analysis of the present invention. Detailed implementation manners

[0050] Now, the exemplary embodiments will be described more comprehensively with reference to the accompanying drawings. However, the exemplary embodiments can be implemented in various forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that the present invention will be more comprehensive and complete, and the concept of the exemplary embodiments will be fully conveyed to those skilled in the art. The described features, structures, or characteristics can be combined in any suitable manner in one or more embodiments.

[0051] Embodiment 1

[0052] Please refer to Figures 1-4 , an embodiment provided by the present invention: a curling motion state evaluation method based on three-dimensional image dynamic analysis, which includes the following specific steps:

[0053] Step 1. Collect data of the curling throwing point;

[0054] In this embodiment, step 1 includes the following specific steps:

[0055] Step 1-1. Collect data of the curling throwing point, and the throwing point data includes rotational speed and initial velocity;

[0056] Step 1-2. Use a camera to record the curling throwing process, automatically identify the feature points on the curling through an image recognition algorithm, and establish the three-dimensional motion trajectory of the curling according to the tracking of the feature points;

[0057] It should be noted that when importing video data recorded by a high-speed camera using video processing software, due to the camera's sensor or environmental factors, the video data may contain noise. First, use a filter to remove the noise and improve the image quality. Then, separate the curling stone and the ice surface from the image (which can be achieved by setting thresholds, using color segmentation, or background subtraction algorithms). Enhance the image features by adjusting the contrast and brightness, etc., to make the feature points more obvious. Use an image recognition algorithm to automatically identify the feature points on the curling stone (such as edge detection, corner detection, or recognition of specific colors / shapes). Use a tracking algorithm (such as the Kalman filter, optical flow method, etc.) to track the feature points in consecutive frames, connect the positions of the feature points tracked in each frame into a trajectory, and obtain the motion trajectory of the feature points on the two-dimensional image plane;

[0058] Use a calibration board (such as a checkerboard) to calibrate the camera, obtain the internal parameters (focal length, principal point, etc.) and external parameters (position and orientation) of the camera through the calibration process, and convert the two-dimensional image coordinates into accurate three-dimensional space coordinates according to the calibration results; Compare the images from different camera perspectives to determine the positions of the feature points in three-dimensional space, and use a matching algorithm (such as block matching, feature matching, or deep learning) to find the corresponding feature points from different perspectives; According to the calibration parameters of the camera and the stereo matching results, use the principle of triangulation to calculate the three-dimensional coordinates of the feature points; Connect the calculated three-dimensional coordinates of the feature points into a trajectory to reconstruct the motion trajectory of the curling stone, and use a smoothing algorithm (such as spline interpolation) to optimize the representation of the trajectory and reduce the noise and errors that may occur during the data acquisition process.

[0059] Step 13: Set the release coordinate of the curling stone before throwing as Marking Point 1, mark the release moment of the curling stone, obtain the coordinate of the curling stone on the three-dimensional motion trajectory at the next set moment, set it as Marking Point 2, and calculate the linear velocity of Marking Point 1 based on the distance and time interval between Marking Point 1 and Marking Point 2. Among them, the calculation formula for the linear velocity is: Among them, d is the distance between Marking Point 1 and Marking Point 2, and t is the time interval between the release moment and the next set moment; Obtain the distance from Marking Point 1 to the center point of the curling stone to calculate the angular velocity of Marking Point 1. Among them, the calculation formula for the angular velocity is: Among them, r is the distance from Marking Point 1 to the center point of the curling stone; Obtain the rotation speed of the curling stone at the release moment based on the angular velocity;

[0060] It should be noted that multiple synchronized high-speed cameras are used to record the curling throwing and sliding processes, ensuring that the cameras cover the entire path of the curling movement and can capture the rotation of the curling from different angles; the collected video data is preprocessed, including denoising, background elimination, etc.; image recognition algorithms are used to identify and track the feature points on the curling, and using the calibration information of the cameras, the two-dimensional image coordinates are converted into three-dimensional space coordinates, and by tracking the marked points, the three-dimensional movement trajectory of the curling is established; the center point of the curling is marked as the rotation center, and the way to obtain the center point of the curling is as follows: obtain the top view of the curling, draw the circumscribed square of the curling, and obtain the intersection point of the two diagonals of the circumscribed square as the center point of the curling, as specifically shown in Figure 4 shown;

[0061] Example: If the camera tracks that the release coordinate of the curling has moved 0.2 meters within 1 second, and the distance from the release coordinate of the curling to the rotation center is 0.1 meter, then the linear velocity v = 0.2 m / s, and the angular velocity ω = 0.2 m / s / 0.1 m = 2 rad / s. Converted to rotational speed, RPM = 2 / (2π) * 60 ≈ 19.1.

[0062] Step 14. Calculate the initial velocity of the curling according to the displacement and time interval from marker point 1 to marker point 2, where the formula for the initial velocity is: where Δx is the displacement from marker point 1 to marker point 2.

[0063] It should be noted that the release coordinate of the curling before throwing is determined, and the moment when the curling leaves the hand is marked, that is, the instant when the curling starts to slide. A short movement trajectory after the curling leaves the hand is selected for analysis (within this trajectory, the acceleration of the curling can be considered constant, or the frictional force and air resistance on the curling are small and can be ignored).

[0064] Example: It is known from video analysis that the curling has moved 1 meter within the first 0.5 seconds after leaving the hand, then the initial velocity is: v0 = 1 m / 0.5 s = 2 m / s;

[0065] The initial velocity can also be obtained through the formula where s is the distance the curling has moved, m is the weight of the curling, and f is the frictional force on the curling (the frictional force and air resistance on the curling are small and can be ignored within this trajectory).

[0066] The weight of the curling affects the sliding distance. The greater the mass of the curling, the farther the sliding distance. Generally, the standard weight of the curling (including the handle and the plug) is 19.96 kg. The curling is made of Scottish natural granite without mica, with a circumference of about 91.44 cm, a height (from the bottom to the top) of 11.43 cm, a standard diameter of 29 cm, and a thickness of 11.5 cm.

[0067] The way to obtain the weight of the curling stone is as follows: directly measure the weight of the curling stone using an accurate scale (such as an electronic scale or a mechanical scale), or find its weight on the bottom or label of the curling stone (the manufacturer will mark the weight of the curling stone to ensure compliance with the competition standards);

[0068] It should be noted that as Figure 3 shown, the way to establish the coordinate system in this embodiment is: taking the intersection of the front delivery line and the center line of the curling competition venue as the coordinate origin, taking the front delivery line as the abscissa, and taking the center line as the ordinate to construct the coordinate system.

[0069] Step 2: Obtain the data of the curling stone throwing point, and predict the stationary point of the curling stone according to the curling stone prediction model;

[0070] In this embodiment, Step 2 includes the following specific steps:

[0071] Step 2-1: Obtain several groups of historical throwing data of the curling stone. The historical throwing data includes the historical release coordinates, historical rotation speed, historical initial speed, and historical actual stationary point. Construct a curling stone motion neural network model with the release coordinates, rotation speed, and initial speed of the curling stone as the input and the stationary point of the curling stone as the output. Divide the historical throwing data of the curling stone into an 80% parameter training set and a 20% parameter test set. Input the 80% parameter training set into the curling stone motion neural network model for training to obtain the initial curling stone motion neural network model. Input the 20% parameter test set into the initial curling stone motion neural network model for testing, and output the optimal initial curling stone motion neural network model with the highest accuracy in judging the stationary point of the curling stone as the curling stone motion neural network model;

[0072] Step 2-2: The output strategy formula of specific neurons in the curling stone motion neural network model is:

[0073]

[0074] where, is the abscissa output of the stationary point of the N neurons in the n + 1 layer, is the ordinate output of the stationary point of the N neurons in the n + 1 layer, σ() is the Sigmoid activation function, is the connection weight between the k-th neuron in the n-th layer and the N neurons in the n + 1 layer, K is the number of neurons in the n-th layer, is the release abscissa input of the k-th neuron in the n-th layer, is the release ordinate input of the k-th neuron in the n-th layer, is the rotation speed input of the k-th neuron in the n-th layer, is the initial speed input of the k-th neuron in the n-th layer, is the bias of the linear relationship between the k-th neuron in the n-th layer and the N neurons in the n + 1 layer;

[0075] Step 2.3: Input the release coordinates, rotation speed, and initial velocity of the curling stone into the curling stone motion neural network model to output the predicted stationary point coordinates.

[0076] Step 3: Obtain the actual stationary point of the curling stone and analyze the deviation between the actual stationary point and the predicted stationary point;

[0077] In this embodiment, Step 3 includes the following specific steps:

[0078] Step 3.1: Obtain the coordinates of the actual stationary point and the predicted stationary point of the curling stone;

[0079] Step 3.2: Calculate the offset displacement of the curling stone according to the offset displacement calculation formula. The offset displacement calculation formula is: where x sj is the abscissa of the actual stationary point, x yc is the abscissa of the predicted stationary point, y sj is the ordinate of the actual stationary point, and y yc is the ordinate of the predicted stationary point;

[0080] Step 3.3: Calculate the offset angle of the curling stone according to the offset angle calculation formula. The offset angle calculation formula is: where arctan is the arctangent function.

[0081] Step 4: Obtain the throwing point data and deviation data, and obtain the positioning of the curling stone anomaly according to the curling stone anomaly positioning model;

[0082] In this embodiment, Step 4 includes the following specific steps:

[0083] Step 4.1: Obtain several groups of historical deviation data, historical rotation speed, and historical initial velocity of the curling stone. The historical deviation data includes historical offset displacement and historical offset angle. Construct a curling stone anomaly positioning neural network model with the rotation speed, initial velocity, offset displacement, and offset angle of the curling stone as the input and the curling stone anomaly positioning point as the output. Divide the historical deviation data, historical rotation speed, and historical initial velocity of the curling stone into an 80% parameter training set and a 20% parameter test set. Input the 80% parameter training set into the curling stone anomaly positioning neural network model for training to obtain the initial curling stone anomaly positioning neural network model. Input the 20% parameter test set into the initial curling stone anomaly positioning neural network model for testing, and output the optimal curling stone anomaly positioning initial neural network model with the highest positioning judgment accuracy for the curling stone anomaly point as the curling stone anomaly positioning neural network model;

[0084] Step 4.2: The output strategy formula of specific neurons in the curling stone anomaly positioning neural network model is:

[0085]

[0086] Among them, is the abscissa output of the curling anomaly point of the I-th neuron in the (i + 1)-th layer, is the ordinate output of the curling anomaly point of the I-th neuron in the (i + 1)-th layer, is the connection weight between the p-th neuron in the i-th layer and the I-th neuron in the (i + 1)-th layer, and P is the number of neurons in the i-th layer, is the offset displacement input of the p-th neuron in the i-th layer, is the offset angle input of the p-th neuron in the i-th layer, is the rotational speed input of the p-th neuron in the i-th layer, V i p is the initial velocity input of the p-th neuron in the i-th layer, is the bias of the linear relationship between the p-th neuron in the i-th layer and the I-th neuron in the (i + 1)-th layer;

[0087] Step Four Three: Input the offset displacement, offset angle, rotational speed, and initial velocity of the curling into the curling anomaly localization neural network model, and output the coordinates of the curling anomaly point.

[0088] Embodiment 2

[0089] Please refer to Figure 5 , a curling motion state evaluation system based on three-dimensional image dynamic analysis, which is used to implement the curling motion state evaluation method based on three-dimensional image dynamic analysis, including a data acquisition module, an image analysis module, a coordinate prediction module, a model construction module, a deviation analysis module, an anomaly localization module, and a control module;

[0090] In this embodiment, the data acquisition module is used to collect curling throwing point data, the actual static point of the curling, and deviation data;

[0091] The image analysis module is used to analyze the curling throwing process recorded by the camera, preprocess the image, identify the feature points of the curling, and obtain the motion trajectory of the curling;

[0092] The model construction module is used to obtain several groups of historical data of the curling, divide the historical data of the curling into an 80% parameter training set and a 20% parameter test set, input the 80% parameter training set into the neural network model for training, and input the 20% parameter test set into the neural network model for testing, so as to construct the neural network model;

[0093] The coordinate prediction module is used to obtain the curling throwing point data and import it into the curling motion neural network model to predict the static point of the curling;

[0094] The deviation analysis module is used to obtain the actual static point of the curling and analyze the deviation displacement and deviation angle between the actual static point and the predicted static point;

[0095] The anomaly localization module is used to obtain the throwing point data and deviation data, import them into the curling anomaly localization neural network model, and obtain the localization of curling anomalies.

[0096] In this embodiment, the control module is used to control the operation of the data acquisition module, image analysis module, coordinate prediction module, model construction module, deviation analysis module, and anomaly localization module.

[0097] Embodiment 3

[0098] This embodiment provides an electronic device, including: a processor and a memory. Among them, a computer program that can be called by the processor is stored in the memory. The processor executes the above-mentioned curling motion state evaluation method based on three-dimensional image dynamic analysis by calling the computer program stored in the memory.

[0099] This electronic device may have relatively large differences due to different configurations or performances, and can include one or more processors (Central Processing Units, CPU) and one or more memories. Among them, at least one computer program is stored in the memory, and this computer program is loaded and executed by the processor to implement the curling motion state evaluation method based on three-dimensional image dynamic analysis provided by the above method embodiment. This electronic device can also include other components for realizing the functions of the device. For example, this electronic device can also have components such as wired or wireless network interfaces and input / output interfaces for inputting and outputting data. This embodiment will not be elaborated here.

[0100] Embodiment 4

[0101] This embodiment proposes a computer-readable storage medium storing instructions. When the computer program runs on a computer device, it enables the computer device to execute the above-mentioned curling motion state evaluation method based on three-dimensional image dynamic analysis.

[0102] For example, the computer-readable storage medium can be a Read-Only Memory (ROM for short), Random Access Memory (RAM for short), Compact Disc Read-Only Memory (CD-ROM for short), magnetic tape, floppy disk, and optical data storage device, etc.

[0103] It should be understood that in various embodiments of the present invention, the magnitudes of the serial numbers of the above processes do not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.

[0104] In the description of this specification, the description referring to terms such as "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in connection with that embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in a suitable manner in any one or more embodiments or examples.

Claims

1. A method for evaluating the state of a curling stone based on dynamic analysis of three-dimensional images, characterized in that, It includes the following specific steps: Step 1: Collect data of the curling throwing point; Step 2: Obtain the data of the curling throwing point, and predict the static point of the curling according to the curling prediction model; Step 3: Obtain the actual static point of the curling, and analyze the deviation between the actual static point and the predicted static point; Step 4: Obtain the throwing point data and deviation data, and obtain the positioning of the curling anomaly according to the curling anomaly positioning model; The said Step 4 includes the following specific steps: Step 4-1: Obtain several groups of historical offset data, historical rotation speed and historical initial speed of the curling. The historical offset data includes historical offset displacement and historical offset angle. Construct a curling anomaly positioning neural network model with the rotation speed, initial speed, offset displacement and offset angle of the curling as the input and the curling anomaly positioning point as the output. Divide the historical offset data, historical rotation speed and historical initial speed of the curling into an 80% parameter training set and a 20% parameter test set. Input the 80% parameter training set into the curling anomaly positioning neural network model for training to obtain an initial curling anomaly positioning neural network model. Input the 20% parameter test set into the initial curling anomaly positioning neural network model for testing, and output the optimal initial curling anomaly positioning neural network model with the highest judgment accuracy for the curling anomaly point positioning as the curling anomaly positioning neural network model; Step 4-2: The output strategy formula of specific neurons in the curling anomaly positioning neural network model is: Among them, is the abscissa output of the curling anomaly point of the I-th neuron in the (i + 1)-th layer, is the ordinate output of the curling anomaly point of the I-th neuron in the (i + 1)-th layer, is the connection weight between the p-th neuron in the i-th layer and the I-th neuron in the (i + 1)-th layer, where P is the number of neurons in the i-th layer, is the offset displacement input of the p-th neuron in the i-th layer, is the offset angle input of the p-th neuron in the i-th layer, is the rotational speed input of the p-th neuron in the i-th layer, is the initial velocity input of the p-th neuron in the i-th layer, is the bias of the linear relationship between the p-th neuron in the i-th layer and the I-th neuron in the (i + 1)-th layer; Step 4-3: Input the offset displacement, offset angle, rotation speed and initial speed of the curling into the curling anomaly positioning neural network model, and output the coordinates of the curling anomaly point.

2. The method for evaluating the curling motion state based on three-dimensional image dynamic analysis according to claim 1, wherein The said Step 2 includes the following specific steps: Step 2-1: Obtain several groups of historical throwing data of the curling. The historical throwing data includes historical release coordinates, historical rotation speed, historical initial speed and historical actual static point. Construct a curling motion neural network model with the release coordinates, rotation speed and initial speed of the curling as the input and the static point of the curling as the output. Divide the historical throwing data of the curling into an 80% parameter training set and a 20% parameter test set. Input the 80% parameter training set into the curling motion neural network model for training to obtain an initial curling motion neural network model. Input the 20% parameter test set into the initial curling motion neural network model for testing, and output the optimal initial curling motion neural network model with the highest judgment accuracy for the static point of the curling as the curling motion neural network model; Step 2-2: The output strategy formula of specific neurons in the curling motion neural network model is: Among them, is the abscissa output of the stationary point of the N neurons in the n+1 layer, is the ordinate output of the stationary point of the N neurons in the n+1 layer, and σ() is the Sigmoid activation function. is the connection weight between the k-th neuron in the n-th layer and the N neurons in the n+1 layer, and K is the number of neurons in the n-th layer. is the off-hand abscissa input of the k-th neuron in the n-th layer, is the off-hand ordinate input of the k-th neuron in the n-th layer, is the rotational speed input of the k-th neuron in the n-th layer, is the initial velocity input of the k-th neuron in the n-th layer, is the bias of the linear relationship between the k-th neuron in the n-th layer and the N neurons in the n+1 layer; Step 2-3: Input the release coordinates, rotation speed and initial speed of the curling into the curling motion neural network model, and output the predicted static point coordinates.

3. The method for evaluating the curling motion state based on three-dimensional image dynamic analysis according to claim 2, wherein The said Step 3 includes the following specific steps: Step 3-1: Obtain the coordinates of the actual static point and the predicted static point of the curling; Step 3-2: Calculate the offset displacement of the curling according to the deviation displacement calculation formula. The deviation displacement calculation formula is: where x sj is the abscissa of the actual stationary point, x yc is the abscissa of the predicted stationary point, y sj is the ordinate of the actual stationary point, y yc is the ordinate of the predicted stationary point; Step 3-3: Calculate the offset angle of the curling according to the deviation angle calculation formula. The deviation angle calculation formula is: Wherein, arctan is the arctangent function.

4. The method for evaluating the curling motion state based on three-dimensional image dynamic analysis according to claim 3, wherein, The said Step 1 includes the following specific steps: Step 1-1: Collect data of the curling throwing point. The throwing point data includes rotation speed and initial speed; Steps 1 and 2: Use a camera to record the curling throwing process, automatically identify the feature points on the curling through an image recognition algorithm, and establish the three-dimensional motion trajectory of the curling based on the tracking of the feature points; Step 13: Set the release coordinate of the curling stone before throwing as Marking Point 1, mark the release moment of the curling stone, obtain the coordinate of the curling stone on the three-dimensional motion trajectory at the next set moment, set it as Marking Point 2, and calculate the linear velocity of Marking Point 1 based on the distance and time interval between Marking Point 1 and Marking Point 2. Among them, the calculation formula of the linear velocity is: Among them, d is the distance between Marking Point 1 and Marking Point 2, and t is the time interval between the release moment and the next set moment; obtain the distance from Marking Point 1 to the center point of the curling stone to calculate the angular velocity of Marking Point 1. Among them, the calculation formula of the angular velocity is: Among them, r is the distance from Marking Point 1 to the center point of the curling stone; obtain the rotation speed of the curling stone at the release moment according to the angular velocity; Step 14: Calculate the initial velocity of the curling stone based on the displacement and time interval from Marking Point 1 to Marking Point 2. The formula for the initial velocity is as follows: where Δx is the displacement from Marking Point 1 to Marking Point 2.

5. A curling motion state evaluation system based on three-dimensional image dynamic analysis, which is used to implement the curling motion state evaluation method based on three-dimensional image dynamic analysis as described in any one of claims 1-4, and is characterized in that, It includes a data acquisition module, an image analysis module, a coordinate prediction module, a model construction module, a deviation analysis module, an anomaly localization module, and a control module; The data acquisition module is used to collect curling throwing point data, the actual stationary point of the curling, and deviation data; The image analysis module is used to analyze the curling throwing process recorded by the camera, preprocess the image, identify the feature points of the curling, and obtain the motion trajectory of the curling; The model construction module is used to obtain several groups of historical data of the curling, divide the historical data of the curling into an 80% parameter training set and a 20% parameter test set, input the 80% parameter training set into the neural network model for training, and input the 20% parameter test set into the neural network model for testing, so as to construct a neural network model; The coordinate prediction module is used to obtain the curling throwing point data and import it into the curling motion neural network model to predict the stationary point of the curling; The deviation analysis module is used to obtain the actual stationary point of the curling and analyze the deviation displacement and deviation angle between the actual stationary point and the predicted stationary point; The anomaly localization module is used to obtain the throwing point data and deviation data and import them into the curling anomaly localization neural network model to obtain the localization of the curling anomaly; The control module is used to control the operation of the data acquisition module, the image analysis module, the coordinate prediction module, the model construction module, the deviation analysis module, and the anomaly localization module.

6. An electronic device, characterized in that, It includes: A memory and a processor. Among them, the memory stores a computer program that can be called by the processor. The processor executes the curling motion state evaluation method based on three-dimensional image dynamic analysis according to any one of claims 1-4 by calling the computer program stored in the memory.

7. A computer-readable storage medium, characterized in that: Instructions are stored. When the instructions run on a computer, the computer is made to execute the curling motion state evaluation method based on three-dimensional image dynamic analysis according to any one of claims 1-4.

Images