Method for calculating a position of an athlete on a sports field

Abstract

A method for calculating the position of an athlete, referred to as target, on a sports field, including estimating an approximate position of the target using a radio-based positioning system, the system including tracking sensors attached to several athletes on the sports field and antennas installed around the sports field, defining a search space around the approximate position, detecting an athlete in the search space using an optical-based positioning system, the system including cameras installed above and/or around the sports field and image recognition device, determining an accurate position of the detected athlete, attributing the accurate position to the target.

Claims

1. A method for calculating a position of an athlete, referred to as a target, on a sports field, comprising: estimating an approximate position of the target using a radio-based positioning system, the radio-based positioning system comprising tracking sensors attached to several athletes on the sports field and antennas installed around the sports field, defining boundaries of a search space, as a first bounded region, around the estimated approximate position such that the estimated approximate position lies within the first bounded region, detecting, within the first bounded region, a second bounded region enclosing a body of the athlete by applying an image recognition algorithm only to the first bounded region corresponding to the defined search space, the first bounded region being defined within one or more images obtained using an optical-based positioning system, the optical-based positioning system comprising cameras installed at least one of above and around the sports field, the first bounded region being smaller than a view region of any particular camera in the optical-based positioning system, and being based only on the estimated approximate position, determining an updated position of the athlete within the detected second bounded region, and attributing the updated position to the target instead of the estimated approximate position.

2. The method according to claim 1, wherein the radio-based positioning system is based on time of arrival measurements.

3. The method according to claim 1, wherein the radio-based positioning system is based on time difference of arrival.

4. The method according to claim 1, wherein the radio-based positioning system is based on angle of arrival measurements.

5. The method according to claim 1, further comprising generating additional information on the target using at least one of an inertial measurement unit attached to the target and the optical-based positioning system.

6. The method according to claim 5, further comprising predicting a next position of the target or a future event using the additional information.

7. The method according to claim 5, further comprising validating the accurate position of the target using the additional information.

8. The method according to claim 5, wherein the additional information comprises an acceleration of the detected athlete.

9. The method according to claim 5, wherein the additional information comprises an orientation of the detected athlete.

10. The method according to claim 5, wherein the additional information comprises a pose of the detected athlete.

11. The method of claim 1, further comprising: determining, using the image recognition algorithm applied to the first bounded region, a position of an object used by the athlete to perform actions on the sports field.

12. A method for calculating a position of an athlete, referred to as target, on a sports field, comprising: estimating an approximate position of the target using a radio-based positioning system, the radio-based positioning system comprising tracking sensors attached to several athletes on the sports field and antennas installed around the sports field, defining boundaries of a search space around the estimated approximate position, detecting, within the first bounded region, a second bounded region enclosing a body of the athlete by applying an image recognition algorithm only to a search region corresponding to the defined search space, the search region being defined within one or more images obtained using an optical-based positioning system, the optical-based positioning system comprising cameras installed at least one of above and around the sports field, the search region being different from a view region of any particular camera in the optical-based positioning system, and being based only on the estimated approximate position, determining an updated position of the athlete within the detected second bounded region, attributing the updated position to the target instead of the estimated approximate position, determining a position of an object used by the athlete, and predicting a future use of the object by the athlete.

13. The method of claim 11, further comprising predicting a future use of the object by the athlete.

Description

BRIEF DESCRIPTION OF THE DRAWING

(1) Other features and advantages of the present invention will appear more clearly upon reading the following detailed description, made with reference to the annexed drawings, given by way of non-limiting example:

(2) FIG. 1 shows a sports field equipped with a radio-based positioning system

(3) FIG. 2 shows said sports field, equipped with an optical-based positioning system comprising cameras located above the field

(4) FIG. 3(a), FIG. 3(b) and FIG. 3(c) show said sport field, equipped an alternative optical-based positioning system comprising cameras located above and around the field

(5) FIG. 4(a) shows a step of estimating an approximate position of a target on said sports field, according to the invention

(6) FIG. 4(b) shows a step of defining a search space around said approximate position, according to the invention

(7) FIG. 4(c) shows a step of detecting and determining an accurate position of an athlete in said search space, according to the invention

(8) FIG. 5 shows a step of generating additional information regarding said detected athlete, according to the invention

(9) FIG. 6 shows steps of a method according to the invention.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

(10) The method according to the invention for calculating the position of a moving athlete—called target—on a sports field requires using a radio-based positioning system RdPS and an optical-based positioning system OpPS.

(11) As illustrated in FIG. 1, the radio-based positioning system RdPS comprises a set of distributed antennas At1, At2 . . . AtM located around the sport field SpF, capable of receiving radio frequency signals transmitted by tracking sensors Snr. Said tracking sensors Snr are attached to the athletes present on the sports field SpF, including the target Tgt. Possible options of attachment of the tracking sensors include the athletes' gear (for example protective gear in sports like ice hockey), the athletes' clothes (for example jerseys), and additional gear designed for this purpose (for example separate chest straps). The position determination of an athlete may be carried out by distance measurements or angle-of-arrival measurements between the tracking sensor attached to said athlete and the antennas. The particular measurements between an attached sensor and the distributed antennas are used by a calculation unit CcRd of the radio-based positioning system RdPS to calculate the position of the sensor.

(12) In an embodiment, the athletes carry inertial measurement units. This way, additional measurements such as acceleration, inclinations and angular rates can be recorded using the installed set of antennas At1, At2 . . . AtM.

(13) As illustrated in FIG. 2, the optical-based positioning system OpPS comprises a set of distributed, calibrated and synchronised cameras Cm1, Cm2 . . . CmN located above. In an alternative embodiment shown in FIG. 3(a), FIG. 3(b) and FIG. 3(c), the cameras Cm1, Cm2 . . . CmN are located both above and around the sports field. Images are gathered from different perspectives such that the camera setup completely captures the sports field. Beside the cameras located above and/or around the sports field, using additional peripheral cameras is possible. All cameras are managed by a camera controller CcOp of the optical-based positioning system OpPS. The detection of athletes in images may be implemented using traditional image recognition approaches. Another powerful detection approach is the application of machine learning algorithms that learn recognition rules and features from provided reference data. Recognition algorithms allow for a detailed semantic segmentation of the athletes, which is a source of additional information.

(14) The method MTH according to the invention comprises the following steps, illustrated in FIG. 6.

(15) In a step Est_ApPos, with reference to FIG. 4(a), an approximate position ApPos of the target Tgt is estimated using the radio-based positioning system RdPS. The approximate position may be determined using angle measurements between the tracking sensor attached to the target and installed antennas of the radio-based positioning system. The approximate position may alternatively be determined using distance measurements between the tracking sensor and the antennas (this is originally referred to as ranging-based positioning).

(16) In a step Def_ScSpa, with reference to FIG. 4(b), a search space ScSpa is defined around said approximate position. This way, the approximate position ApPos of the target Tgt from the radio-based positioning system is used to limit the search space for consequent calculations. This results in a reduced computational effort and a simplification of detection techniques.

(17) In a step Dtc_Ath, with reference to FIG. 4(c), an athlete Ath is detected in said search space ScSpa using the optical-based positioning system OpPS. As explained in the previous paragraph, image recognition algorithms are limited to the predefined region of interest (search space ScSpa) in order to reduce the calculations. The box shown on FIG. 4(c) shows the result of the detection of an athlete's body.

(18) In a step Dtm_AcPos, as illustrated in FIG. 4(c), an accurate and precise position AcPos of said detected athlete Ath is determined.

(19) In a step Atr_AcPos, the accurate position AcPos is attributed to the target Tgt.

(20) Modern machine learning based recognition algorithms are able to carry out object detection and localization in real time. The main shortcoming of these algorithms is that they are not able to unambiguously identify the respective athlete. In contrast, the position data gathered from radio-based positioning allows for an explicit allocation of the position data of an attached sensor to the target (signals received from multiple sensors can be discriminated by the radio-based positioning system RdPS and the assignment of a sensor to an athlete is defined by the user). Thanks to the method of the invention, accurate positions gathered from optical-based positioning can be assigned to their matching position gathered from radio-based positioning. Thus, an unambiguous allocation of the accurate position measurements to the correct athletes is possible.

(21) In addition, the method MTH may comprises the following steps.

(22) In a step Gen_AdInf, with reference to FIG. 5, additional information AdInf on the target Tgt is generated using at least one inertial measurement unit attached to said target and/or the optical-based positioning system OpPS. Such an additional information may be the acceleration of the target, their pose, their orientation, the orientation of their racket or stick, etc.

(23) In a step Val_AcPos, the additional information AdInf gathered from the optical-based positioning system OpPS and/or from the inertial measurement unit(s) is used to complement and validate the information from the radio-based positioning system. Information about additional objects, such as rackets or sticks, and a detailed analysis of the target may result in a higher position accuracy that are not available only with a radio-based positioning method.

(24) In a step Pr_NxPos, respectively Pr_FtEvt, the additional information AdInf gathered from the optical-based positioning system OpPS and/or from the inertial measurement unit(s) is used to predict a next position NxPos of the target Tgt, respectively a future event FtEvt (for instance, the target is going to use their racket or stick).

(25) Of course, the present invention is not limited to the illustrated example but may be subject to various variants and alterations, which will be apparent to those skilled in the art.