Method and computer for finding a launch position of a golf ball
The automated method for determining golf ball launch position through video frame analysis and blob ranking in golf ball tracking systems addresses the inefficiency of manual marker placement, achieving precise and cost-effective tracking.
Patent Information
- Application Number
- PCT/EP2026/062478
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2025-05-09
- Filing Date
- 2026-04-30
- Publication Date
- 2026-11-12
AI Technical Summary
Existing golf ball tracking systems require manual placement of virtual markers for determining the launch position, increasing operational costs and inefficiency.
An automated method using a computer to analyze video frames before and after the launch event, identify blobs, rank them based on shape, position, and distance, and extrapolate the launch position, eliminating the need for manual intervention.
Accurately determines the launch position of a golf ball with high precision, reducing operational costs by automating the process and eliminating the need for manual marker placement.
Smart Images

Figure EP2026062478_12112026_PF_FP_ABST
Abstract
Description
[0001] METHOD AND COMPUTER FOR
[0002] FINDING A LAUNCH POSITION OF A GOLF BALL
[0003] TECHNICAL FIELD
[0004] The embodiments herein relate to systems for tracking golf balls, such as tracking systems for production of video from a golf event, such as a competition, a driving range, or any other location where players hit golf balls. In particular, the embodiments disclosed herein relate to a method and a computer for finding a launch position of a golf ball, e.g. tracked by a tracking unit.
[0005] BACKGROUND
[0006] During a golf event, it is often desirable to broadcast video from, for example, a tee, at which a player may swing a golf club to shoot a golf ball towards e.g. a flag, a hole, out onto a range, a field or the like. Additionally, the broadcast video may include a visual trace indicating the flight path of the golf ball after it has been hit, thereby enhancing the viewing experience for the audience.
[0007] There are several known methods for generating such a trace of the golf ball's flight path. For example, a so-called capturing system can be used. The capturing system can include a video camera and a tracking unit. The video camera captures a scene in which a trace, captured by tracking unit can be rendered. The tracking unit captures, e.g. using radar and / or image sensors, the trace of the golf ball. The trace obtained by the tracking unit typically starts at a point in time, or space, from when the golf ball was actually launched, because the tracking unit may need to evaluate many different false positive traces and can only determine with sufficient confidence after some time that a particular trace corresponds to a launched golf ball.
[0008] As a result, in order to produce a complete trace of the golf ball's flight— from its launch position to its current or last detected position— one or more human operators are typically required to manually place a virtual marker in the video stream captured by the video camera, which includes the launch position within its field of view. This marker represents the estimated initial position of the golf ball prior to launch. The need for manual placementof such a virtual marker increases the operational costs associated with the capturing system.
[0009] SUMMARY
[0010] The present inventors have realized an automated method of finding the launch position would save both time and cost when operating a system of the abovementioned kind.
[0011] An object may hence be to eliminate, or at least reduce, the abovementioned problem and / or disadvantages.
[0012] This object, and possibly other objects, is achieved by the aspects of the present invention as set forth herein.
[0013] According to an aspect, there is provided a method, performed by a computer, for finding a launch position of a golf ball tracked by a tracking unit. The computer obtains, from a video capturing device, a set of video frames of, e.g. illustrating, capturing, showing, depicting, or the like, a scene in which the golf ball is visible. The computer obtains a message indicating a launch event relating to the golf ball. The launch event is associated with a launch point in time. The computer determines a first point in time that occurs before the launch point in time and a second point in time that occurs after the launch point in time.
[0014] Furthermore, the computer identifies, within the set of video frames, a first frame at the first point in time and a second frame at the second point in time. The computer determines a differential frame by computing a difference between the first and second frames. The computer forms a set of blobs of pixels in the differential frame.
[0015] The computer also ranks the set of blobs by ranking each blob of the set of blobs based on one or more of:
[0016] • a shape of said each blob,
[0017] • a position of said each blob, e.g. in relation to standard rules of golf, in relation to tee markers, or the like,
[0018] • a number of pixels of said each blob,• a blob location difference given by a comparison of a blob distance between said each blob and the video capturing device to a launch area distance between a launch area and the video capturing device, or
[0019] • the like, or
[0020] • a combination thereof,
[0021] thereby forming a ranked set of blobs. The computer identifies an extremum, such as a minimum, a maximum, or the like, in the ranked set of blobs as a candidate blob. The computer generates an extrapolated trace by extrapolating a trace obtained by the tracking unit to a position of the candidate blob. The computer verifies that the extrapolated trace fulfils conditions of expected motion of the golf ball, assumed to be freely moving after the launch point in time and as constrained by the trace. Next, the computer provides the position of the candidate blob as the launch position of the golf ball.
[0022] Thanks to the ranking of the blobs, the computer can find out which blob has a position that most likely corresponds to the launch position. Accordingly, in this manner, the computer can determine the launch position of a golf ball within a frame, such as a camera image, e.g. from the video capturing device, such as a TV / Broadcast camera, with high accuracy and while taking one or more of the shape of said each blob, the position of said each blob, the number of pixels of said each blob, the blob location difference, or the combination thereof, into account. The launch position can be a two dimensional position in the frame, i.e. having an X and Y coordinate, e.g. expressed in pixel values within the frame.
[0023] An advantage is that the manual step in the process of tracking objects according to known solutions can be avoided, or even eliminated. As a consequence, cost of operating the system can be reduced, since one or more operators, for whom their only task is to place a "ball marker", can be reduced, or even eliminated. The embodiments herein provide an automated ball tracking method that after setup is free from manual steps of placing a ball marker and can give a complete trace to visualize the flight of an object, such as a golf ball.
[0024] According to another aspect, the object is achieved by a computer configured for finding a launch position of a golf ball tracked by a tracking unit. The computer is configured for obtaining, from a video capturing device, a set of video frames of a scene in which the golfball is visible. The computer is configured for obtaining a message indicating a launch event relating to the golf ball. The launch event is associated with a launch point in time. The computer is configured for determining a first point in time that occurs before the launch point in time and a second point in time that occurs after the launch point in time. The computer is configured for identifying, within the set of video frames, a first frame at the first point in time and a second frame at the second point in time. The computer is also configured for determining a differential frame by computing a difference between the first and second frames. The computer is configured for forming a set of blobs of pixels in the differential frame. The computer is configured for ranking the set of blobs by ranking each blob of the set of blobs based on one or more of:
[0025] • a shape of said each blob,
[0026] • a position of said each blob, e.g. in relation to standard rules of golf, in relation to tee markers, or the like,
[0027] • a number of pixels of said each blob,
[0028] • a blob location difference given by a comparison of a blob distance between said each blob and the video capturing device to a launch area distance between a launch area and the video capturing device, or
[0029] • a combination thereof,
[0030] thereby forming a ranked set of blobs. The computer is configured for identifying an extremum in the ranked set of blobs as a candidate blob. The computer is configured for generating an extrapolated trace by extrapolating a trace obtained by the tracking unit to a position of the candidate blob. The computer is configured for verifying that the extrapolated trace fulfils conditions of expected motion of the golf ball, assumed to be freely moving after the launch point in time and as constrained by the trace. The computer is configured for providing the position of the candidate blob as the launch position of the golf ball.
[0031] In some embodiments, the ranking of said each blob comprises calculating a respective score based on one or more of:
[0032] the shape of said each blob,
[0033] the position of said each blob,• the number of pixels of said each blob,
[0034] • the blob location difference,
[0035] • the like, or
[0036] • the combination thereof.
[0037] In some embodiments, the calculating of the respective score comprises calculating the respective score as a weighted average of one or more of, preferably all of, the shape of said each blob, the position of said each blob, the number of pixels of said each blob, and the blob location difference.
[0038] In some embodiments, the forming of the set of blobs comprises the ranking of the set of blobs and / or the ranking of the set of blobs is performed after the forming of the set of blobs.
[0039] In some embodiments, the generating of the extrapolated trace comprises backward extrapolating the trace to the position of the candidate blob.
[0040] In some embodiments, the expected motion of the golf ball conforms to physics of projectile motion, wherein preferably the expected motion comprises an expected trace as constrained by classical physics.
[0041] In some embodiments, the message is obtained from one or more of the tracking unit, the video capturing device, and a launch event detecting device.
[0042] In some embodiments, the obtaining of the message comprises generating the message using one or more of:
[0043] • radar and / or optical detection,
[0044] • image processing on the set of video frames to identify characteristics of launcher when launching the golf ball,
[0045] • sound detection,
[0046] • the like, ora combination thereof.
[0047] In some embodiments, the determining of the first point in time and the second point in time comprises determining the first point in time to be a first number of milliseconds, "ms", before the launch point in time and determining the second point in time to be a second number of ms after the launch point in time.
[0048] In some embodiments, the determining of the differential frame comprises computing a pixel-wise or pixel-group-wise difference between the first and second frames, or wherein the determining of the differential frame comprises using one or more of pixel-wise differencing, block-wise differencing, background subtraction, thresholding, optical flow estimation, or temporal median filtering, or a combination thereof, to determine the differential frame.
[0049] In some embodiments, the providing of the position comprises providing the extrapolated trace including the position.
[0050] In some embodiments, the providing of the extrapolated trace comprises:
[0051] drawing the extrapolated trace and the trace in video frames received from the video capturing device, and / or
[0052] providing access to the extrapolated trace and / or the trace via an application programming interface.
[0053] In some embodiments, the method is performed by a system comprising the computer and at least one of the tracking unit and the video capturing device.
[0054] BRIEF DESCRIPTION OF THE DRAWINGS
[0055] The various aspects of embodiments disclosed herein, including particular features and advantages thereof, will be readily understood from the following detailed description and the accompanying drawings, which are briefly described in the following.Figure 1 is a schematic overview of an example of a system in which embodiments herein may be implemented.
[0056] Figure 2 is a flowchart illustrating embodiments of the method herein, e.g. performed by the computer.
[0057] Figure 3 is a schematic timeline illustrating frames before and after a point in time of a launch event.
[0058] Figure 4, Figure 5 and Figure 6 are 2D image frames relating to the first and second point in time and to an example of a related differential frame.
[0059] Figure 7 is a block diagram illustrating embodiments of the computer.
[0060] DETAILED DESCRIPTION
[0061] Throughout the following description, similar reference numerals have been used to denote similar features, such as nodes, actions, modules, circuits, parts, items, elements, units or the like, when applicable.
[0062] Figure 1 shows an exemplifying system 100 according to the embodiments herein. The system is configured to track a golf ball 101 in flight and to provide a video stream including a trace 105 of the golf ball's trajectory in the air. Two tee markers 108, 109 indicate a launch area, from which the golf ball 101 can be hit by a golf player 103. The two tee markers 108, 109 include a left tee marker 108 and a right tee marker 109. The launch area can thus be a physical area on the ground, i.e. as above indicated by the tee markers 108, 109 and constrained by standard rules of golf.
[0063] Figure 1 also shows a launch position Pl, e.g. an actual launch position, a true launch position, or the like. Additionally, Figure 1 shows another launch position P2 for evaluation, e.g. a candidate launch position, a false positive launch position, or the like.
[0064] The system 100 includes a computer 110, a tracking unit 120, a video capturing device 130 and optionally a launch event detecting device 140.
[0065] The tracking unit 120 can be configured to capture at least tracking data, from which a trace of the golf ball can be derived. The tracking unit 120 may further be configured to capture a video, whose video frames may be provided, e.g. rendered in or on, with at least a portion ofthe trace. Accordingly, the tracking unit 120 is configured to provide tracking data, such as tracking image frames, tracking frames, or the like, to e.g. the computer 110. In more detail, the tracking unit 120 may comprise one or more of a radar, a lidar, a video capturing device, a camera, a video recorder, a high-speed camera, a high-speed video recording device, a combination thereof, or the like. In this context, it may be noted that high-speed refers to an image capturing rate, often measured in frames per second (fps), of 500 fps, 1000 fps, 10000 fps, 100000 fps or more.
[0066] Furthermore, the video capturing device 130, such as a broadcast camera, a video camera, a camcorder, or the like, can be configured to provide a video stream processable by the computer 110, or any other device configured to provide the video stream with the trace of the golf ball 101. The launch area can possibly be seen as a portion of a frame in a set of frames, e.g. video frames, from the video capturing device 130. The portion can be given by the 2D positions of the markers in the frame. E.g. below a line LI, see Figure 6, and limited in width by vertical lines at, e.g. with some margin, the left and right tee marker 108, 109.
[0067] In Figure 1, the launch event detecting device 140 can comprise the tracking unit 120 and / or the video capturing device 130. In particular, the launch event detecting device 140 can include a microphone for picking up sounds from the launch area. Furthermore, in some examples, the computer 110 can comprise the launch event detecting device 140.
[0068] As used herein, a blob comprises a set of pixels, i.e. one or more pixels, that have been identified as being a moving collection of pixels from one frame to another subsequent frame, e.g. by use of a blob detection procedure. In related literature, e.g. Computer Vision (CV), many known blob detection procedures are described. For example, differential blob detection, spatial blob detection, grey-level blob detection or the like.
[0069] As used herein, a trace comprises a series of blobs detected in subsequent frames, where the series of blobs is determined to pertain to the same moving object, such as sports projectile or the like. A known or proprietary tracing procedure may be used for this purpose. Again, in related literature many known tracing procedures are described. Blobdetection and / or tracking procedure are known from e.g. US11335013, US 10338209 and other documents.
[0070] It may here be noted that the tracking unit 120 can perform blob detection as part of its procedure to form a trace of the golf ball 101 after being launched. The tracking unit 120 can employ any known blob detection procedure that is appropriate for supporting tracking of moving objects may be applied to tracking image frames captured by the tracking unit 120.
[0071] As will be described below, blobs are identified by the computer 110 in frames from the video capturing device 130. The purpose of this blob detection is to find the launch position, or potential launch positions. Here, as described in the following the blob detection, e.g. the computer 110, ranks blobs in a way specifically adapted to accurately find the actual launch position Pl.
[0072] As used herein, the terms "golf ball", "tracked object", "sports projectile" or the like, may have been used interchangeably. A tracked object may thus be a sports projectile, a golf ball, a ball, a tennis ball, a boule ball, a football, a soccer ball, a badminton ball, a curling stone or the like.
[0073] Figure 2 shows an example of the method for finding the launch position Pl of the golf ball 101. As an example, the computer 110 performs a method for finding the launch position Pl of the golf ball 101 tracked by the tracking unit 120. In some examples, the method is performed by the system 100, which again can comprise the computer 110 and at least one of the tracking unit 120 and the video capturing device 130.
[0074] Initially, the computer 110 may establish wired or wireless connections to the tracking unit 120 and the video capturing device 130.
[0075] One or more of the following actions may be performed in any suitable order.
[0076] Action A110
[0077] The computer 110 obtains, such as receives, fetches, retrieves, or the like, from the video capturing device 130 or from a frame queue or the like, a set of video frames of a scene inwhich the golf ball 101 is visible, i.e. the golf ball 101 is visible in the set of video frames at least before the launch event in action A120. The set of video frames have been, or are being, captured by the video capturing device 130.
[0078] Action A120
[0079] The computer 110 obtains, such as receiving, fetches, retrieves, or the like, a message indicating a launch event relating to the golf ball 101. The launch event is associated with a launch point in time TO.
[0080] The message can be obtained from one or more of the tracking unit 120, the video capturing device 130, and the launch event detecting device 140.
[0081] The message can be generated, e.g. by the tracking unit 120 and / or the video capturing device 130 and / or the launch event detecting device 140 and / or the computer 110, using one or more of:
[0082] • radar and / or optical detection,
[0083] • image processing on the set of video frames to identify characteristics of launcher, i.e. the golf player 103, when launching the golf ball 101, e.g. when the video capturing device 130 generates the message,
[0084] • sound detection, e.g. when the tracking unit 120 and / or the video capturing device 130 and / or the launch event detecting device 140 generates the message, or
[0085] • a combination thereof.
[0086] Any device, such as the tracking unit 120, the video capturing device 130, the launch event detecting device 140, and the like, that has a microphone can be used when generating the message using sound detection. As an example, the microphone can pickup sounds from the launch area and identify the sound of when the golf ball 101 is hit by the golf player 103. Then the launch point in time will be the point in time of that sound.
[0087] See also section "Generating the message indicating the launch event" below.
[0088] Action A130
[0089] The computer 110 determines, such as establishes, calculates, or the like, a first point in time T1 that occurs before the launch point in time TO and a second point in time T2 that occurs after the launch point in time TO.The computer 110 determine the first point in time and the second point in time by determining the first point in time to be a first number of milliseconds, "ms", before the launch point in time and by determining the second point in time to be a second number of ms after the launch point in time. The first number of ms can be 10 ms, 20 ms, 30 ms, 40 ms, 50 ms, 100 ms, 200 ms, 500 ms, 1000 ms, or the like. The second number of ms can be 10 ms, 20 ms, 30 ms, 40 ms, 50 ms, 100 ms, 200 ms, 500 ms, 1000 ms, or the like. The frame rate can be 30, 50, 60 fps, or the like depending on the video capturing device 130 in use. For any frame rates, X fps, minimum values of the first and second number can be calculated as 1 / (2*X). The first and second number can often be 1 / X ms, or more in some cases. For a frame rate of 30 frames per second, a duration of one frame corresponds to 33 ms.
[0090] Therefore, for a frame rate of 30 fps, both the first and the second number can often be 33 ms divided by two, or as mentioned above 33 ms.
[0091] Action A140
[0092] The computer 110 identifies, within the set of video frames, a first frame Fl at the first point in time T1 and a second frame F2 at the second point in time T2. Typically, the first point in time T1 and the second point in time T2 are not completely, or not even close to being, aligned with points in time associated with the video frames. Therefore, the first frame Fl can be the frame whose associated point in time is closest to the first point in time T1 or whose associated point in time is closest before the first point in time Tl. Correspondingly, the second frame F2 can be the frame whose associated point in time is closest to the second point in time T2 or whose associated point in time is closest after the second point in time T2.
[0093] The golf ball 101 is visible in the first frame Fl. Clearly, since the second frame F2 is at the second point in time T2 after the launch event, the golf ball 101 is not visible in the second frame F2, or at least not visible the launch position Pl. That is, the golf ball 101 is non-visible in the second frame F2, i.e. the golf ball 101 is non-visible at the launch position Pl.
[0094] Action A150
[0095] The computer 110 determines a differential frame DF by computing a difference between the first and second frames Fl, F2. In more detail, since the golf ball 101 is visible in the first frame Fl, but not in the second frame F2 (the golf ball 101 is at least not visible at the launchposition Pl in the second frame F2), a blob in the differential frame DF can indicate the launch position. However, the differential frame DF can also include one or more false positive blobs that are not the actual launch position.
[0096] As an example, the computer 110 can determine the differential frame by computing a pixelwise or pixel-group-wise difference between the first and second frames Fl, F2.
[0097] Alternatively or additionally, the computer 110 can determine the differential frame by using one or more of pixel-wise differencing, block-wise differencing, background subtraction, thresholding, optical flow estimation, or temporal median filtering, or a combination thereof. That is, the computer 110 can use any known algorithm for determining the differential frame DF based on the first and second frames Fl, F2.
[0098] The determining of the differential frame DF can be performed per channel of the frames, i.e. per color channel of the frames. The computer 110 can therefore typically determine the differential frame using the blue channel of a RGB color coded frame. This can be beneficial when the golf ball 101 is white. For a yellow golf ball, it can be preferred to use one or more other channels, such as the green channel, or the like. In other examples, it can be beneficial to transform the frames into another color space, such as Red Green Blue (RGB), Cyan Magenta Yellow Black (CMYK), Hue Saturation Value (HSV), Hue Saturation Lightness (HSL), Luma Blue-difference Red-difference (YCbCr), Lightness a* b* (Lab), X Y Z (CIE XYZ), Standard Red Green Blue (sRGB), etc..
[0099] In a more detailed example, a differential frame can be calculated by subtracting one image or frame from another to find their difference. This can be done by subtracting the value of each pixel in one of the images from the corresponding pixel in the other image. This can be done by either looking at the pixel value as a combination of all the color channels present in the image (RGB and / or YUV values), or the individual color channels, or using a representation of the image converted to grayscale. The result of the subtraction is then compared to a set threshold value to eliminate residual background noise. If the difference is above this value it is considered as a difference (movement) in the image. As a consequence, an order of the terms in the subtraction can matter. This can be mitigated by either checking if the value is below the threshold instead of above or one can use the absolute value of the difference. The differences can be represented in a binary format where the movement is stored as ones, e.g. black, and no differences ("background") as zeroes, e.g. white, resultingin a black and white image. Alternatively, the movement can be stored as zeros and no difference as ones.
[0100] Action A160
[0101] The computer 110 forms a set of blobs of pixels, such as clusters / groups of pixels, in the differential frame DF, e.g. using any known blob detection algorithm. The blob detection algorithm can use connected-component labelling (CCL) with a 24-way connectivity, e.g. looking both at nearest and one step away, or the like. In other examples, other blob or clustering algorithms can be used.
[0102] In some examples, the computer 110 can obtain a classification of objects identified in a neighborhood of the first and / or second frame Fl, F2, e.g. using a segmentation or classification procedure. Here, the neighborhood of the first and / or second frame Fl, F2 can refer to any frame that is located between the first and second frames Fl, F2, any frame that is located directly, or almost directly, before or after the first and second frames Fl, F2, or it can refer to any one or both of the first and second frames Fl, F2. This means that the computer 110 can execute a classification procedure in order to obtain the classification of the objects. Alternatively, or even additionally, the classification procedure can be executed by another entity (not shown) that provides the classification of objects to the computer 110.
[0103] The classification of objects includes a type and a position of the object. This means that the classification procedure can result in e.g. a list of object types and corresponding positions (and optionally further information, such as width, length or object, etc.), i.e. results from the classification procedure. The object types can be "a player", "an active player", "a spectator", "a golf ball", "a tee marker", "a golf bag", "a caddie", etc. As above, each object type can be associated with a corresponding position, e.g. in two dimensions.
[0104] From the results, the computer 110 can extract one or more of:
[0105] • distance (and / or position) relative to the most likely player, wherein the most likely player can be an object type in the results from the classification procedure. For example, the most likely player can be the player that has a sufficient approximatesize, or height, and is closest to a center of a launch area, defined the tee markers 108, 109.
[0106] • a position, e.g. in pixels, a cluster, or blob, of the player can be limited by a set of rules. For instance, the rules can include that a cluster with a Y (height) coordinate above or below the bounds of a rectangle including the player is unlikely to be the launch position and thus resulting in a lower score, e.g. if not already removed from consideration by the filters. Similarly, an X coordinate value too close to a player or too far away from a player can also be used to give lower scores to a particular cluster. Thus, an interval on X values and / or an ROI on X and Y can be defined to conversely result in higher scores for clusters within these bounds. The size of the interval and / or size and geometry of the ROI can be dynamically assessed with known or estimated sensor size and focal length of the video capturing device 130 and / or a measured distance to launch area.
[0107] • Objects classified as "a ball" or "a golf ball", i.e. when the classification procedure has identified an object as a golf ball, this classification can be used to give the clusters and / or blobs with a position corresponding to the classified golf ball object, e.g. within a threshold radius of the position of the classified golf ball object, a higher score, or
[0108] • the like.
[0109] Action A170
[0110] The computer 110 ranks the set of blobs as elaborated on in subsequent action A172.
[0111] Action A172
[0112] The computer 110 ranks each blob of the set of blobs based on one or more of:
[0113] • a shape of said each blob,
[0114] • a position of said each blob, e.g. in relation to standard rules of golf, in relation to tee markers, in relation to the player, e.g. the most likely player, or the like,
[0115] • a number of pixels of said each blob,• a blob location difference given by a comparison of a blob distance between said each blob and the video capturing device 130 to a launch area distance between a launch area and the video capturing device 130, or
[0116] • a classification, or classification object, from the classification procedure,
[0117] • the like, or
[0118] • a combination thereof,
[0119] thereby forming a ranked set of blobs.
[0120] The computer 110 can perform the ranking A172 of said each blob by calculating A174 a respective score based on one or more of:
[0121] • the shape of said each blob,
[0122] • the position of said each blob,
[0123] • the number of pixels of said each blob,
[0124] • the blob location difference,
[0125] • the like, or
[0126] • the combination thereof.
[0127] In some examples, the computer 110 can calculate A174 the respective score by calculating the respective score as a weighted average of one or more of, preferably all of, the shape of said each blob, the position of said each blob, the number of pixels of said each blob, and the blob location difference, or the like. In this manner, the influence of each of the shape of said each blob, the position of said each blob, the number of pixels of said each blob, the blob location difference, and / or the like, on the respective score can be chosen. As an example, it may be preferred that e.g. the shape of said each blob has a greater contribution than e.g. the blob distance to the respective score, then a weight for the shape of said each blob should be greater than a weight for the blob distance.
[0128] In the following examples, high score is assumed to mean that said each blob matches blobs that potentially originate from golf balls and low score is assumed to mean that the blob is probably not a golf ball. In other examples, high score can mean that the blob is not a golf ball and a low score can mean that the blob originates from a golf ball.As an example relating to the shape of said each blob, said each blob can be ranked as a more promising blob the closer the shape of said each blob is to a circle, an ellipse, or the like. A shape of the golf ball 101 is circular when stationary. Sometimes, the shape can become stretched due to various reasons, such as due to movement. As an example, the computer 110 calculates a matching score that is higher the more closely the shape of said each blob corresponds to a shape of a circle, an ellipse, or the like. Various known algorithms for circle and / or ellipse similarity can be used, such as circularity factor (4*pi*area / perimeterA2), Hough circle transform, least square circle fit, least square ellipse fit, and the like.
[0129] As an example relating to the position of said each blob, the computer 110 can discard said each blob, when the position is above a line LI in the differential frame. Discard can mean that a low score, or even zero, is awarded to said each blob. The position of said each blob can typically refer to the position of the blob within the frame, e.g. in terms of pixels, scaled or non-scaled x and y position, e.g. in a length unit, such as meters, or the like. The line LI can be defined as coinciding with a virtual line running through both the two markers 108, 109, or slightly above the virtual line, e.g. in order to add some margin. The margin can be useful to ensure that the golf ball is not accidentally discarded, e.g. in case the golf ball is placed at said virtual line, e.g. on a tee. The margin can also be useful in case a player accidentally places the golf ball closer to the target, e.g. a hole, a flag, or the like, than the virtual line, i.e. against the standard rules of golf. Accordingly, when said each blob is below the line LI, the computer 110 can rank said each blob high, e.g. by assigning a high score to said each blob.
[0130] As an example relating to the number of pixels of said each blob, it can be that it has been found that blobs of golf balls typically include X pixels, such as about 10, 20, 30, etc. pixels. The value of X can be find empirically, by testing, by tuning, etc. and can assume any suitable value depending on the particular use case. The expected size of the ball, given as the number of pixels of said each blob, is dependent on the distance from the video capturing device to the golf ball. The computer 110 can therefore assign said each blob a high score when the number of pixels is close to X and a low score when the number of pixels is further away from X. As a further example, given that a size of the golf ball 101 is known, e.g. minimum of 42.67mm in diameter, and a maximum deviation should be less than tenths of amillimeter, and the video capturing device's 130 sensor size and resolution is known, the number of pixels of blobs originating from the golf ball 101 can be estimated.
[0131] As an example relating to the blob location difference, it can be that the computer 110 can calculate, or estimate, etc., the difference between the blob distance and the launch area distance to obtain the blob location difference. In more detail, the blob location difference can be obtained, e.g. by the computer 110, relatively the video capturing device 130, as the blob distance compared to the launch area distance.
[0132] In some examples, the smaller the blob location difference is the higher the score of said each blob will be given by the computer 110. The launch area distance can be the distance between the tracking unit 120 and / or the video capturing device 130 and a point at the ground within the launch area, i.e. the physical launch area.
[0133] In some examples, the computer 110 forms the set of blobs as part of the ranking A170 of the set of blobs and / or the computer 110 can perform the ranking A170 of the set of blobs after the forming A160 of the set of blobs.
[0134] As can be understood from the foregoing, the distinction between blob forming and ranking can be subjective. For example, the shape of said each blob can be a condition when forming the blobs, but the shape can also, or alternatively, be a condition when ranking said each blob. This same, or similar, applies to the number of pixels of said each blob. However, the position of said each blob and / or the blob distance can preferably be applied in the ranking of said each blob.
[0135] However, when using weights for each of the shape of said each blob, the position of said each blob, the number of pixels of said each blob, and the blob location difference, it may be preferred to apply the weighting in either the blob forming context or the ranking context, in order to control relative influence of each parameter. Here, it is understood that each parameter refers to each parameter of a set of parameters, wherein the set of parameters comprises the shape of said each blob, the position of said each blob, the number of pixels of said each blob, the blob location difference, or the like, or the combination thereof.
[0136] Action A180
[0137] The computer 110 identifies an extremum in the ranked set of blobs as a candidate blob. The extremum can be a maximum, a minimum, or the like.As an example, the computer 110 can identify the candidate blob as the blob that has a highest, or a lowest, respective score among the ranked blobs. Depending on whether high or low score is chosen as criteria for indicating that the blob is likely a golf ball, high or low weighted average score can indicate a most probable blob.
[0138] Sometimes, it may be that the highest scored candidate is only evaluated in action A190 below, when a difference in score to a second highest score is above a threshold value. In this manner, it can be ensured that the candidate blob stands out, e.g. in terms of the score, as compared to scores of other blobs.
[0139] Action A190
[0140] The computer 110 generates an extrapolated trace 106 by extrapolating a trace 105 obtained by the tracking unit 120 to a position Pl of the candidate blob.
[0141] The computer 110 can further generate the extrapolated trace 106 by backward extrapolation of the trace 105 to the position of the candidate blob.
[0142] Action A200
[0143] The computer 110 verifies that the extrapolated trace 106 fulfils conditions of expected motion of the golf ball 101, assumed to be freely moving after the launch point in time TO and as constrained by the trace 105.
[0144] As an example, if the combination of the trace 105 and the extrapolated trace 106 includes a jerk, or jaggedness, the computer 110 can conclude that the extrapolated trace 106 does not fulfil conditions of expected motion of the golf ball 101, e.g. when moving freely. However, if the combination of the trace 105 and the extrapolated trace 106 bends smoothly as expected due to gravity, the computer 110 can conclude that the extrapolated trace 106 fulfils the conditions of expected motion. In addition to gravity, launch parameters, such as speed, angle, etc., as well as already detected position of the ball in flight influences the conditions of the expected motion, or trace.
[0145] As another example, the computer 110 can check that the position Pl of the candidate blob qualifies as the launch position by extrapolation from the trace of the tracked golf ball. The candidate blob can be said to qualify as the launch position if the extrapolated trace is smooth and has a curvature as if the golf ball 101 is influenced by gravity, launch parameters, and / or already detected positions.In some examples, the expected motion of the golf ball conforms to physics of projectile motion. The expected motion can further comprise an expected trace as constrained by classical physics, such as physics of projectile motion, physics of moving mechanical objects, or the like.
[0146] As a further example, the computer 110 can verify that the extrapolated trace 106 matches, such as conforms to, corresponds to, aligns with, or the like, at least to some extent the trace 105, e.g. when considering laws of classical physics. To some extent can refer to that a certain margin can be allowed, i.e. exact match or conformity is not required.
[0147] Action A210
[0148] The computer 110 provides the position Pl of the candidate blob as the launch position of the golf ball 101. In this manner, the computer 110 can communicate a result of the method herein, e.g. to be displayed, to be processed by another device, to be communicated to a user, or the like.
[0149] Action A215
[0150] The computer 110 can provide the position Pl by providing the extrapolated trace 106 including the position Pl.
[0151] In some examples, the providing A215 of the extrapolated trace 106 can comprise action A217 and / or action A219 as below.
[0152] Action A217
[0153] The computer 110 may draw the extrapolated trace 106 and the trace 105 in video frames received from the video capturing device 130.
[0154] Action A219
[0155] The computer 110 may provide access to the extrapolated trace 106 and / or the trace 105 via an application programming interface, or API for short.
[0156] Figure 3 shows an example of a set of video frames captured by the video capturing device 130. The launch point in time TO occurs at the launch event, detected according to any one or more of the examples herein. Prior to the launch point in time TO, the first point in timeT1 occurs. After the launch point in time TO, the second point in time T2 occurs. Sometimes, the launch point in time TO can be adjusted to be synchronized and in phase with the set of video frames. Then, the first and second points in time Tl, T2 can be associated with a first and second frame Fl, F2, respectively, where each frame Fl, F2 is at an integer multiple of a frame duration away, i.e. before or after, the launch point in time TO. By means of operator D, e.g. for calculating differential frames, a differential frame DF is determined as the difference between the first and second frame Fl, F2.
[0157] As mentioned, the computer 110 determines the first point in time Tl that occurs before the launch point in time TO and the second point in time T2 that occurs after the launch point in time TO.
[0158] Figure 4 illustrates an example of the first frame Fl. In the first frame Fl, the golf ball 101 is visible. The golf ball 101 is resting on a launch area, e.g. on a tee.
[0159] Figure 5 illustrates an example of the second frame F2. In the second frame F2, the golf ball 101 is visible, e.g. in the air. The golf ball 101 has been hit and is flying away over a fairway, a range, or the like.
[0160] Figure 6 illustrates an example of the differential frame DF. The computer 110 has formed several blobs, performed ranking of the blobs and identified a candidate blob whose position reasonably fits with the trace 105. The differential frame DF is in this examples a binary frame difference obtained from the first and second frames Fl, F2. Here, white pixels in Figure 6 are considered to not be any movement and black pixels are movements, detected above a filtering threshold, in the frame, or image. The line LI shows a limiting border of the tee markers, which can be used as a filter, i.e. filtering out blobs located above the line LI. The crosses 108c, 109c shows the tee markers. Blobs are indicated as circles. The estimated launch position 603 from the tracking process is shown as +. In some examples, the computer 110 can filter based on said each blob's distance to the estimated launch position 603, e.g. assign higher scores for said each blob that are closer to the estimated launch position 603.A dot with X indicates the best cluster 602 based on the ranking, e.g. the scores from the ranking. A bounding box 601 of the player can be obtained from the classificationprocedure. For example, a low score can be awarded to blobs that are located inside the bounding box 601, i.e. too close to the player.
[0161] Generating the message indicating the launch event
[0162] Various actions can be performed to initiate, or execute, the generation of a message indicating the occurrence of a launch event. The actions can serve as the initial step in a method for detecting and processing a launch, and is preferably performed before, or as part of, action A120 above. These actions can be called triggers.
[0163] It is noted that, for tracking-based triggers, the specific implementation details for detecting a launch event are not provided, as it is assumed that a person skilled in the art would understand how the respective technologies may be used to detect the motion of a launched object such as a golf ball. Furthermore, although the minimum requirement for a trigger is to distinguish between a launch event and non-launch, many of the triggering systems are capable of generating additional useful data, which can be leveraged by subsequent processing modules in the system, or the computer.
[0164] In one embodiment, a launch event is detected using a tracking system comprising at least one radar sensor, optionally of the Frequency Modulated Continuous Wave (FMCW) type. Upon detection of a moving object indicative of a launch, the radar system generates a trigger signal. This trigger may include, in addition to the launch indication, data such as the estimated radial distance to the object, radial velocity, and the estimated two-dimensional or three-dimensional launch position. This information is propagated to other system components, along with the estimated launch time.
[0165] In another embodiment, a launch event is detected based on a combination of radar and optical tracking systems. The fusion of radar and optical observations may be used to increase reliability and reduce false positives. For instance, a trigger generated by the radar system may be suppressed or validated based on the absence or presence of motion within a predefined region of interest in the optical data. Upon confirmation of a launch, the trigger is propagated along with associated metadata such as estimated launch position and time.In a further embodiment, optical tracking systems alone may be used to detect a launch event. This may include using data from a single camera or from multiple cameras configured in a stereo arrangement. In the case of multiple cameras, a calibrated stereo vision system may estimate a three-dimensional launch position. Upon detecting motion indicative of a launch, the system generates a trigger signal including launch metadata such as estimated launch time and position.
[0166] In yet another embodiment, a trigger may be generated based on characteristics of the launcher, such as a human player or mechanical launching device. In the absence of external tracking systems, launch detection may rely on analysis of video frames showing the launcher. For example, a machine learning model such as a pose estimation network (e.g., Pose-YOLO or Google's Pose Landmarker) may be used to extract body keypoints from one or more video frames. These keypoints may then be classified using a classifier model (e.g., XGBoost) into states such as "idle," "preparing to swing," and "swinging." A heuristic may then be applied to these classified states to determine whether a launch has occurred. Based on such analysis, a launch trigger may be generated and propagated to the rest of the system.
[0167] In a still other embodiment, a launch event is detected by analyzing audio data associated with the golf shot. This audio may be captured by a microphone, e.g. integrated into the video capturing device, a dedicated microphone positioned on the tee box to record the sound of the swing and ball impact, or any other microphone-equipped device located near the ball's launch position. Further, the microphone can be an example of the launch event detection device 140. By examining the characteristics of the captured audio signal, it is possible to identify potential launch events by detecting transient spikes or other anomalies in either the time or frequency domain. When such indicators are found, a launch trigger can be generated and propagated throughout the system, e.g. sent to the computer 110.
[0168] With reference to Figure 7, a schematic block diagram of embodiments of the computer 110 of Figure 1 is shown.
[0169] The computer 110 can include a processing module 701 for performing the methods described herein. The processing module can be embodied in the form of one or morehardware modules and / or one or more software modules. The term "module" may thus refer to a circuit, a software block or the like according to various embodiments as described below.
[0170] The computer 110 may further include a memory 702. The memory can include, such as contain or store, instructions, e.g., in the form of a computer program 703, which can include computer readable code units.
[0171] According to some embodiments herein, the computer 110 and / or the processing module 701 includes a processing circuit 704 as an exemplifying hardware module, which can include one or more processors. Accordingly, the processing module 701 may be embodied in the form of, or 'realized by', the processing circuit 704. The instructions may be executable by the processing circuit 704, whereby the computer 110 is operative to perform the methods of Figure 2. As another example, the instructions, when executed by the computer 110 and / or the processing circuit 704, may cause the computer 110 to perform the method according to Figure 2. The memory 702 can contain the instructions executable by said processing circuit 704 whereby the computer 110 is operative for performing the method according to Figure 2.
[0172] Figure 7 further illustrates a carrier 705, or program carrier, which provides, such as comprises, mediates, supplies and the like, the computer program 703 as described directly above. The carrier 705 may be one of an electronic signal, an optical signal, a radio signal, a computer readable medium, a non-transitory computer readable medium, and a computer program product, and the like.
[0173] In some embodiments, the computer 110 and / or the processing module 701 may comprise one or more of an obtaining module 710, a determining module 730, an identifying module 730, a ranking module 740, a generating module 750, a verifying module 760, and so on, as exemplifying hardware modules. The term "module" may refer to a circuit when the term "module" refers to a hardware module. In other examples, one or more of the aforementioned exemplifying hardware modules may be implemented as one or more software modules.
[0174] Moreover, the computer 110 and / or the processing module 701 may comprise an Input / Output module 706, which may be exemplified by the receiving module and / or the sending module when applicable. Furthermore, each of the various modules may be separated into one or more further modules, e.g. the receiving module can comprise a firstreceiving module and a second receiving module, and so forth, each being associated with a respective receiving action. The same reasoning can be applied for any one of the other modules herein, i.e. in particular those associated with two or more actions.
[0175] Accordingly, the computer 110 is configured for finding a launch position Pl of a golf ball 101 tracked by a tracking unit 120.
[0176] Therefore, according to the various embodiments described above, the computer 110 and / or the processing module 701 and / or the obtaining module 710 is configured for obtaining, from a video capturing device 130, a set of video frames of a scene in which the golf ball 101 is visible.
[0177] The computer 110 and / or the processing module 701 and / or the obtaining module 710 is configured for obtaining a message indicating a launch event relating to the golf ball 101. The launch event is associated with a launch point in time TO.
[0178] The computer 110 and / or the processing module 701 and / or the determining module 720 is configured for determining a first point in time T1 that occurs before the launch point in time TO and a second point in time T2 that occurs after the launch point in time TO.
[0179] The computer 110 and / or the processing module 701 and / or the identifying module 730 is configured for identifying, within the set of video frames, a first frame Fl at the first point in time T1 and a second frame F2 at the second point in time T2.
[0180] The computer 110 and / or the processing module 701 and / or the determining module 720 is configured for determining a differential frame DF by computing a difference between the first and second frames Fl, F2.
[0181] The computer 110 and / or the processing module 701 and / or the forming module 795 is configured for forming a set of blobs of pixels.
[0182] The computer 110 and / or the processing module 701 and / or the ranking module 740 is configured for ranking the set of blobs by ranking each blob of the set of blobs based on one or more of:
[0183] • a shape of said each blob,
[0184] • a position of said each blob, e.g. in relation to standard rules of golf, in relation to tee markers, or the like,
[0185] • a number of pixels of said each blob,• a blob location difference given by a comparison of a blob distance between said each blob and the video capturing device 130 to a launch area distance between a launch area and the video capturing device 130,
[0186] • the like, or
[0187] • a combination thereof,
[0188] thereby forming a ranked set of blobs,
[0189] The computer 110 and / or the processing module 701 and / or the identifying module 730 is configured for identifying an extremum in the ranked set of blobs as a candidate blob. The computer 110 and / or the processing module 701 and / or the generating module 750 is configured for generating an extrapolated trace 106 by extrapolating a trace 105 obtained by the tracking unit 120 to a position Pl of the candidate blob.
[0190] The computer 110 and / or the processing module 701 and / or the verifying module 760 is configured for verifying that the extrapolated trace 106 fulfils conditions of expected motion of the golf ball, assumed to be freely moving after the launch point in time TO and as constrained by the trace 105.
[0191] The computer 110 and / or the processing module 701 and / or the providing module 770 is configured for providing the position Pl of the candidate blob as the launch position of the golf ball 101.
[0192] The computer 110 and / or the processing module 701 and / or the ranking module 740 can be configured for ranking said each blob by calculating a respective score based on one or more of:
[0193] • the shape of said each blob,
[0194] • the position of said each blob,
[0195] • the number of pixels of said each blob,
[0196] • the blob location difference,
[0197] • the like, or
[0198] • the combination thereof.
[0199] The computer 110 and / or the processing module 701 and / or the calculating module (not shown) may be configured for calculating the respective score as a weighted average of oneor more of, preferably all of, the shape of said each blob, the position of said each blob, the number of pixels of said each blob, and / or the blob location difference.
[0200] The computer 110 and / or the processing module 701 and / or the forming module 795 may be configured for ranking the set of blobs and / or the computer 110 can be configured for ranking of the set of blobs after the forming of the set of blobs.
[0201] In some embodiments, the computer 110 and / or the processing module 701 and / or the generating module 750 is configured for generating the extrapolated trace 106 by backward extrapolating the trace 105 to the position of the candidate blob.
[0202] In some embodiments, the expected motion of the golf ball conforms to physics of projectile motion, wherein preferably the expected motion comprises an expected trace as constrained by classical physics.
[0203] The computer 110 and / or the processing module 701 and / or the obtaining module 710 may be configured for obtaining the message by generating, e.g. by the tracking unit 120 and / or the video capturing device 130 and / or the launch event detecting device 140 and / or the computer 110, the message using one or more of:
[0204] • radar and / or optical detection,
[0205] • image processing on the set of video frames to identify characteristics of launcher when launching the golf ball 101,
[0206] • sound detection, or
[0207] • a combination thereof.
[0208] In some embodiments, the computer 110 is configured for obtaining the message from one or more of the tracking unit 120, the video capturing device 130, and a launch event detecting device 140.
[0209] The computer 110 and / or the processing module 701 and / or the determining module 720 may be configured for determining the first point in time to be a first number of1
[0210] milliseconds, "ms", before the launch point in time. The computer 110 and / or the processing module 701 and / or the determining module 720 may be configured for determining the second point in time to be a second number of ms after the launch point in time. The first number of ms is 100 ms, 200 ms, 500 ms, 1000 ms, or the like. The second number of ms is 100 ms, 200 ms, 500 ms, 1000 ms, or the like.
[0211] The computer 110 and / or the processing module 701 and / or the computing module (not shown) may be configured for computing a pixel-wise or pixel-group-wise difference between the first and second frames Fl, F2.
[0212] Alternatively, the computer 110 and / or the processing module 701 and / or the using module 410 may be configured for using one or more of pixel-wise differencing, block-wise differencing, background subtraction, thresholding, optical flow estimation, or temporal median filtering, or a combination thereof, to determine the differential frame.
[0213] The computer 110 and / or the processing module 701 and / or the providing module 770 may be configured for providing the extrapolated trace 106 including the position Pl.
[0214] The computer 110 and / or the processing module 701 and / or the drawing module 780 may be configured for drawing the extrapolated trace 106 and the trace 105 in video frames received from the video capturing device 130, and / or
[0215] The computer 110 and / or the processing module 701 and / or the providing module 770 may be configured for providing access to the extrapolated trace 106 and / or the trace 105 via an application programming interface.
[0216] In some embodiments, the computer 110 is comprised in a system 100 comprising at least one of the tracking unit 120 and the video capturing device 130.
[0217] As used herein, the term "module" may refer to one or more functional units, each of which may be implemented as one or more hardware modules and / or one or more software modules and / or a combined software / hardware module in a node, such as a physical or virtual node, a computer, or the like. In some examples, the module may represent a functional unit realized as software and / or hardware of the node.
[0218] As used herein, the term "computer program carrier", "program carrier", or "carrier", may refer to one of an electronic signal, an optical signal, a radio signal, and a computer readablemedium. In some examples, the computer program carrier may exclude transitory, propagating signals, such as the electronic, optical and / or radio signal. Thus, in these examples, the computer program carrier may be a non-transitory carrier, such as a non-transitory computer readable medium.
[0219] As used herein, the term "processing unit" may include one or more hardware units, one or more software units or a combination thereof. Any such unit, be it a hardware, software or a combined hardware-software unit, may be a determining means, estimating means, capturing means, associating means, comparing means, identification means, selecting means, receiving means, sending means or the like as disclosed herein. As an example, the expression "means" may be a unit corresponding to the units listed above in conjunction with the Figures.
[0220] As used herein, the term "software module" may refer to a software application, a Dynamic Link Library (DLL), a software component, a software routine or procedure, a software object, a React component, an object according to Component Object Model (COM), a software function, a software engine, an executable binary software file or the like.
[0221] The terms "processing module" or "processing circuit" may herein comprise one or more processors, an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA) or the like. The processing circuit or the like may comprise one or more processor kernels.
[0222] As used herein, the expression "configured to / for" can refer to that a processing circuit can be configured to, such as adapted to or operative to, by means of software configuration and / or hardware configuration, perform one or more of the actions described herein.
[0223] As used herein, the term "action" may refer to an action, a step, an operation, a response, a reaction, an activity or the like. It shall be noted that an action herein may be split into two or more sub-actions as applicable. Moreover, also as applicable, it shall be noted that two or more of the actions described herein may be merged into a single action.As used herein, the term "memory" may refer to a hard disk, a magnetic storage medium, a portable computer diskette or disc, flash memory, random access memory (RAM) or the like. Furthermore, the term "memory" may refer to an internal register memory of a processor or the like.
[0224] As used herein, the term "computer readable medium" may be a Universal Serial Bus (USB) memory, a Digital Versatile Disc (DVD), a Blu-ray disc, a software unit that is received as a stream of data, a Flash memory, a hard drive, a memory card, such as a MemoryStick, a Multimedia Card (MMC), Secure Digital (SD) card, etc. One or more of the aforementioned examples of computer readable medium may be provided as one or more computer program products.
[0225] As used herein, the term "computer readable code units" may be text of a computer program, parts of or an entire binary file representing a computer program in a compiled format or anything there between.
[0226] Any feature disclosed for one example and / or embodiment may be combined with one or more features disclosed for one or more other examples and / or embodiments without departing from the scope herein.
Claims
CLAIMS1. A method, performed by a computer (110), for finding a launch position (Pl) of a golf ball (101) tracked by a tracking unit (120), wherein the method comprises: receiving (A110), by the computer (110) and from a video capturing device (130), a set of video frames of a scene in which the golf ball (101) is visible,obtaining (A120) a message indicating a launch event relating to the golf ball (101), wherein the launch event is associated with a launch point in time (TO), determining (A130) a first point in time (Tl) that occurs before the launch point in time (TO) and a second point in time (T2) that occurs after the launch point in time (TO),identifying (A140), within the set of video frames, a first frame (Fl) at the first point in time (Tl) and a second frame (F2) at the second point in time (T2), determining (A150) a differential frame (DF) by computing a difference between the first and second frames (Fl, F2),forming (A160) a set of blobs of pixels in the differential frame (DF),ranking (A170) the set of blobs by ranking (A172) each blob of the set of blobs based on one or more of:• a shape of said each blob,• a position of said each blob,• a number of pixels of said each blob,• a blob location difference given by a comparison of a blob distance between said each blob and the video capturing device (130) to a launch area distance between a launch area and the video capturing device (130), or • a combination thereof,thereby forming a ranked set of blobs, wherein the ranking (A172) of said each blob is performed by calculating (A174) a respective score based on one or more of:o the shape of said each blob,o the position of said each blob,o the number of pixels of said each blob,o the blob location difference, oro the combination thereof,identifying (A180) an extremum in the ranked set of blobs as a candidate blob, generating (A190) an extrapolated trace (106) by extrapolating a trace (105) obtained by the tracking unit (120) to a position (Pl) of the candidate blob,verifying (A200) that the extrapolated trace (106) fulfils conditions of expected motion of the golf ball, assumed to be freely moving after the launch point in time (TO) and as constrained by the trace (105), andproviding (A210) the position (Pl) of the candidate blob as the launch position of the golf ball (101).
2. The method according to the preceding claim, wherein the calculating (A174) of the respective score comprises calculating the respective score as a weighted average of one or more of, preferably all of, the shape of said each blob, the position of said each blob, the number of pixels of said each blob, and the blob location difference.
3. The method according to any one of the preceding claims, wherein the forming (A160) of the set of blobs comprises the ranking (A170) of the set of blobs and / or the ranking (A170) of the set of blobs is performed after the forming (A160) of the set of blobs.
4. The method according to any one of the preceding claims, wherein the generating (A190) of the extrapolated trace (106) comprises backward extrapolating the trace (105) to the position of the candidate blob.
5. The method according to any one of the preceding claims, wherein the expected motion of the golf ball conforms to physics of projectile motion, wherein preferably the expected motion comprises an expected trace as constrained by classical physics.
6. The method according to any one of the preceding claims, wherein the message is obtained from one or more of the tracking unit (120), the video capturing device (130), and a launch event detecting device (140).
7. The method according to the preceding claim, wherein the obtaining (A120) of the message comprises generating the message using one or more of:• radar and / or optical detection,• image processing on the set of video frames to identify characteristics of launcher when launching the golf ball (101),• sound detection, or• a combination thereof.
8. The method according to any one of the preceding claims, wherein the determining (A130) of the first point in time (Tl) and the second point in time (T2) comprises determining the first point in time (Tl) to be a first number of milliseconds, "ms", before the launch point in time and determining the second point in time (T2) to be a second number of ms after the launch point in time.
9. The method according to any one of the preceding claims, wherein the determining (A150) of the differential frame comprises computing a pixel-wise or pixel-group-wise difference between the first and second frames (Fl, F2), or wherein the determining of the differential frame comprises using one or more of pixel-wise differencing, block-wise differencing, background subtraction, thresholding, optical flow estimation, or temporal median filtering, or a combination thereof, to determine the differential frame.
10. The method according to any one of the preceding claims, wherein the providing (A210) of the position (Pl) comprises providing (A215) the extrapolated trace (106) including the position (Pl).
11. The method according to the preceding claim, wherein the providing (A215) of the extrapolated trace (106) comprises:drawing (A217) the extrapolated trace (106) and the trace (105) in video frames received from the video capturing device (130), and / orproviding (A219) access to the extrapolated trace (106) and / or the trace (105) via an application programming interface.
12. The method according to any one of the preceding claims, wherein the method is performed by a system (100) comprising the computer (110) and at least one of the tracking unit (120) and the video capturing device (130).
13. A computer (110) configured for finding a launch position (Pl) of a golf ball (101) tracked by a tracking unit (120), wherein the computer (110) is configured for: receiving, by the computer (1109 and from a video capturing device (130), a set of video frames of a scene in which the golf ball (101) is visible,obtaining a message indicating a launch event relating to the golf ball (101), wherein the launch event is associated with a launch point in time (TO),determining a first point in time (Tl) that occurs before the launch point in time (TO) and a second point in time (T2) that occurs after the launch point in time (TO), identifying, within the set of video frames, a first frame (Fl) at the first point in time (Tl) and a second frame (F2) at the second point in time (T2),determining a differential frame (DF) by computing a difference between the first and second frames (Fl, F2),forming a set of blobs of pixels in the differential frame (DF),ranking the set of blobs by ranking each blob of the set of blobs based on one or more of:• a shape of said each blob,• a position of said each blob,• a number of pixels of said each blob,• a blob location difference given by a comparison of a blob distance between said each blob and the video capturing device (130) to a launch area distance between a launch area and the video capturing device (130), or• a combination thereof,thereby forming a ranked set of blobs, wherein the computer (110) is configured to perform the ranking of said each blob by calculating a respective score based on one or more of:• the shape of said each blob,• the position of said each blob,• the number of pixels of said each blob,• the blob location difference, or• the combination thereof,identifying an extremum in the ranked set of blobs as a candidate blob, generating an extrapolated trace (106) by extrapolating a trace (105) obtained by the tracking unit (120) to a position (Pl) of the candidate blob,verifying that the extrapolated trace (106) fulfils conditions of expected motion of the golf ball (101), assumed to be freely moving after the launch point in time (TO) and as constrained by the trace (105), andproviding the position (Pl) of the candidate blob as the launch position of the golf ball (101).
