Smart baseball batting motion analysis system and method

Abstract

A smart baseball batting motion analysis system includes: an inertial sensor, an image capture device, a bat swing analysis server and a batting posture analysis server. The bat swing analysis server includes a swing phased analysis unit and a swing index analysis unit. The swing phased analysis unit is used for generating a swing phased data based on the inertial data obtained when a batter swings a bat, and the swing index analysis unit is used for generating a swing index data accordingly. The batting posture analysis server includes a posture key point analysis unit and a posture index analysis unit. The posture key point analysis unit is for generating a posture key point data of the batter based on the inertial data and a corresponding batting image data. The posture index analysis unit is for generating the posture index data of the batter based on the swing phased data, the inertia data and the posture key point data to determine the batter's batting performance.

Description

Intelligent Baseball Hitting Motion Analysis System and Method This disclosure relates to a smart baseball batting motion analysis system and method. Traditionally, baseball has relied heavily on the accumulation of experience among athletes, who then refine their form and improve their hitting performance through self-study or coaching advice. However, sometimes a lack of personnel or the inability to find practice partners promptly can prevent immediate corrections during self-training, leading to repeated use of incorrect postures. Even with guidance, subtle adjustments may be difficult to pinpoint, requiring repeated trial and error. All these factors contribute to poor training effectiveness. Therefore, obtaining immediate analysis and evaluation of movements can significantly improve training efficiency and results. This disclosure presents a smart baseball batting motion analysis system, comprising: a plurality of inertial sensors, a batting swing analysis server, an image capturing device, and a batting posture analysis server. The plurality of inertial sensors are used to sense and generate a plurality of inertial data corresponding to the batter's swing. The batting swing analysis server, communicatively connected to the plurality of inertial sensors, includes: a swing phase analysis unit, used to phase the swing based on the plurality of inertial data to generate swing phase data. The image capturing device is used to capture and generate batting image data of the batter during the swing. The batting posture analysis server, communicatively connected to the plurality of inertial sensors, the batting swing analysis server, and the image capturing device, includes: a posture key point analysis unit and a posture index analysis unit. The posture key point analysis unit is used to analyze and generate the batter's posture key point data based on the plurality of inertial data and batting image data. The posture index analysis unit is used to analyze and generate multiple posture index data of the batter based on swing phase data, as well as at least one of multiple inertial data and posture key point data, and to judge the batter's batting performance. In one embodiment, the attitude keypoint analysis unit includes: a motion feature extraction module for extracting motion feature data from the strike image data; a two-dimensional keypoint analysis module for generating two-dimensional keypoint data of the striker based on the motion feature data using a two-dimensional keypoint analysis model; a three-dimensional attitude angle analysis module for calculating and generating three-dimensional attitude angle data of the striker based on a plurality of inertial data; and a three-dimensional keypoint analysis module for generating three-dimensional keypoint data of the striker as attitude keypoint data based on the two-dimensional keypoint data and the three-dimensional attitude angle data using a three-dimensional keypoint analysis model. In one embodiment, the attitude keypoint data includes the three-dimensional spatial coordinates of at least one joint of the striker. In one embodiment, the plurality of inertial data used to generate attitude key point data corresponds to the plurality of limbs of the striker. In one embodiment, the plurality of posture index data includes at least one of the following: elbow flexion angle, maximum elbow extension angular velocity, upper torso rotation angle, maximum upper torso rotation angular velocity, pelvic rotation angle, maximum pelvic rotation angular velocity, knee flexion angle, maximum knee extension angular velocity, stride length, and impact kinetic chain. In one embodiment, the plurality of inertial data used to generate swing phase data correspond to at least one of the bat used by the batter and a plurality of the batter's limbs. In one embodiment, the swing phase data includes dividing the swing into at least one of an initial standoff, a preparatory phase, a step phase, a swing phase, a post-swing phase, and an end-standoff phase. In one embodiment, the intelligent baseball batting motion analysis system further includes: a swing index analysis unit, used to analyze and generate multiple swing index data of the batter based on swing phase data and multiple inertial data. In one embodiment, the plurality of swing index data includes at least one of the following: swing trajectory, time of impact, maximum angular velocity of the bat at the top, linear velocity of the bat at the top, maximum linear velocity of the grip, vertical angle of the bat, and angle of attack. This disclosure further proposes a smart baseball batting motion analysis method, which includes: sensing and generating multiple inertial data corresponding to the batter's swing; dividing the swing into phases based on the multiple inertial data to generate swing phase data; capturing and generating batting video data of the batter during the swing; analyzing and generating the batter's posture key point data based on the multiple inertial data and batting video data; and analyzing and generating multiple posture index data of the batter based on the swing phase data and at least one of the multiple inertial data and posture key point data, and judging the batter's batting performance accordingly. The embodiments disclosed herein are discussed in detail below. However, it will be understood that the embodiments provide many applicable concepts that can be implemented in a wide variety of specific situations. The discussed and disclosed embodiments are for illustrative purposes only and are not intended to limit the scope of this disclosure. Figure 1 is a schematic diagram of a smart baseball batting motion analysis system 100 according to an embodiment of this disclosure. As shown in Figure 1, the smart baseball batting motion analysis system 100 includes a batting swing analysis server 110, a batting posture analysis server 120, multiple inertial sensors 130, an image capturing device 140, and a batting motion analysis database 150. The batting swing analysis server 110 is communicatively connected to the multiple inertial sensors 130, while the batting posture analysis server 120 is communicatively connected to the multiple inertial sensors 130, the batting swing analysis server 110, and the image capturing device 140. In one embodiment, both the batting swing analysis server 110 and the batting posture analysis server 120 are communicatively connected to a batting motion analysis database 150. Furthermore, the batting swing analysis server 110 includes a swing phase analysis unit 111 and a swing index analysis unit 112, while the batting posture analysis server 120 includes a posture key point analysis unit 121 and a posture index analysis unit 122. The intelligent baseball batting motion analysis system 100 uses multiple inertial sensors 130 and an image capturing device 140 to collect relevant sensing data of the batter during the swing, and transmits it to the batting swing analysis server 110 and the batting posture analysis server 120 for analysis, and stores the analysis results in the batting motion analysis database 150. In one embodiment, the intelligent baseball batting motion analysis system 100 can also be accessed and displayed by other devices such as mobile devices, or the analysis results can be accessed by other servers for further data analysis. Figure 2 is a schematic diagram showing the installation positions of the inertial sensor 130 and the image capturing device 140 in one embodiment of this disclosure. As shown in Figure 2(a), the inertial sensor 130 can be installed on the bottom of the bat B and on the batter's body H, such as the left and right lower legs, left and right thighs, left and right upper arms, left and right forearms, pelvis, and back. The inertial sensor 130 is only required to measure the changes in the movement of the bat B and the body H when the batter swings the bat, and its position and number are not limited to these. The image capturing device 140 is used to capture the action image of the batter swinging the bat and generate batting image data. As shown in Figure 2(b), while the inertial sensor 130 is installed on the batter's body H, the image capturing device 140 is set up in front of the batter and continuously captures the action image of the batter swinging the bat. The image capturing device 140 is only required to capture the complete action image of the batter swinging the bat, and its installation position and number are not limited to these. Figure 3 is a flowchart of a smart baseball batting motion analysis method 300 according to an embodiment of this disclosure. The smart baseball batting motion analysis method 300 of this disclosure will be described below with reference to Figures 1 to 3. First, step S310 is performed, where the inertial sensor 130 senses the inertial data of the batter during the swing. As shown in Figure 2(a), the inertial sensor 130 can be installed on multiple limbs of the batter's body H, or on the bat B. In one embodiment, step S310 can also be performed using only the inertial sensor 130 at the pelvis of the bat B and the body H, and the inertial data can be transmitted to the batting swing analysis server 110. Next, step S320 is performed, where the swing phase analysis unit 111 phases the swing based on the inertial data to generate swing phase data. Step S320, swing phasing, is used to segment the signal during the swing-hitting process from before the swing to the stationary period after the swing, allowing for more detailed analysis of each action during the swing-hitting process. In one embodiment, inertial data can be obtained by correcting the signals from the accelerometer, gyroscope, and magnetometer of the inertial sensor 130 and filtering out high noise and motion noise through data processing, such as using a low-pass filter; this disclosure is not particularly limited in this regard. After obtaining inertial data suitable for swing-hitting phase analysis, the swing-hitting phase is determined by the resultant force of the three-axis acceleration and the resultant force of the three-axis angular velocity in the inertial data, as explained below. Figure 4 is a schematic diagram of the swing and hitting phases in one embodiment of this disclosure. In one embodiment of this disclosure, the swing and hitting process is divided into six phases (ST1~ST6) with five phase points (P1~P5), in the following order: initial stationary phase ST1, preparatory phase start point P1, preparatory phase ST2, lead foot off the ground P2, step phase ST3, lead foot landing P3, swing phase ST4, impact point P4, afterburner phase ST5, afterburner end point P5, and end stationary phase ST6. Starting from the initial stationary phase ST1, the preparatory phase start point P1 is defined as the starting point of the swing, which is the instant shown in the inertial data that the batter begins to move. This marks the beginning of the preparatory phase ST2, during which the batter assumes a ready stance and makes slight adjustments to their body angle. Next, the batter prepares to swing, at which point the lead foot leaves the ground. This is the phase point, P2, where the lead foot leaves the ground. Since the body's center of gravity shifts towards the rear foot when the lead foot leaves the ground, the moment the first peak of the X-direction acceleration signal in the pelvic inertial data is taken as P2. Here, the X-direction is towards the pitcher, i.e., the direction the ball approaches; however, the direction can be defined as needed, and this disclosure does not impose any particular restrictions. After the lead foot leaves the ground at P2, the stepping phase ST3 begins. This is the period after the lead foot leaves the ground and the body's center of gravity shifts briefly to the rear foot, during which the batter gradually transfers their weight forward to the lead foot to increase hitting momentum. When the center of gravity shifts forward and the leading foot touches the ground, the moment when the minimum value of the X-direction acceleration signal in the inertial data of the pelvis is defined as the leading foot landing P3, and then the swing phase ST4 begins. The swing phase (ST4) is the period from when the batter swings the bat B until it makes contact with the ball. During ST4, the batter sequentially forms a kinetic chain from the knees, waist, upper torso, and hands to transfer the swing energy to the bat B. When the bat B makes contact with the ball, its angular velocity usually reaches its maximum value. Therefore, the moment when the resultant force of the three-axis angular velocities of the bat B reaches its maximum value in the inertial data is defined as the impact point (P4), followed by the aftermath phase (ST5). During the aftermath phase (ST5), the bat B continues to move along the swing path, while its swing velocity gradually decreases to zero. Therefore, the moment when the bat B shows itself stationary in the inertial data is defined as the end of the aftermath phase (P5), followed by the end-of-swing phase (ST6), at which point the batter performs the finishing swing motion, and the swing is complete. This embodiment uses the sensing data from the inertial sensors 130 at the pelvis and bat B as the basis for batting swing periodization. However, more inertial sensors 130 and further analysis can be used to obtain more refined periodization results. Furthermore, a single periodization point or period can be selected as the analysis target as needed. As long as the inertial sensors 130 can be used to periodize this process, this disclosure does not particularly limit the location and number of inertial sensors 130, the number of periods, or the method of determining periodization points. On the other hand, in step S330, the image capturing device 140 captures the striking image data of the batter swinging the bat. Here, the capture of the batter's image is performed simultaneously with the sensing of inertial data so that the inertial data can correspond to the striking image data. However, in another embodiment, the image capturing and inertial data sensing can be performed separately and multiple times for comparison, or the order of steps can be adjusted as needed. This disclosure does not impose any particular limitation on this. In addition, the image capturing device 140 can use, for example, a global shutter camera to obtain images of the batter's rapid movement. In one embodiment, it can also be combined with an optical motion capture system to obtain more accurate human spatial coordinates. As long as the images of the batter's rapid movement during the swing can be clearly recorded and used for subsequent analysis, the captured striking image data can include dynamic or static images, and the camera used is not limited to this. After the batter completes the swinging motion, the battering posture analysis server 120 receives the battering image data from the image capturing device 140 and the inertial data from the inertial sensor 130. After processing the inertial data as described above, step S340 is performed, whereby the posture key point analysis unit 121 analyzes and generates the batter's posture key point data based on the inertial data and the battering image data. Figure 5 is a schematic diagram of the positions of inertial sensing points I and attitude key points K in one embodiment of this disclosure. In Figure 5(a), the batter's body H has 10 inertial sensing points I at the corresponding positions of the inertial sensor 130, as shown in Figure 5(b). In this embodiment, after the attitude key point analysis unit 121 analyzes the inertial data and batting image data, 16 attitude key points K of the body H and their corresponding three-dimensional spatial coordinates can be obtained as attitude key point data. Among them, the attitude key points K correspond to the head, back, spine, pelvis, right wrist joint, right elbow joint, right shoulder joint, right hip joint, right knee joint, right ankle joint, left wrist joint, left elbow joint, left shoulder joint, left hip joint, left knee joint, and left ankle joint of the body H, respectively. The three-dimensional spatial coordinates are shown in Figure 2(b), where the X direction is defined as the batter's direction towards the pitcher (i.e., the direction in which the ball approaches), the Y direction is to the left, and the Z direction is upward. This disclosure takes obtaining 16 attitude key points K and their corresponding three-dimensional spatial coordinates as an example, but the number of attitude key points K, their position on the human body H, and the direction of the three-dimensional spatial coordinates can be set according to requirements, and this disclosure does not impose any special restrictions on this. After obtaining inertia data, swing phase data, and posture key point data in steps S310-S340, step S350 is performed where the posture index analysis unit 122 analyzes and generates the batter's posture index data based on the swing phase data, inertia data, and posture key point data. The posture index data includes various posture indicators that can be used to evaluate the batter's batting swing at different phases during the action. Since posture indicators generally correspond to batting performance such as swing speed and motion smoothness, obtaining the data corresponding to each posture indicator allows for the assessment and evaluation of the batter's batting swing, thereby enabling the development of improvement plans. In one embodiment, the attitude index analysis unit 122 obtains the batting swing stage based on the swing stage data, and performs attitude index analysis based on the inertia data and attitude key point data under this stage to obtain attitude index data. Table 1 shows an example of attitude indices and their definitions. The attitude indices, their corresponding stages, and the related attitude key points K and inertia data can be adjusted according to the index definitions and the desired level of detail in the evaluation. This disclosure does not impose any particular limitations on this. Table 1 After obtaining posture index data in step S350, the posture index analysis unit 122 can judge and evaluate the batter's posture during the swing based on the posture index data, and transmit it to the batting motion analysis database 150 for storage. Alternatively, the posture index data can be transmitted to a mobile device for professional evaluation by baseball professionals, and an improvement plan can be established accordingly. Since the index is related to batting performance, the corresponding batting performance can be judged by setting the trend of the index and comparing it with preset values ​​or preset ranges. For example, a larger forward elbow flexion angle and a smaller backward elbow flexion angle at the point of impact (P4) indicate greater range of motion and a faster swing speed. Larger maximum elbow extension angular velocity, upper torso rotation angle, upper torso rotation angular velocity, pelvic rotation angular velocity, and knee extension angular velocity, smaller pelvic rotation angle, and a knee flexion angle of 11°–20° may indicate better hitting performance. Furthermore, stride length can be used to assess the impact on the batter, and the correctness of the swing power can be judged by whether the batter's kinetic chain is sequentially pelvis, upper torso, shoulder, and bat B. This disclosure is not particularly limited in its application to assessing batter posture indicators. Figure 6 is a schematic diagram of the attitude key point analysis unit 121 according to an embodiment of this disclosure, and Figure 7 is a flowchart of the attitude key point data analysis method 700 according to an embodiment of this disclosure. As shown in Figure 6, the attitude key point analysis unit 121 includes a motion feature extraction module 1211, a two-dimensional key point analysis module 1212, a three-dimensional attitude angle analysis module 1213, and a three-dimensional key point analysis module 1214. The attitude key point data analysis method 700 according to an embodiment will be described in detail below with reference to Figures 6 and 7. After the strike posture analysis server 120 receives the strike image data of the striker captured by the image capturing device 140 and the inertial data of the inertial sensor 130 in step S330, step S340 is performed, in which the posture key point analysis unit 121 analyzes and generates the striker's posture key point data based on the inertial data and the strike image data. First, in step S710, the motion feature extraction module 1211 extracts motion feature data from the strike image data. The extraction of motion feature data can use convolutional neural networks (CNNs) or other methods that can extract features from images. Next, in step S720, the 2D keypoint analysis module 1212 generates 2D keypoint data of the striker based on the motion feature data using a 2D keypoint analysis model. After the motion feature data is input into the 2D keypoint analysis model, the model obtains the probability of each keypoint in the 2D image based on the motion feature data, thus obtaining 2D probability heatmaps corresponding to multiple keypoints. The image coordinates of the location where the maximum probability occurs are the 2D keypoint coordinates, which are used as the striker's 2D keypoint data. In one embodiment, a total of 16 2D keypoints corresponding to the joints of the human body can be obtained. In training the 2D keypoint analysis model, to improve the accuracy of 2D keypoint analysis, images of human 2D keypoints annotated by an optical motion capture system can be used as a standard for training. In one embodiment, the 2D keypoint analysis model can use a stacked hourglass network, a high-resolution network (HRNet), or YoLo pose estimation (You Only Look Once for Pose Estimation, YoLoPose). This disclosure is not limited to any method that can perform 2D keypoint analysis on images. On the other hand, although the posture key point K can be analyzed based solely on the image, the image may still be obscured due to changes in movement during the image acquisition process, resulting in incomplete information. Therefore, in one embodiment, inertial data can be further combined to enable perceptual fusion of visual and inertial data for more accurate analysis. Therefore, in step S730, the three-dimensional attitude angle analysis module 1213 calculates the three-dimensional attitude angle data of the striker using inertial data. In this embodiment, the inertial data includes inertial data corresponding to the limbs received by 10 inertial sensors 130 installed on the human body H. The calculation method can use an adaptive cumber Kalman filter (ACKF), an extended Kalman filter, etc., and calculate the three-dimensional quaternion attitude angles of the 10 limbs when the striker swings the bat based on this inertial data. In one embodiment, in order to perform perceptual fusion with the two-dimensional keypoint data obtained from the image, it is necessary to further convert the inertial coordinate system of the three-dimensional quaternion attitude angles to the photographic coordinate system as three-dimensional attitude angle data that can be used for perceptual fusion. Among them, steps S710-S720, which obtain two-dimensional keypoint data from the strike image data, and step S730, which obtains three-dimensional attitude angle data from the sensing data, can be performed simultaneously, and the order of the steps can be adjusted according to needs. This disclosure does not particularly limit this. After obtaining the two-dimensional keypoint data and three-dimensional attitude angle data through steps S710-S730, step S740 is performed, where the three-dimensional keypoint analysis module 1214 generates three-dimensional keypoint data of the striker as attitude keypoint data based on the two-dimensional keypoint data and three-dimensional attitude angle data using the three-dimensional keypoint analysis model. Specifically, the two-dimensional keypoint data corresponding to the 16 joints and the three-dimensional attitude angle data corresponding to the 10 limbs are input into the three-dimensional keypoint analysis model for three-dimensional keypoint analysis. Since the two-dimensional keypoint data corresponds to the joints of the human body, and the three-dimensional attitude angle data corresponds to the limbs, the three-dimensional keypoint analysis model maps the inertial data of the limbs to the joints, that is, it distributes the influence of the three-dimensional attitude angle data to the two-dimensional keypoints. In the training of the 3D keypoint analysis model, in order to obtain the coordinates of 3D keypoints corresponding to each joint of the human body as 3D keypoint data and improve the accuracy of 3D keypoint analysis, while the image capturing device 140 captures the image of the striker, an optical motion capture system can be used to obtain the standard of the striker's corresponding image coordinates (i.e., the standard image depth) as a reference for training the 3D keypoint coordinate prediction. In one embodiment, the 3D keypoint analysis model can use a Semantic Graph Convolutional Network (SEMGCN), a Videopose 3D model, or a Strided Transformer. This disclosure is not limited to any method that can perform 3D keypoint analysis on images. Furthermore, in another embodiment, when using a stacked hourglass network, high-resolution network (HRNet), or YoLo pose estimation (You Only Look Once for Pose Estimation, YoLoPose) as a two-dimensional keypoint analysis model to generate two-dimensional keypoint data in step S720, a deep regression subnetwork can be further used. This subnetwork takes the two-dimensional keypoint data as input and the image coordinates obtained by the optical motion capture system as the standard to obtain three-dimensional keypoint estimation data. This three-dimensional keypoint estimation data is then further processed in step S740 and input together with the three-dimensional pose angle data into the three-dimensional keypoint analysis model to obtain more accurate three-dimensional keypoint coordinates corresponding to each joint as three-dimensional keypoint data. Through steps S710-740, the attitude key point analysis unit 121 generates three-dimensional key point data of the striker as attitude key point data for analyzing the striker's attitude indicators. In one embodiment, since the three-dimensional key point data generated in step S740 is data in a photographic coordinate system, the three-dimensional key point data can be further converted into three-dimensional spatial coordinates as shown in Figure 2(b), with the striker's front, left, and top as the X, Y, and Z directions, and the attitude key point data in the three-dimensional spatial coordinate system is used for attitude indicator analysis. In another embodiment, in addition to the batter's posture, swing index analysis can also be performed by the swing index analysis unit 112 based on the inertial data corresponding to the bat B and the human body H at each stage to obtain swing data. Table 2 shows an example of swing indexes and definitions, in which the batting swing trajectory algorithm calculates the corresponding coordinates of the bat B in three-dimensional space and the batting swing trajectory based on the acceleration signal, angular velocity signal and magnetic force signal generated by the bat B during the swing. Table 2 Since swing indices are related to the batter's swing direction, angle, and speed when the bat is swung (B), as well as the direction and angle of the ball's flight after being hit, they can also be used as indicators to evaluate the batter's hitting performance. The swing indices, their corresponding phases, and the inertial data involved can be adjusted according to the definition of the indices and the desired level of precision in the evaluation; this disclosure does not impose any particular limitations on this. Although this disclosure has been disclosed above with reference to embodiments, it is not intended to limit this disclosure. Anyone skilled in the art may make some modifications and refinements without departing from the spirit and scope of this disclosure. Therefore, the scope of protection of this disclosure shall be determined by the appended claims. 100: Intelligent Baseball Hitting Motion Analysis System 110: Hitting Swing Analysis Server 111: Swing Phase Analysis Unit 112: Swing Index Analysis Unit 120: Hitting Posture Analysis Server 121: Posture Keypoint Analysis Unit 1211: Motion Feature Extraction Module 1212: Two-Dimensional Keypoint Analysis Module 1213: Three-Dimensional Posture Angle Analysis Module 1214: Three-Dimensional Keypoint Analysis Module 122: Posture Index Analysis Unit 130: Inertial Sensor 140: Image Capture Device 150: Hitting Motion Analysis Database 300: Smart Baseball Hitting Motion Analysis Method 700: Posture Key Point Data Analysis Method B: Bat H: Human Body I: Inertial Sensing Point K: Posture Key Point P1: Start Point of Preparation Phase P2: Leading Foot Leaves the Ground P3: Leading Foot Lands P4: Hitting Point P5: End Point of Aftermath S310~S350, S710~S740: Steps ST1: Initial Stationary Phase ST2: Preparation Phase ST3: Step Phase ST4: Swing Phase ST5: Aftermath Phase ST6: End of Stationary Phase X, Y, Z: Direction To gain a more complete understanding of the embodiments and their advantages, the following description is made in conjunction with the accompanying drawings, in which: Figure 1 is a schematic diagram of a smart baseball batting motion analysis system according to an embodiment of the present disclosure; Figure 2 is a schematic diagram of the installation positions of an inertial sensor and image capturing device according to an embodiment of the present disclosure; Figure 3 is a flowchart of a smart baseball batting motion analysis method according to an embodiment of the present disclosure; Figure 4 is a schematic diagram of swing phases according to an embodiment of the present disclosure; Figure 5 is a schematic diagram of the positions of inertial sensing points and attitude key points according to an embodiment of the present disclosure; Figure 6 is a schematic diagram of an attitude key point analysis unit according to an embodiment of the present disclosure; and Figure 7 is a flowchart of an attitude key point data analysis method according to an embodiment of the present disclosure. 100: Intelligent Baseball Hitting Analysis System 110: Batting Swing Analysis Server 111: Swing Phase Analysis Unit 112: Swing Index Analysis Unit 120: Attack Attitude Analysis Server 121: Attitude Key Point Analysis Unit 122: Attitude Index Analysis Unit 130: Inertial sensor 140: Image capturing device 150: Striking Movement Analysis Database

Claims

1. A smart baseball batting motion analysis system includes: a plurality of inertial sensors for sensing and generating a plurality of inertial data corresponding to a batter's swing; a batting swing analysis server communicatively connected to the plurality of inertial sensors, including: a swing phase analysis unit for phased swings based on the plurality of inertial data to generate swing phase data; an image capturing device for capturing and generating batting image data of the batter's swing; and a batting posture analysis server communicatively connected to the plurality of inertial sensors, the batting swing analysis server, and the image capturing device, the batting posture analysis server including: A posture key point analysis unit is used to analyze and generate posture key point data of the batter based on the plurality of inertial data and the batting image data; and a posture index analysis unit is used to analyze and generate a plurality of posture index data of the batter based on the swing phase data and at least one of the plurality of inertial data and the posture key point data, and to judge the batter's batting performance accordingly. The intelligent baseball batting motion analysis system as described in claim 1, wherein the posture keypoint analysis unit comprises: a motion feature extraction module for extracting motion feature data from the batting image data; a two-dimensional keypoint analysis module for generating two-dimensional keypoint data of the batter based on the motion feature data using a two-dimensional keypoint analysis model; a three-dimensional posture angle analysis module for calculating and generating three-dimensional posture angle data of the batter based on the plurality of inertial data; and a three-dimensional keypoint analysis module for generating three-dimensional keypoint data of the batter as the posture keypoint data based on the two-dimensional keypoint data and the three-dimensional posture angle data using a three-dimensional keypoint analysis model. The intelligent baseball batting motion analysis system as described in claim 1, wherein the posture key point data includes the three-dimensional spatial coordinates of at least one joint of the batter. The intelligent baseball batting motion analysis system as described in claim 1, wherein the plurality of inertial data used to generate the attitude key point data corresponds to the plurality of limbs of the batter. The intelligent baseball batting motion analysis system as described in claim 1, wherein the plurality of posture index data includes at least one of the following: elbow flexion angle, maximum elbow extension angular velocity, upper torso rotation angle, maximum upper torso rotation angular velocity, pelvic rotation angle, maximum pelvic rotation angular velocity, knee flexion angle, maximum knee extension angular velocity, stride length, and batting kinetic chain. The intelligent baseball batting motion analysis system as described in claim 1, wherein the plurality of inertial data used to generate the swing phase data corresponds to one of the bats used by the batter and at least one of the plurality of limbs of the batter. The intelligent baseball batting motion analysis system as described in claim 1, wherein the swing phase data includes at least one of the following: an initial stationary phase, a preparatory phase, a step phase, a swing phase, a afterburner phase, and an end stationary phase. The intelligent baseball batting motion analysis system as described in claim 1 further includes: a swing index analysis unit for analyzing and generating multiple swing index data of the batter based on the swing phase data and the plurality of inertial data. The intelligent baseball batting motion analysis system as described in claim 8, wherein the plurality of swing index data includes at least one of the following: swing trajectory, batting time, maximum angular velocity of the bat at the top of the bat, linear velocity of the bat at the top of the bat, maximum linear velocity of the grip, vertical angle of the bat, and angle of attack. A smart baseball batting motion analysis method includes: sensing and generating a plurality of inertial data corresponding to a batter's swing; dividing the swing into phases based on the plurality of inertial data to generate swing phase data; capturing and generating batting image data of the batter during the swing; analyzing and generating posture key point data of the batter based on the plurality of inertial data and the batting image data; and analyzing and generating a plurality of posture index data of the batter based on the swing phase data and at least one of the plurality of inertial data and the posture key point data, and judging the batter's batting performance accordingly.

Images