A visual and PVDF piezoelectric fusion table tennis AI automatic penalty device and method

Abstract

The application discloses a ping-pong ball AI automatic penalty device and method fusing vision and PVDF piezoelectricity, wherein the penalty device comprises a double-sided net column integrated collection component, a table surface PVDF falling point sensing module, an embedded heterogeneous SoC master control unit, a man-machine interaction terminal module and a wireless complaint peripheral module; the double-sided net column vision and double-layer PVDF piezoelectricity multimodal sensing are fused, and full-process unmanned penalty is realized by relying on the embedded edge AI; the device can be directly installed on the existing ping-pong table in the campus, can independently operate in the case of network interruption, can automatically identify the net wiping, edge wiping and falling point and report the score by voice, can be matched with the player complaint review function, can automatically store the training and competition data, and is suitable for sports course teaching and intramural ping-pong events, can reduce the pressure and economic investment of the executive staff, can perfect the digital matching of the smart campus sports venue, and can be used for major ping-pong events by further updating the system and perfecting the hardware.

Description

Technical Field

[0001] This invention belongs to the field of smart sports equipment technology, specifically relating to a table tennis AI automatic judgment device and method that integrates vision and PVDF piezoelectricity. Background Technology

[0002] Currently, table tennis competitions at all levels, as well as training sessions in schools and commercial venues, generally rely on human referees. Each match is officiated by at least one head referee, with additional sideline referees for key events. This human refereeing model has the following inherent flaws: 1) The physiological limitations of the human eye lead to frequent misjudgments and missed judgments. In the high-speed rallies of table tennis, the ball speed can reach 20m / s. Key actions such as grazing the edge, slightly touching the net, and the height of the serve toss are fleeting. The player's body and racket can easily obstruct the referee's view. The human eye cannot capture millimeter-level subtle contact. Industry data shows that the accuracy of human judgment on edge balls is significantly insufficient. Minor net touches and shots hitting the edge of the table are very easy to be misjudged.

[0003] 2) Inconsistent subjective judgment standards Referees' professional level, on-court condition, fatigue level, and personal experience all affect the judgment criteria; for ambiguous scenarios such as serve toss, net touch, and edge touch, there is no objective quantitative basis for judgment, and it all depends on subjective experience, making it difficult to guarantee the fairness of the competition.

[0004] 3) Scarcity of professional referee resources and high operating costs The training period for high-level certified referees can take several years. Campus leagues and club point games are numerous and the venues are scattered. In most amateur games, players can only score against each other, which leads to frequent disputes.

[0005] 4) The process of reviewing disputed balls is cumbersome and disrupts the rhythm of the game. Since human referees lack objective replays and sensor evidence, they can only rely on the referee's recollection when disputes arise, without reliable evidence to support their decisions.

[0006] 5) Manual methods cannot simultaneously retain complete match data. Human referees only record the score and cannot automatically collect training and analysis data such as the landing point, number of net touches, and ball toss parameters for each round.

[0007] To alleviate the pressure on human referees, the industry has developed simple scoring devices such as pure vision-based eagle-eye systems, single-plane piezoelectric sensors, and single-side cameras; however, all of these solutions have significant shortcomings: 1) A pure multi-camera Hawk-Eye visual judgment system requires the installation of 3 or more high-speed cameras on the field. The equipment is expensive, players' bodies can easily block the ball, frames are lost in strong light, there is no vibration sensor to verify the judgment, and slight net contact can easily be missed.

[0008] 2) Table tennis tables with only piezoelectric sensors can only identify the landing point of the ball, but cannot determine the height of the serve, re-serve after a net touch, or other rule-based penalties, and cannot completely replace human referees.

[0009] 3) The single-sided net post visual solution has a large blind spot. When the player turns to the side or swings his arm, he will completely block the net area, making it impossible to reliably capture the net-touch scene.

[0010] 4) The existing processing unit hardware architecture is unreasonable. Most of them use general-purpose industrial control computer CPUs, which require image acquisition cards and synchronous clock peripherals, resulting in a large size that cannot be easily scaled up.

[0011] Currently, there is a lack of integrated judgment schemes in the published literature and patents that integrate high-speed cameras and PVDF piezoelectric films for net cleaning on both sides of the net posts, PVDF landing point sensing across the entire table surface, and single heterogeneous SoC hardware to synchronize all multi-source signals. There is also a lack of automated adjudication systems that can completely replace human referees. Summary of the Invention

[0012] To address the aforementioned issues, this invention discloses a visual and PVDF piezoelectric fusion AI-powered automatic judging device and method for table tennis. This multimodal automatic judging device integrates high-speed visual acquisition from both sides of the net posts, PVDF net rubbing sensing from the net posts, and PVDF landing point sensing across the entire table surface. It is equipped with an embedded heterogeneous SoC main controller to achieve full-process AI-powered automatic judging and human-computer interaction appeals. It can replace traditional human referees to achieve unmanned, standardized, and high-precision automatic adjudication.

[0013] To achieve the above objectives, the technical solution of the present invention is as follows: A table tennis AI automatic judgment device that integrates vision and PVDF piezoelectricity comprises five core components: a dual-sided net post integrated acquisition component, a tabletop PVDF landing point sensing module, an embedded heterogeneous SoC main control unit, a human-computer interaction terminal module, and a wireless appeal peripheral module.

[0014] 1.1 Dual-sided grid post integrated data acquisition component Each grid post is a hollow modified column, integrating three major sub-units inside: High-speed camera unit An angled viewfinder is set on the inside of the pillar, with a built-in 500~1000fps miniature high-speed industrial camera. The lens is angled downward to cover the entire net area, the serve and toss area, and half of the table. The left and right pillar lenses have complementary perspectives, completely eliminating blind spots caused by the player's body and racket. The camera outputs a MIPI high-speed image stream with built-in hardware microsecond-level timestamps.

[0015] PVDF mesh cleaning piezoelectric induction unit An ultra-thin flexible PVDF piezoelectric film is placed above the net posts, and is attached to the inner side of the posts near the net cable slots, with the film tightly attached to the net cable support position. When the ball hits or rubs against the net, the net cable causes the film to deform and generate a voltage signal. A miniature charge amplifier circuit and a low-pass filter circuit are provided to filter out noise vibration signals caused by net cable swaying and court resonance, and output the net rubbing vibration amplitude and trigger timestamp.

[0016] Synchronous clock submodule It receives the global synchronization reference clock sent by the embedded heterogeneous SoC and synchronizes the camera frame signal of this column with the PVDF piezoelectric vibration signal.

[0017] 1.2 Tabletop PVDF Landing Point Sensing Module An ultra-thin PVDF piezoelectric film matrix of 0.15mm is used and set in the interlayer under the bottom or surface of a standard ping-pong table. It is divided into a left half-table sensing area and a right half-table sensing area. Edge-rubbing sensing strips are added around the table surface. Each sensing array unit is independently matched with a high-speed ADC acquisition channel to collect the X / Y two-dimensional coordinates, peak voltage, and timing of the ping-pong ball impacting the table surface in real time.

[0018] Setting shock-absorbing pads on the surface of the ultra-thin flexible PVDF piezoelectric film matrix can provide buffering and protection.

[0019] 1.3 Embedded Heterogeneous SoC Main Control Unit The device's sole core processor is a heterogeneous edge SoC chip specifically designed for table tennis officiating, which integrates: 1) Multi-core CPU processing kernel: responsible for logical operations, game score storage, peripheral drivers, and human-computer interaction scheduling; 2) NPU AI hardware acceleration core: local lightweight object detection, sphere trajectory fitting neural network inference; 3) Multi-channel integrated ADC: Simultaneously receives all simulated vibration signals from the left grid post PVDF, the right grid post PVDF, and the full-area PVDF array on the table via the main line; 4) Dual-channel MIPI image hardware decoding interface: directly connects to the image streams of the left and right high-speed cameras; 5) Global hardware timing synchronization module: outputs a unified reference clock to synchronize the timestamps of all sensor data; 6) Peripheral driver unit: integrates Bluetooth driver, voice amplifier driver, touch screen driver, and local storage control.

[0020] 1.4 Human-Computer Interaction Terminal Module 1) 7-inch narrow bezel touch screen: Normally displays the real-time score and heat map of the landing points of both sides; when a player appeals, a pop-up window simultaneously displays the replay of the left and right net posts from two perspectives and the piezoelectric vibration waveform.

[0021] 2) Voice broadcast unit: Automatically broadcasts the results of the rulings such as net touch replay, valid score, out-of-bounds, and service violation.

[0022] The embedded heterogeneous SoC main control unit and human-computer interaction terminal module are set on the side of the ping-pong table, which is convenient for observation and does not affect the athlete's performance.

[0023] 1.5 Wireless Appeal Peripheral Module It includes two independent Bluetooth wireless buttons (or other wireless communication methods), assigned to the left and right players respectively; when a player disagrees with a ruling in the current round, pressing the button will automatically lock and cache all objective supporting data for the current round.

[0024] 2. AI-based automatic judgment method Includes the following complete execution steps: S1 device power-on synchronous calibration initialization The embedded heterogeneous SoC main controller outputs a global synchronization clock to complete the timing alignment of the dual-side cameras, net post PVDF, and table PVDF sensors; the AI ​​lightweight table tennis detection model is loaded into the NPU (multi-channel signal processor) acceleration core; the table boundary and net height are automatically identified and the ITTF judgment threshold is entered.

[0025] S2 Multi-source signal synchronous parallel acquisition during game process The high-speed cameras on the left and right columns continuously output high-speed image streams with uniform timestamps; the PVDF piezoelectric films on the left and right columns continuously collect simulated signals of screen vibration; and the PVDF piezoelectric matrix at the bottom of the table continuously collects the impact coordinates, impact intensity, and impact timing of the falling ball.

[0026] S3 SoC main control layered filtering and multimodal fusion verification Noise filtering: The CPU core executes a digital filtering algorithm to filter out piezoelectric noise generated by footsteps, racket hitting the table, and net swaying; Net contact joint judgment: Simultaneously reads the overlapping pixel area of ​​the ball and net in the images from both sides of the camera, and compares the vibration amplitude of the PVDF on the left and right net posts for double verification; Landing point validity judgment: AI tracks the ball's landing time and matches the timestamp and XY coordinates of the PVDF impact signal on the table; Serving compliance verification: AI identifies the ball toss height and serving offset angle.

[0027] S4 Comprehensive Judgment Output and Routine Interaction The SoC fusion judgment result is pushed to the touch terminal for display, and the judgment information is broadcast by voice.

[0028] S5 Player Appeal Human-Computer Interaction Review Process The player presses the appeal button, and the system locks and caches all data for the current round; the referee retrieves the dual-view synchronized replay and sensor evidence data, and chooses to uphold the original decision or modify the score.

[0029] S6 Post-Match Data Export Referees can export all round landing data, penalty records, and appeal replay videos with a single click via a touch terminal.

[0030] The beneficial effects of this invention are as follows: 1) Objective and accurate judgment, eliminating subjective bias. Bilateral vision eliminates blind spots, the net wiping system uses image + vibration dual verification, and the piezoelectric table accurately identifies millimeter-level landing points, resulting in a judgment accuracy rate far exceeding that of human eyes.

[0031] 2) Unmanned and automated officiating significantly reduces labor costs. The entire system can independently complete all the referee's work, eliminating the need for a certified full-time referee for each match.

[0032] 3) The equipment is lightweight, highly versatile, and has a low barrier to entry. It can be customized into a dedicated table tennis table (with a PVDF piezoelectric array sandwiched in the tabletop and an integrated visual net device) or it can be installed on an existing table tennis table without customization (with a PVDF piezoelectric array installed under the tabletop and a small 180° fisheye camera installed on the net posts), at a cost far lower than that of a professional eagle-eye system.

[0033] 4) High integration of processing units. The heterogeneous SoC integrates all functions on a single chip, eliminating the need for additional acquisition cards and synchronization boards, thus avoiding the drawbacks of complex and bulky peripherals in general-purpose CPUs.

[0034] 5) Local, low-latency, autonomous operation. It does not rely on external cloud networks, providing millisecond-level feedback on rulings; for controversial plays, it can retrieve objective supporting evidence with a single click, significantly reducing match pause times.

[0035] 6) Full-round data traceability. Automatically stores data for each round, and allows for one-click export of technical statistics after the match, assisting players in technical training and coaches in tactical analysis. Attached Figure Description

[0036] Figure 1 This is a schematic diagram of the overall assembly structure of the device of the present invention; Figure 2 This is a cross-sectional view of the integrated acquisition component with dual-sided grid pillars. Figure 3 A schematic diagram showing the PVDF landing point sensing module installed below the table surface; Figure 4 A schematic diagram showing the PVDF landing point sensing module installed in the interlayer of the table; Figure 5 This is a block diagram of the signal connection of the embedded heterogeneous SoC main control hardware. Figure 6 A schematic diagram of the appeal review interface for a human-computer interaction terminal; Figure 7 This is a flowchart illustrating the overall process of the AI-based automatic judgment method of the present invention.

[0037] List of identifiers in attached diagrams: 1. Table surface, 2. Net, 3. Net posts, 4. PVDF landing point sensor module on the table surface, 5. Shock-absorbing pad, 6. Heterogeneous SoC main controller, 7. Human-computer interaction components, 8. Edge rubbing sensor strip, 9. PVDF net rubbing piezoelectric film, 10. Camera, 11. Local synchronization clock submodule, 12. Main cable. Detailed Implementation

[0038] The present invention will be further illustrated below with reference to the accompanying drawings and specific embodiments. It should be understood that the following specific embodiments are for illustrative purposes only and are not intended to limit the scope of the invention.

[0039] Example 1: Complete Set of Standard Indoor Professional Training Table Tennis Tables An ultra-thin PVDF piezoelectric film matrix is ​​installed on the interlayer or back of the tabletop 1 and fully adhered to the bottom of a standard table tennis table. Edge-rubbing sensor strips 8 are installed around the tabletop. The sensor array is connected to a hidden side junction box via a main line 12 and wired to an embedded heterogeneous SoC main controller 6. The original ordinary left and right net posts are removed and replaced with the integrated acquisition net posts 3 of this invention. The data lines of the two posts are uniformly connected to the SoC main controller 6 to complete global clock synchronization. The embedded heterogeneous SoC main controller 6 is placed on a small storage bracket on the side of the table, which connects to the double net posts, the tabletop sensor junction box, a 7-inch touch interactive terminal, and a voice broadcast module. Two Bluetooth appeal buttons are paired and placed at both ends of the table. Upon power-up, the device automatically calibrates timing synchronization and loads the AI ​​model. During the game, it automatically and synchronously acquires images from both sides, net-rubbing vibrations, and ball landing coordinates. The AI ​​fusion completes automatic scoring and voice scoring, eliminating the need for a dedicated referee for the entire match.

[0040] Example 2: Low-Cost Simplified Example for Campus The PVDF film directly mounted on the underside of table 1 is simplified to a zoned discrete sensing unit, retaining the left and right ball landing areas and the four edge-rubbing sensing strips; the frame rate of the net post camera is reduced to 500fps; the SoC retains the CPU, basic AI inference, and multi-channel ADC functions, while the high-performance NPU is removed; the human-machine interface only retains basic functions such as score and replay. The school's intramural league does not require hiring professional referees; only one organizer is needed to complete the entire event.

[0041] Example 3: Example of a major table tennis event A PVDF piezoelectric array is installed in the interlayer of the tabletop, and the left and right ball landing areas and the four edge rubbing sensing strips 8 are retained. An integrated ball net device integrating PVDF pressure sensing and high-definition camera is assembled. The latest table tennis rules design an automatic judgment algorithm, and the program firmware is burned into the main control chip to complete real-time recognition and scoring of landing point, out of bounds, net rubbing, edge rubbing, and player serving fouls.

[0042] It should be noted that the above content merely illustrates the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. For those skilled in the art, various improvements and modifications can be made without departing from the principle of the present invention, and all such improvements and modifications fall within the scope of protection of the claims of the present invention.

Claims

1. A visual and PVDF piezoelectric fusion AI automatic judgment device for table tennis, characterized in that, include: Dual-sided grid column integrated acquisition component, tabletop PVDF landing point sensing module, embedded heterogeneous SoC main control unit, human-computer interaction terminal module, and wireless appeal peripheral module; The dual-sided net post integrated acquisition component includes hollow posts located on the left and right sides of the net. Each post is embedded with a high-speed camera and a PVDF piezoelectric film for net contact. The high-speed camera is used to acquire image streams of the ball's flight, net contact, and the player's serve. The PVDF piezoelectric film for net contact is set close to the net lines to sense the vibration signals generated by the ball's friction and impact with the net. Each post also has a local synchronization clock submodule. The tabletop PVDF impact point sensing module is a flexible PVDF piezoelectric film matrix. The flexible PVDF piezoelectric film matrix is ​​set in the interlayer under the tabletop or attached to the bottom of the table tennis table, and is used to collect the two-dimensional coordinates and impact timing signals of the table tennis ball impacting the tabletop. The embedded heterogeneous SoC main control unit is electrically connected to the dual-sided grid pillar integrated acquisition component, the tabletop PVDF landing point sensing module, and the human-computer interaction terminal module, respectively. The embedded heterogeneous SoC main control unit integrates a multi-core CPU computing core, an AI hardware acceleration core, multiple ADC acquisition channels, a dual-channel MIPI image decoding interface, and a global hardware timing synchronization module. The global hardware timing synchronization module sends a unified reference clock to both sides of the grid pillars to synchronize and align all sensor data, including camera images, grid pillar piezoelectricity, and platform landing point piezoelectricity. The embedded heterogeneous SoC main control unit has built-in multimodal fusion verification logic and table tennis competition rule judgment logic, and automatically outputs the judgment result by simultaneously fusing the visual images from both sides, the vibration signals of the left and right net posts rubbing the net, and the coordinate signal of the ball landing on the table. The human-computer interaction terminal module includes a touch screen and a voice broadcasting unit, used to display the score, replay dual-view video, piezoelectric sensing waveforms and broadcast penalty information; The wireless appeal peripheral module includes a pair of wireless communication devices, configured for athletes on both sides, to send round appeal trigger signals to the embedded heterogeneous SoC main controller.

2. The table tennis AI automatic judgment device according to claim 1, characterized in that, The PVDF wiping piezoelectric film is equipped with a micro charge amplifier circuit and a low-pass filter circuit, which output the wiping vibration amplitude and trigger timestamp.

3. The table tennis AI automatic judgment device according to claim 1, characterized in that, The PVDF piezoelectric film matrix on the platform is divided into a left half sensing area and a right half sensing area. Edge sensing strips are installed around the platform, and shock-absorbing pads are installed between the flexible PVDF piezoelectric film matrix and the platform.

4. The table tennis AI automatic judging device according to claim 1, characterized in that, The high-speed camera has a frame rate of 500-1000fps, outputs a MIPI high-speed image stream, and has its own hardware microsecond-level timestamp.

5. The table tennis AI automatic judgment device according to claim 1, characterized in that, The thickness of the PVDF piezoelectric film matrix on the platform is 0.15 mm, and each sensing array unit is independently matched with a high-speed ADC acquisition channel.

6. The table tennis AI automatic judgment device according to claim 1, characterized in that, When the human-computer interaction terminal module's touch screen is in the appeal state, a pop-up window simultaneously displays the dual-view playback images of the left and right net columns, as well as the PVDF mesh vibration waveform diagram of the net columns and the thermal distribution diagram of the PVDF landing point on the table.

7. A method for automatic AI-based judgment in table tennis, integrating vision and PVDF piezoelectricity, applied to the apparatus described in any one of claims 1 to 6, characterized in that... Includes the following steps: S1: The embedded heterogeneous SoC main controller outputs a global synchronization clock to complete the timing calibration of dual-side cameras, net post PVDF, and table PVDF sensors, loads the AI ​​detection model, resets the score to zero, pairs with the wireless appeal button, identifies the table boundary and net height, and records the ITTF judgment threshold. S2: During the game, the embedded heterogeneous SoC main controller synchronously acquires high-speed images from the dual-sided net pillar cameras, the net pillar PVDF acquires the net wiping vibration signal, and the table PVDF acquires the ball impact coordinates. All data with a unified timestamp is transmitted to the embedded heterogeneous SoC main controller. S3: The embedded heterogeneous SoC main control performs noise filtering and uses image pixel overlap matching and piezoelectric vibration amplitude of the net post to determine net grazing; it combines the PVDF impact timing and coordinates of the table surface to determine the landing point is valid, off the table, or grazing; and it uses AI to recognize the ball throwing action to verify the compliance of the serve. S4: The embedded heterogeneous SoC main control integrates multi-source sensor data to output the judgment result, pushes it to the touch terminal and announces it by voice, and automatically records the score; S5: When the athlete presses the wireless appeal button, the main control locks all video and piezoelectric sensor data of the current round, retrieves dual-view replay and piezoelectric waveform through the touch terminal to complete the review, updates the score and archives the round data; S6: After the match, export all round penalty data and replay videos via touch terminal.

8. The AI-based automatic penalty judgment method for table tennis according to claim 7, characterized in that, The specific logic for the joint determination of net contact in step S3 is as follows: when the images from both cameras detect that the ball is in contact with the net area, and the piezoelectric vibration amplitude of any net post exceeds a preset threshold, the ball is confirmed to have contacted the net. If only a single image overlaps or only there is a slight vibration, it is determined to be interference and the net contact instruction is not triggered.

9. The AI-based automatic penalty judgment method for table tennis according to claim 7, characterized in that, The specific logic for determining the validity of the landing point in step S3 is as follows: the AI ​​trajectory fits the moment the ball lands, and matches the timestamp and XY coordinates of the PVDF impact signal on the table. If the coordinates fall within the opponent's table, it is determined to be a valid shot and a point is scored. If the coordinates exceed the table boundary, it is determined to be off the table and a point is lost. If the impact occurs, the sensor strip around the table is widened to determine if it is a glancing edge. During the serve phase, it is verified whether the first landing point is located in the player's half of the court.

10. The AI-based automatic judgment method for table tennis according to claim 7, characterized in that, In step S5, when the athlete presses the wireless appeal button, the main control locks and caches all dual-channel video, net post piezoelectric waveform and table landing point sensor data for the current round. The interactive terminal then pops up an appeal review window. The athlete can select to maintain the original judgment or modify the score via touch screen. After confirmation, the SoC updates the score and broadcasts the appeal review result via voice. The complete sensor and video data for this round are permanently stored in local storage for archiving.

Images